This page provides a detailed overview about the Vision panel. Further information can be reviewed within the Vision Success Guide.
- Snapshot Gallery
- Snapshot Editor
- Training a Surface
- Train Snapshot
- Search Area & Default Settings
The Snapshot Gallery is used to train an store and vision objects. Vision objects include landmarks, objects to be detected using the arm camera, and objects to be detected using an external camera. The Snapshot Galley is accessed by clicking the icon in the tool bar highlighted in the image below.
After clicking the icon in the tool bar, the Snapshot Gallery will open.
- To train a vision object, click on the "+Snapshot" button at the top of the Snapshot Gallery or click the "Add New" button under one of the 3 available options (Arm Camera, External Camera, Landmarks).
- Clicking any of the available options will open the Snapshot Editor.
- Note: When adding a Snapshot using the "+Snapshot" button, the Snapshot editor will open to the Locator tab under Arm Camera.
The Snapshot Editor contains two primary tabs for the vision system type, Arm Camera and External Camera.
Arm Cam Tab
The images below show the three options available when using the arm camera; Locator, Inspector, or Landmark. Each of the three options uses one or more of the following concepts; Training a Surface/Landmark, Training a Snapshot, and modifying the Search Area and Settings. These concepts are details in the sections below.
See the External Vision page for more information.
Training a Surface
Landmarks can be used as reference frames for re-registering task positions or to set an image surface when training a snapshot. The instructions below detail how to train a new landmark, training a surface for a snapshot is a very similar process and is detailed in the Training a Snapshot section below.
Note: The surface type option is only available for Landmarks. It is not available for a Locator or Inspector snapshot. This is because the Landmark surface frame is specifically for the Robot Positioning System (RPS) Landmark registration process.
- Open the Snapshot Gallery and click the Add New button (+ symbol) under the section titled Landmarks. The snapshot editor will open to the landmark tab.
- Fix the landmark to a solid surface that is larger than the size of the landmark to ensure the landmark cannot become bent or deformed. Note that the landmark should be placed as close as possible to the arm locations that will be referenced in order to ensure an accurate re-registration.
- Move the robot arm to a position where the landmark is within the view of the arm camera. Note that a message may appear stating that the exposure should be adjusted. Do not change any settings at this time as the robot has not yet optimized the arm position. See the image below for reference.
- Note: If no Landmark is present and a surface is being trained for a vision locator, the active endpoint must be used to define the surface. Position the arm such that the active endpoint is defining the desired frame. If the surface is horizontal or vertical, use the cartesian panel to make sure the arm is pointing perfectly downward or straight, and make sure the tool center point Z direction is accurate. Press ‘TOOL SET’ to set the surface. It is generally not recommended to use the active endpoint to define the surface because it can be difficult to align the arm perfectly with the surface if it is not parallel to one of the planes defined by the base frame.
- Select a Landmark surface type option:
- Any- Landmark registration will update the position of all axes (X, Y, Z, Rx, Ry, Rz) of the Landmark surface frame upon re-registration. This option is subject to the most registration error, but should be used when the parallel movement and distance to the original Landmark location, or the Landmark relative to the the robot arm camera cannot be controlled, or guaranteed. This is the default setting.
- Fixed, Variable Z- The position of Rx and Ry axes of the Landmark surface frame are fixed upon establishing the Landmark and will not be updated upon a Landmark re-registration. A Landmark re-registration will update the positions of the X, Y, Z, and Rz axes of the Landmark surface frame. This option helps to reduce re-registration error due roll and pitch variations.
- Fixed- The position of Rx, Ry, and Z axes of the Landmark surface frame are fixed upon establishing the Landmark and will not be updated upon a Landmark re-registration. A Landmark re-registration will update the position of the X, Y, and Rz axes of the Landmark surface frame. This option should be used if there is a constant distance between the Landmark surface frame and the robot arm camera but can move parallel and relative to a consistent surface. For example, this option is suggested if the robot arm camera distance (Z) will remain consistent relative to a surface (e.g. table), but where something on a table (e.g. fixture) may move in one plane only. This option is subject to the least re-registration error.
- Click the SET button in order to register the landmark.
- If the ‘AUTO’ toggle is on, Sawyer will automatically move the arm to the optimal position and register the surface.
- If the ‘AUTO’ toggle is off, users will see and additional ‘OPTIMIZE’ button. Clicking ‘OPTIMIZE’ will move the arm to the optimal position. Alternatively, users can adjust the arm pose based on their preference. In either case, press ