SilhouetteMatch

Introduction

The SilhouetteMatch module is an optional on-board module of the rc_visard and requires a separate SilhouetteMatch license to be purchased.

The module detects objects by matching a predefined silhouette (“template”) to edges in an image.

For the SilhouetteMatch module to work, special object templates are required for each type of object to be detected. Roboception offers a template generation service on their website, where the user can upload CAD files or recorded data of the objects and request object templates for the SilhouetteMatch module.

The object templates consist of significant edges of each object. These template edges are matched to the edges detected in the left and right camera images, considering the actual size of the objects and their distance from the camera. The poses of the detected objects are returned and can be used for grasping, for example.

The SilhouetteMatch module offers:

  • A dedicated page on the rc_visard Web GUI for easy setup, configuration, testing, and application tuning.
  • A REST-API interface and a KUKA Ethernet KRL Interface.
  • The definition of 2D regions of interest to select relevant parts of the camera image (see Setting a region of interest).
  • A load carrier detection functionality for bin-picking applications (see LoadCarrier), to provide grasps for objects inside a bin only.
  • Storing of up to 50 templates.
  • The definition of up to 50 grasp points for each template via an interactive visualization in the Web GUI.
  • Support for static and robot-mounted cameras and optional integration with the Hand-eye calibration module, to provide grasps in the user-configured external reference frame.
  • Selection of a sorting strategy to sort the detected objects and returned grasps.
  • 3D visualization of the detection results with grasp points and gripper animations in the Web GUI.

Suitable objects

The SilhouetteMatch module is intended for objects which have significant edges on a common plane that is parallel to the base plane on which the objects are placed. This applies to flat, nontransparent objects, such as routed, laser-cut or water-cut 2D parts and flat-machined parts. More complex parts can also be detected if there are significant edges on a common plane, e.g. a special pattern printed on a flat surface.

The SilhouetteMatch module works best for objects on a texture-free base plane. The color of the base plane should be chosen such that a clear contrast between the objects and the base plane appears in the intensity image.

Suitable scene

The scene must meet the following conditions to be suitable for the SilhouetteMatch module:

  • The objects to be detected must be suitable for the SilhouetteMatch module as described above.
  • Only objects belonging to one specific template are visible at a time (unmixed scenario). In case other objects are visible as well, a proper region of interest (ROI) must be set.
  • All visible objects are lying on a common base plane, which has to be calibrated.
  • The offset between the base plane normal and the camera’s line of sight does not exceed 10 degrees.
  • The objects are not partially or fully occluded.
  • All visible objects are right side up (no flipped objects).
  • The object edges to be matched are visible in both, left and right camera images.

Base-plane calibration

Before objects can be detected, a base-plane calibration must be performed. Thereby, the distance and angle of the plane on which the objects are placed is measured and stored persistently on the rc_visard.

Separating the detection of the base plane from the actual object detection renders scenarios possible in which the base plane is temporarily occluded. Moreover, it increases performance of the object detection for scenarios where the base plane is fixed for a certain time; thus, it is not necessary to continuously re-detect the base plane.

The base-plane calibration can be performed in three different ways, which will be explained in more detail further down:

  • AprilTag based
  • Stereo based
  • Manual

The base-plane calibration is successful if the normal vector of the estimated base plane is at most 10 degrees offset to the camera’s line of sight. If the base-plane calibration is successful, it will be stored persistently on the rc_visard until it is removed or a new base-plane calibration is performed.

Note

To avoid privacy issues, the image of the persistently stored base-plane calibration will appear blurred after rebooting the rc_visard.

In scenarios where the base plane is not accessible for calibration, a plane parallel to the base-plane can be calibrated. Then an offset parameter can be used to shift the estimated plane onto the actual base plane where the objects are placed. The offset parameter gives the distance in meters by which the estimated plane is shifted towards the camera.

In the REST-API, a plane is defined by a normal and a distance. normal is a normalized 3-vector, specifying the normal of the plane. The normal points away from the camera. distance represents the distance of the plane from the camera along the normal. Normal and distance can also be interpreted as \(a\), \(b\), \(c\), and \(d\) components of the plane equation, respectively:

\[ax + by + cz + d = 0\]

AprilTag based base-plane calibration

AprilTag detection (ref. TagDetect) is used to find AprilTags in the scene and fit a plane through them. At least three AprilTags must be placed on the base plane so that they are visible in the left and right camera images. The tags should be placed such that they are spanning a triangle that is as large as possible. The larger the triangle, the more accurate is the resulting base-plane estimate. Use this method if the base plane is untextured and no external random dot projector is available. This calibration mode is available via the REST-API interface and the rc_visard Web GUI.

Stereo based base-plane calibration

The 3D point cloud computed by the stereo matching module is used to fit a plane through its 3D points. Therefore, the region of interest (ROI) for this method must be set such that only the relevant base plane is included. The plane_preference parameter allows to select whether the plane closest to or farthest from the camera should be used as base plane. Selecting the closest plane can be used in scenarios where the base plane is completely occluded by objects or not accessible for calibration. Use this method if the base plane is well textured or you can make use of a random dot projector to project texture on the base plane. This calibration mode is available via the REST-API interface and the rc_visard Web GUI.

Manual base-plane calibration

The base plane can be set manually if its parameters are known, e.g. from previous calibrations. This calibration mode is only available via the REST-API interface and not the rc_visard Web GUI.

Setting a region of interest

If objects are to be detected only in part of the camera’s field of view, a 2D region of interest (ROI) can be set accordingly as described in Region of interest.

Setting of grasp points

To use SilhouetteMatch directly in a robot application, grasp points can be defined for each template. A grasp point represents the desired position and orientation of the robot’s TCP (Tool Center Point) to grasp an object as shown in Fig. 35.

_images/grasp_points_silm.svg

Fig. 35 Definition of grasp points with respect to the robot’s TCP

Each grasp consists of an id which must be unique within all grasps for an object template, the template_id representing the template to which the grasp should be attached, and the pose in the coordinate frame of the object template. Grasp points can be set via the REST-API interface, or by using the interactive visualization in the Web GUI. The rc_visard can store up to 50 grasp points per template.

When a grasp is defined on a symmetric object, all grasps symmetric to the defined one will automatically be considered in the SilhouetteMatch module’s detect_object service call. Symmetric grasps for a given grasp point can be retrieved using the get_symmetric_grasps service call and visualized in the Web GUI.

Setting grasp points in the Web GUI

The rc_visard Web GUI provides an intuitive and interactive way of defining grasp points for object templates. In a first step, the object template has to be uploaded to the rc_visard. This can be done in the Web GUI in any pipeline under Modules ‣ SilhouetteMatch by clicking on + Add a new Template in the Templates and Grasps section, or in Database ‣ Templates in the SilhouetteMatch Templates and Grasps section. Once the template upload is complete, a dialog with a 3D visualization of the object template is shown for adding or editing grasp points. The same dialog appears when editing an existing template. If the template contains a collision model or a visualization model, this 3D model is visualized as well.

This dialog provides two ways for setting grasp points:

  1. Adding grasps manually: By clicking on the + symbol, a new grasp is placed in the object origin. The grasp can be given a unique name which corresponds to its ID. The desired pose of the grasp can be entered in the fields for Position and Roll/Pitch/Yaw which are given in the coordinate frame of the object template represented by the long x, y and z axes in the visualization. The grasp point can be placed freely with respect to the object template - inside, outside or on the surface. The grasp point and its orientation are visualized in 3D for verification.
  2. Adding grasps interactively: Grasp points can be added interactively by first clicking on the Add Grasp button in the upper right corner of the visualization and then clicking on the desired point on the object template visualization. If the 3D model is displayed, the grasps will be attached to the surface of the 3D model. Otherwise, the grasp is attached to the template surface. The grasp orientation is a right-handed coordinate system and is chosen such that its z axis is perpendicular to the surface pointing inside the template at the grasp position. The position and orientation in the object coordinate frame is displayed on the right. The position and orientation of the grasp can also be changed interactively. In case Snap to surface is enabled in the visualization (default), the grasp can be moved along the template or model surface by clicking on the Translate button in the visualization and then clicking on the grasp point and dragging it to the desired position. The orientation of the grasp around the surface normal can also be changed by choosing Rotate and then rotating the grasp with the cursor. In case Snap to surface is disabled, the grasp can be translated and rotated freely in all three dimensions.

If the object template has symmetries, the grasps which are symmetric to the defined grasps can be displayed by clicking on show symmetric grasps.

Setting grasp points via the REST-API

Grasp points can be set via the REST-API interface using the set_grasp or set_all_grasps services (see Services). In the SilhouetteMatch module a grasp consists of the template_id of the template to which the grasp should be attached, an id uniquely identifying the grasp point and the pose. The pose is given in the coordinate frame of the object template and consists of a position in meters and an orientation as quaternion.

Setting the preferred orientation of the TCP

The SilhouetteMatch module determines the reachability of grasp points based on the preferred orientation of the gripper or TCP. The preferred orientation can be set via the set_preferred_orientation service or on the SilhouetteMatch page in the Web GUI. The resulting direction of the TCP’s z axis is used to reject grasps which cannot be reached by the gripper. Furthermore, the preferred orientation can be used to sort the reachable grasps by setting the corresponding sorting strategy.

The preferred orientation can be set in the camera coordinate frame or in the external coordinate frame, in case a hand-eye calibration is available. If the preferred orientation is specified in the external coordinate frame and the sensor is robot mounted, the current robot pose has to be given to each object detection call, so that the preferred orientation can be used for filtering and, optionally, sorting the grasps on the detected objects. If no preferred orientation is set, the orientation of the left camera is used as the preferred orientation of the TCP.

Setting the sorting strategies

The objects and grasps returned by the detect_object service call are sorted according to a sorting strategy which can be chosen by the user. The following sorting strategies are available and can be set in the Web GUI or using the set_sorting_strategies service call:

  • preferred_orientation: objects and grasp points with minimal rotation difference between their orientation and the preferred orientation of the TCP are returned first,
  • direction: objects and grasp points with the shortest distance along a defined direction vector in a given pose_frame are returned first.

If no sorting strategy is set or default sorting is chosen in the Web GUI, sorting is done based on a combination of preferred_orientation and the minimal distance from the camera along the z axis of the preferred orientation of the TCP.

Detection of objects

Objects can only be detected after a successful base-plane calibration. It must be ensured that the position and orientation of the base plane does not change before the detection of objects. Otherwise, the base-plane calibration must be renewed.

For triggering the object detection, in general, the following information must be provided to the SilhouetteMatch module:

  • The template of the object to be detected in the scene.
  • The coordinate frame in which the poses of the detected objects shall be returned (ref. Hand-eye calibration).

Optionally, further information can be given to the SilhouetteMatch module:

  • An offset in case the objects are lying not on the base plane but on a plane parallel to it. The offset is the distance between both planes given in the direction towards the camera. If omitted, an offset of 0 is assumed.
  • The ID of the load carrier which contains the objects to be detected.
  • The ID of the region of interest where to search for the load carrier if a load carrier is set. Otherwise, the ID of the region of interest where the objects should be detected. If omitted, objects are matched in the whole image.
  • The current robot pose in case the camera is mounted on the robot and the chosen coordinate frame for the poses is external or the preferred orientation is given in the external frame.
  • Collision detection information: The ID of the gripper to enable collision checking and optionally a pre-grasp offset to define a pre-grasp position. Details on collision checking are given below in CollisionCheck.

On the Web GUI the detection can be tested in the Try Out section of the SilhouetteMatch page. The result is visualized as shown in Fig. 36.

_images/silhouetteMatchResults_en.png

Fig. 36 Result image of the SilhouetteMatch module as shown in the Web GUI

The left image shows the calibrated base plane in blue and the template to be matched in red with the defined grasp points in green (see Setting of grasp points). The template is warped to the size and tilt matching objects on the calibrated base plane would have.

The right image shows the detection result. The shaded blue area on the left is the region of the left camera image which does not overlap with the right image, and in which no objects can be detected. The chosen region of interest is shown as bold petrol rectangle. The detected edges in the image are shown in light blue and the matches with the template (instances) are shown in red. The blue circles are the origins of the detected objects as defined in the template and the green circles are the reachable grasp points. Unreachable grasp points will be visualized as red dots (not shown in the figure).

The poses of the object origins in the chosen coordinate frame are returned as results. The orientation of the detected objects is aligned with the normal of the base plane. If the chosen template also has grasp points attached, a list of grasps for all objects is returned in addition to the list of detected objects. The grasp poses are given in the desired coordinate frame and the grasps are sorted according to the selected sorting strategy (see Setting the sorting strategies). There are references between the detected object instances and the grasps via their uuids.

In case the templates have a continuous rotational symmetry (e.g. cylindrical objects), all returned object poses will have the same orientation. Furthermore, all grasps symmetric to each grasp point on an object are checked for reachability and collisions, and only the best one according to the given sorting strategy is returned.

For objects with a discrete symmetry (e.g. prismatic objects), all collision-free symmetries of each grasp point which are reachable according to the given preferred TCP orientation are returned, ordered by the given sorting strategy.

The detection results and runtimes are affected by several run-time parameters which are listed and explained further down. Improper parameters can lead to timeouts of the SilhouetteMatch module’s detection process.

Interaction with other modules

Internally, the SilhouetteMatch module depends on, and interacts with other on-board modules as listed below.

Note

All changes and configuration updates to these modules will affect the performance of the SilhouetteMatch module.

Stereo camera and stereo matching

The SilhouetteMatch module makes internally use of the rectified images from the Camera module (rc_camera). Thus, the exposure time should be set properly to achieve the optimal performance of the module.

For base-plane calibration in stereo mode the disparity images from the Stereo matching module (rc_stereomatching) are used. Apart from that, the stereo-matching module should not be run in parallel to the SilhouetteMatch module, because the detection runtime increases.

For best results it is recommended to enable smoothing for Stereo matching.

Estimation of gravity vector

For each object detection inside a load carrier, the module estimates the gravity vector as described in Estimation of gravity vector.

IO and Projector Control

In case the rc_visard is used in conjunction with an external random dot projector and the IO and Projector Control module (rc_iocontrol), the projector should be used for the stereo-based base-plane calibration.

The projected pattern must not be visible in the left and right camera images during object detection as it interferes with the matching process. Therefore, it must either be switched off or operated in ExposureAlternateActive mode.

Hand-eye calibration

In case the camera has been calibrated to a robot, the SilhouetteMatch module can automatically provide poses in the robot coordinate frame. For the SilhouetteMatch node’s Services, the frame of the input and output poses and plane coordinates can be controlled with the pose_frame argument.

Two different pose_frame values can be chosen:

  1. Camera frame (camera). All poses and plane coordinates provided to and by the module are in the camera frame.
  2. External frame (external). All poses and plane coordinates provided to and by the module are in the external frame, configured by the user during the hand-eye calibration process. The module relies on the on-board Hand-eye calibration module to retrieve the camera mounting (static or robot mounted) and the hand-eye transformation. If the sensor mounting is static, no further information is needed. If the sensor is robot-mounted, the robot_pose is required to transform poses to and from the external frame.

All pose_frame values that are not camera or external are rejected.

Note

If no hand-eye calibration is available, all pose_frame values should be set to camera.

Note

If the hand-eye calibration has changed after base-plane calibration, the base-plane calibration will be marked as invalid and must be renewed.

If the sensor is robot-mounted, the current robot_pose has to be provided depending on the value of pose_frame and the definition of the preferred TCP orientation:

  • If pose_frame is set to external, providing the robot pose is obligatory.
  • If the preferred TCP orientation is defined in external, providing the robot pose is obligatory.
  • If pose_frame is set to camera and the preferred TCP orientation is defined in camera, providing the robot pose is optional.

If the current robot pose is provided during calibration, it is stored persistently on the rc_visard. If the updated robot pose is later provided during get_base_plane_calibration or detect_object as well, the base-plane calibration will be transformed automatically to this new robot pose. This enables the user to change the robot pose (and thus camera position) between base-plane calibration and object detection.

Note

Object detection can only be performed if the limit of 10 degrees angle offset between the base plane normal and the camera’s line of sight is not exceeded.

CollisionCheck

Collision checking can be easily enabled for grasp computation of the SilhouetteMatch module by passing a collision_detection argument to the detect_object service call. It contains the ID of the used gripper and optionally a pre-grasp offset. The gripper has to be defined in the GripperDB module (see Setting a gripper) and details about collision checking are given in Collision checking within other modules. In addition to collision checking between the gripper and the detected load carrier, collisions between the gripper and the calibrated base plane will be checked, if the run-time parameter check_collisions_with_base_plane is true. If the selected SilhouetteMatch template contains a collision model and the run-time parameter check_collisions_with_matches is true, also collisions between the gripper and all other detected objects (not limited to max_number_of_detected_objects) will be checked. The object on which the grasp point to be checked is located, is excluded from the collision check.

If collision checking is enabled, only grasps which are collision free will be returned. However, the visualization images on the SilhouetteMatch page of the Web GUI also show colliding grasp points in red. The objects which are considered in the collision check are also visualized with their edges in red.

The CollisionCheck module’s run-time parameters affect the collision detection as described in CollisionCheck Parameters.

Parameters

The SilhouetteMatch software module is called rc_silhouettematch in the REST-API and is represented in the Web GUI under Modules ‣ SilhouetteMatch. The user can explore and configure the rc_silhouettematch module’s run-time parameters, e.g. for development and testing, using the Web GUI or the REST-API interface.

Parameter overview

This module offers the following run-time parameters:

Table 38 The rc_silhouettematch module’s run-time parameters
Name Type Min Max Default Description
check_collisions_with_-base_plane bool false true true Whether to check for collisions between gripper and base plane
check_collisions_with_matches bool false true true Whether to check for collisions between gripper and detected matches
edge_sensitivity float64 0.1 1.0 0.6 Sensitivity of the edge detector
match_max_distance float64 0.1 10.0 2.5 Maximum allowed distance in pixels between the template and the detected edges in the image
match_percentile float64 0.7 1.0 0.85 Percentage of template pixels that must be within the maximum distance to successfully match the template
max_number_of_detected_objects int32 1 20 10 Maximum number of detected objects
quality string - - High Quality: [Low, Medium, High]

Description of run-time parameters

Each run-time parameter is represented by a row on the Web GUI’s SilhouetteMatch page. The name in the Web GUI is given in brackets behind the parameter name and the parameters are listed in the order they appear in the Web GUI:

max_number_of_detected_objects (Maximum Object Number)

This parameter gives the maximum number of objects to detect in the scene. If more than the given number of objects can be detected in the scene, only the objects matching best to the given sorting strategy are returned.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?max_number_of_detected_objects=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?max_number_of_detected_objects=<value>

quality (Quality)

Object detection can be performed on images with different resolutions: High (full image resolution), Medium (half image resolution) and Low (quarter image resolution). The lower the resolution, the lower the detection time, but the fewer details of the objects are visible.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?quality=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?quality=<value>

match_max_distance (Maximum Matching Distance)

This parameter gives the maximum allowed pixel distance of an image edge pixel from the object edge pixel in the template to be still considered as matching. If the object is not perfectly represented in the template, it might not be detected when this parameter is low. High values, however, might lead to false detections in case of a cluttered scene or the presence of similar objects, and also increase runtime.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?match_max_distance=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?match_max_distance=<value>

match_percentile (Matching Percentile)

This parameter indicates how strict the matching process should be. The matching percentile is the ratio of template pixels that must be within the Maximum Matching Distance to successfully match the template. The higher this number, the more accurate the match must be to be considered as valid.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?match_percentile=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?match_percentile=<value>

edge_sensitivity (Edge Sensitivity)

This parameter influences how many edges are detected in the camera images. The higher this number, the more edges are found in the intensity image. That means, for lower numbers, only the most significant edges are considered for template matching. A large number of edges in the image might increase the detection time.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?edge_sensitivity=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?edge_sensitivity=<value>

check_collisions_with_base_plane (Check Collisions with Base Plane)

If this parameter is set to true, and collision checking is enabled by passing a gripper to the detect_object service call, all grasp points will be checked for collisions between the gripper and the calibrated base plane, and only grasp points at which the gripper would not collide with the base plane will be returned.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?check_collisions_with_base_plane=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?check_collisions_with_base_plane=<value>

check_collisions_with_matches (Check Collisions with Matches)

If this parameter is set to true, and collision checking is enabled by passing a gripper to the detect_object service call, all grasp points will be checked for collisions between the gripper and all other detected objects (not limited to max_number_of_detected_objects), and only grasp points at which the gripper would not collide with any other detected object will be returned.

Via the REST-API, this parameter can be set as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/parameters/parameters?check_collisions_with_matches=<value>
PUT http://<host>/api/v1/nodes/rc_silhouettematch/parameters?check_collisions_with_matches=<value>

Status values

This module reports the following status values:

Table 39 The rc_silhouettematch module’s status values
Name Description
data_acquisition_time Time in seconds required by the last active service to acquire images
last_timestamp_processed The timestamp of the last processed dataset
load_carrier_detection_time Processing time of the last load carrier detection in seconds
processing_time Processing time of the last detection (including load carrier detection) in seconds

Services

The user can explore and call the rc_silhouettematch module’s services, e.g. for development and testing, using the REST-API interface or the rc_visard Web GUI.

The SilhouetteMatch module offers the following services.

detect_object

Triggers an object detection as described in Detection of objects and returns the pose of all found object instances.

Details

All images used by the service are guaranteed to be newer than the service trigger time.

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/detect_object
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/detect_object

Required arguments:

object_id in object_to_detect: ID of the template which should be detected.

pose_frame: see Hand-eye calibration.

Potentially required arguments:

robot_pose: see Hand-eye calibration.

Optional arguments:

offset: offset in meters by which the base-plane calibration will be shifted towards the camera.

load_carrier_id: ID of the load carrier which contains the items to be detected.

collision_detection: see Collision checking within other modules.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "collision_detection": {
      "gripper_id": "string",
      "pre_grasp_offset": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      }
    },
    "load_carrier_id": "string",
    "object_to_detect": {
      "object_id": "string",
      "region_of_interest_2d_id": "string"
    },
    "offset": "float64",
    "pose_frame": "string",
    "robot_pose": {
      "orientation": {
        "w": "float64",
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "position": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      }
    }
  }
}

The maximum number of returned instances can be controlled with the max_number_of_detected_objects parameter.

object_id: ID of the detected template.

instances: list of detected object instances, ordered according to the chosen sorting strategy.

grasps: list of grasps on the detected objects, ordered according to the chosen sorting strategy. The instance_uuid gives the reference to the detected object in instances this grasp belongs to.

load_carriers: list of detected load carriers.

timestamp: timestamp of the image set the detection ran on.

return_code: holds possible warnings or error codes and messages.

The definition for the response with corresponding datatypes is:

{
  "name": "detect_object",
  "response": {
    "grasps": [
      {
        "id": "string",
        "instance_uuid": "string",
        "pose": {
          "orientation": {
            "w": "float64",
            "x": "float64",
            "y": "float64",
            "z": "float64"
          },
          "position": {
            "x": "float64",
            "y": "float64",
            "z": "float64"
          }
        },
        "pose_frame": "string",
        "timestamp": {
          "nsec": "int32",
          "sec": "int32"
        },
        "uuid": "string"
      }
    ],
    "instances": [
      {
        "grasp_uuids": [
          "string"
        ],
        "id": "string",
        "object_id": "string",
        "pose": {
          "orientation": {
            "w": "float64",
            "x": "float64",
            "y": "float64",
            "z": "float64"
          },
          "position": {
            "x": "float64",
            "y": "float64",
            "z": "float64"
          }
        },
        "pose_frame": "string",
        "timestamp": {
          "nsec": "int32",
          "sec": "int32"
        },
        "uuid": "string"
      }
    ],
    "load_carriers": [
      {
        "id": "string",
        "inner_dimensions": {
          "x": "float64",
          "y": "float64",
          "z": "float64"
        },
        "outer_dimensions": {
          "x": "float64",
          "y": "float64",
          "z": "float64"
        },
        "overfilled": "bool",
        "pose": {
          "orientation": {
            "w": "float64",
            "x": "float64",
            "y": "float64",
            "z": "float64"
          },
          "position": {
            "x": "float64",
            "y": "float64",
            "z": "float64"
          }
        },
        "pose_frame": "string",
        "rim_ledge": {
          "x": "float64",
          "y": "float64"
        },
        "rim_step_height": "float64",
        "rim_thickness": {
          "x": "float64",
          "y": "float64"
        },
        "type": "string"
      }
    ],
    "object_id": "string",
    "return_code": {
      "message": "string",
      "value": "int16"
    },
    "timestamp": {
      "nsec": "int32",
      "sec": "int32"
    }
  }
}

calibrate_base_plane

Triggers the calibration of the base plane, as described in Base-plane calibration.

Details

A successful base-plane calibration is stored persistently on the rc_visard and returned by this service. The base-plane calibration is persistent over firmware updates and rollbacks.

All images used by the service are guaranteed to be newer than the service trigger time.

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/calibrate_base_plane
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/calibrate_base_plane

Required arguments:

plane_estimation_method: method to use for base-plane calibration. Valid values are STEREO, APRILTAG, MANUAL.

pose_frame: see Hand-eye calibration.

Potentially required arguments:

plane if plane_estimation_method is MANUAL: plane that will be set as base-plane calibration.

robot_pose: see Hand-eye calibration.

region_of_interest_2d_id: ID of the region of interest for base-plane calibration.

Optional arguments:

offset: offset in meters by which the estimated plane will be shifted towards the camera.

plane_preference in stereo: whether the plane closest to or farthest from the camera should be used as base plane. This option can be set only if plane_estimation_method is STEREO. Valid values are CLOSEST and FARTHEST. If not set, the default is FARTHEST.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "offset": "float64",
    "plane": {
      "distance": "float64",
      "normal": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      }
    },
    "plane_estimation_method": "string",
    "pose_frame": "string",
    "region_of_interest_2d_id": "string",
    "robot_pose": {
      "orientation": {
        "w": "float64",
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "position": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      }
    },
    "stereo": {
      "plane_preference": "string"
    }
  }
}

plane: calibrated base plane.

timestamp: timestamp of the image set the calibration ran on.

return_code: holds possible warnings or error codes and messages.

The definition for the response with corresponding datatypes is:

{
  "name": "calibrate_base_plane",
  "response": {
    "plane": {
      "distance": "float64",
      "normal": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "pose_frame": "string"
    },
    "return_code": {
      "message": "string",
      "value": "int16"
    },
    "timestamp": {
      "nsec": "int32",
      "sec": "int32"
    }
  }
}

get_base_plane_calibration

Returns the configured base-plane calibration.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/get_base_plane_calibration
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/get_base_plane_calibration

Required arguments:

pose_frame: see Hand-eye calibration.

Potentially required arguments:

robot_pose: see Hand-eye calibration.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "pose_frame": "string",
    "robot_pose": {
      "orientation": {
        "w": "float64",
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "position": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      }
    }
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "get_base_plane_calibration",
  "response": {
    "plane": {
      "distance": "float64",
      "normal": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "pose_frame": "string"
    },
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

delete_base_plane_calibration

Deletes the configured base-plane calibration.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/delete_base_plane_calibration
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/delete_base_plane_calibration
This service has no arguments.

The definition for the response with corresponding datatypes is:

{
  "name": "delete_base_plane_calibration",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

set_preferred_orientation

Persistently stores the preferred orientation of the gripper to compute the reachability of the grasps, which is used for filtering and, optionally, sorting the grasps returned by the detect_object service (see Setting the preferred orientation of the TCP).

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/set_preferred_orientation
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/set_preferred_orientation

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "orientation": {
      "w": "float64",
      "x": "float64",
      "y": "float64",
      "z": "float64"
    },
    "pose_frame": "string"
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "set_preferred_orientation",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

get_preferred_orientation

Returns the preferred orientation of the gripper to compute the reachability of the grasps, which is used for filtering and, optionally, sorting the grasps returned by the detect_object service (see Setting the preferred orientation of the TCP).

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/get_preferred_orientation
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/get_preferred_orientation
This service has no arguments.

The definition for the response with corresponding datatypes is:

{
  "name": "get_preferred_orientation",
  "response": {
    "orientation": {
      "w": "float64",
      "x": "float64",
      "y": "float64",
      "z": "float64"
    },
    "pose_frame": "string",
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

set_sorting_strategies

Persistently stores the sorting strategy for sorting the grasps and detected objects returned by the detect_object service (see Detection of objects).

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/set_sorting_strategies
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/set_sorting_strategies

Only one strategy may have a weight greater than 0. If all weight values are set to 0, the module will use the default sorting strategy.

If the weight for direction is set, the vector must contain the direction vector and pose_frame must be either camera or external.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "direction": {
      "pose_frame": "string",
      "vector": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "weight": "float64"
    },
    "preferred_orientation": {
      "weight": "float64"
    }
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "set_sorting_strategies",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

get_sorting_strategies

Returns the sorting strategy for sorting the grasps and detected objects returned by the detect_object service (see Detection of objects).

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/get_sorting_strategies
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/get_sorting_strategies
This service has no arguments.

All weight values are 0 when the module uses the default sorting strategy.

The definition for the response with corresponding datatypes is:

{
  "name": "get_sorting_strategies",
  "response": {
    "direction": {
      "pose_frame": "string",
      "vector": {
        "x": "float64",
        "y": "float64",
        "z": "float64"
      },
      "weight": "float64"
    },
    "preferred_orientation": {
      "weight": "float64"
    },
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

reset_defaults

Resets all parameters of the module to its default values, as listed in above table. The reset does not apply to templates, base-plane calibration, preferred orientation and sorting strategies.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/reset_defaults
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/reset_defaults
This service has no arguments.

The definition for the response with corresponding datatypes is:

{
  "name": "reset_defaults",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

set_region_of_interest_2d (deprecated)

Persistently stores a 2D region of interest on the rc_visard.

Details

This service can be called as follows.

This service is not available in API version 2. Use set_region_of_interest_2d in rc_roi_db instead.
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/set_region_of_interest_2d

get_regions_of_interest_2d (deprecated)

Returns the configured 2D regions of interest with the requested region_of_interest_2d_ids.

Details

This service can be called as follows.

This service is not available in API version 2. Use get_regions_of_interest_2d in rc_roi_db instead.
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/get_regions_of_interest_2d

delete_regions_of_interest_2d (deprecated)

Deletes the configured 2D regions of interest with the requested region_of_interest_2d_ids.

Details

This service can be called as follows.

This service is not available in API version 2. Use delete_regions_of_interest_2d in rc_roi_db instead.
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/delete_regions_of_interest_2d

Internal services

The following services for configuring grasps can change in future without notice. Setting, retrieving and deleting grasps is recommend to be done via the Web GUI.

set_grasp

Persistently stores a grasp for the given object template on the rc_visard. All configured grasps are persistent over firmware updates and rollbacks.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/set_grasp
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/set_grasp

Details for the definition of the grasp type are given in Setting of grasp points.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "grasp": {
      "id": "string",
      "pose": {
        "orientation": {
          "w": "float64",
          "x": "float64",
          "y": "float64",
          "z": "float64"
        },
        "position": {
          "x": "float64",
          "y": "float64",
          "z": "float64"
        }
      },
      "template_id": "string"
    }
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "set_grasp",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

set_all_grasps

Replaces the list of grasps for the given object template on the rc_visard.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/set_all_grasps
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/set_all_grasps

Details for the definition of the grasp type are given in Setting of grasp points.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "grasps": [
      {
        "id": "string",
        "pose": {
          "orientation": {
            "w": "float64",
            "x": "float64",
            "y": "float64",
            "z": "float64"
          },
          "position": {
            "x": "float64",
            "y": "float64",
            "z": "float64"
          }
        },
        "template_id": "string"
      }
    ],
    "template_id": "string"
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "set_all_grasps",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

get_grasps

Returns all configured grasps which have the requested grasp_ids and belong to the requested template_ids.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/get_grasps
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/get_grasps

If no grasp_ids are provided, all grasps belonging to the requested template_ids are returned. If no template_ids are provided, all grasps with the requested grasp_ids are returned. If neither IDs are provided, all configured grasps are returned.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "grasp_ids": [
      "string"
    ],
    "template_ids": [
      "string"
    ]
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "get_grasps",
  "response": {
    "grasps": [
      {
        "id": "string",
        "pose": {
          "orientation": {
            "w": "float64",
            "x": "float64",
            "y": "float64",
            "z": "float64"
          },
          "position": {
            "x": "float64",
            "y": "float64",
            "z": "float64"
          }
        },
        "template_id": "string"
      }
    ],
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

delete_grasps

Deletes all grasps with the requested grasp_ids that belong to the requested template_ids.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/delete_grasps
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/delete_grasps

If no grasp_ids are provided, all grasps belonging to the requested template_ids are deleted. The template_ids list must not be empty.

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "grasp_ids": [
      "string"
    ],
    "template_ids": [
      "string"
    ]
  }
}

The definition for the response with corresponding datatypes is:

{
  "name": "delete_grasps",
  "response": {
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

get_symmetric_grasps

Returns all grasps that are symmetric to the given grasp.

Details

This service can be called as follows.

PUT http://<host>/api/v2/pipelines/0/nodes/rc_silhouettematch/services/get_symmetric_grasps
PUT http://<host>/api/v1/nodes/rc_silhouettematch/services/get_symmetric_grasps

The definition for the request arguments with corresponding datatypes is:

{
  "args": {
    "grasp": {
      "id": "string",
      "pose": {
        "orientation": {
          "w": "float64",
          "x": "float64",
          "y": "float64",
          "z": "float64"
        },
        "position": {
          "x": "float64",
          "y": "float64",
          "z": "float64"
        }
      },
      "template_id": "string"
    }
  }
}

The first grasp in the returned list is the one that was passed with the service call. If the object template does not have an exact symmetry, only the grasp passed with the service call will be returned. If the object template has a continuous symmetry (e.g. a cylindrical object), only 12 equally spaced sample grasps will be returned.

Details for the definition of the grasp type are given in Setting of grasp points.

The definition for the response with corresponding datatypes is:

{
  "name": "get_symmetric_grasps",
  "response": {
    "grasps": [
      {
        "id": "string",
        "pose": {
          "orientation": {
            "w": "float64",
            "x": "float64",
            "y": "float64",
            "z": "float64"
          },
          "position": {
            "x": "float64",
            "y": "float64",
            "z": "float64"
          }
        },
        "template_id": "string"
      }
    ],
    "return_code": {
      "message": "string",
      "value": "int16"
    }
  }
}

Return codes

Each service response contains a return_code, which consists of a value plus an optional message. A successful service returns with a return_code value of 0. Negative return_code values indicate that the service failed. Positive return_code values indicate that the service succeeded with additional information.

Table 40 Return codes of the SilhouetteMatch module services
Code Description
0 Success
-1 An invalid argument was provided
-3 An internal timeout occurred, e.g. during object detection
-4 Data acquisition took longer than allowed
-7 Data could not be read or written to persistent storage
-8 Module is not in a state in which this service can be called. E.g. detect_object cannot be called if there is no base-plane calibration.
-10 New element could not be added as the maximum storage capacity of regions of interest or templates has been exceeded
-100 An internal error occurred
-101 Detection of the base plane failed
-102 The hand-eye calibration changed since the last base-plane calibration
-104 Offset between the base plane normal and the camera’s line of sight exceeds 10 degrees
10 The maximum storage capacity of regions of interest or templates has been reached
11 An existing element was overwritten
100 The requested load carrier was not detected in the scene
101 None of the detected grasps is reachable
102 The detected load carrier is empty
103 All detected grasps are in collision
107 The base plane was not transformed to the current camera pose, e.g. because no robot pose was provided during base-plane calibration
108 The template is deprecated.
109 The plane for object detection does not fit to the load carrier, e.g. objects are below the load carrier floor.
151 The object template has a continuous symmetry
999 Additional hints for application development

Template API

For template upload, download, listing and removal, special REST-API endpoints are provided. Templates can also be uploaded, downloaded and removed via the Web GUI. The templates include the grasp points, if grasp points have been configured. Up to 50 templates can be stored persistently on the rc_visard.

GET /templates/rc_silhouettematch

Get list of all rc_silhouettematch templates.

Template request

GET /api/v2/templates/rc_silhouettematch HTTP/1.1

Template response

HTTP/1.1 200 OK
Content-Type: application/json

[
  {
    "id": "string"
  }
]
Response Headers:
 
Status Codes:
Referenced Data Models:
 
GET /templates/rc_silhouettematch/{id}

Get a rc_silhouettematch template. If the requested content-type is application/octet-stream, the template is returned as file.

Template request

GET /api/v2/templates/rc_silhouettematch/<id> HTTP/1.1

Template response

HTTP/1.1 200 OK
Content-Type: application/json

{
  "id": "string"
}
Parameters:
  • id (string) – id of the template (required)
Response Headers:
 
  • Content-Type – application/json application/octet-stream
Status Codes:
  • 200 OK – successful operation (returns Template)
  • 404 Not Found – node or template not found
Referenced Data Models:
 
PUT /templates/rc_silhouettematch/{id}

Create or update a rc_silhouettematch template.

Template request

PUT /api/v2/templates/rc_silhouettematch/<id> HTTP/1.1
Accept: multipart/form-data application/json

Template response

HTTP/1.1 200 OK
Content-Type: application/json

{
  "id": "string"
}
Parameters:
  • id (string) – id of the template (required)
Form Parameters:
 
  • file – template file (required)
Request Headers:
 
  • Accept – multipart/form-data application/json
Response Headers:
 
Status Codes:
Referenced Data Models:
 
DELETE /templates/rc_silhouettematch/{id}

Remove a rc_silhouettematch template.

Template request

DELETE /api/v2/templates/rc_silhouettematch/<id> HTTP/1.1
Accept: application/json
Parameters:
  • id (string) – id of the template (required)
Request Headers:
 
Response Headers:
 
Status Codes:
  • 200 OK – successful operation
  • 403 Forbidden – forbidden, e.g. because there is no valid license for this module.
  • 404 Not Found – node or template not found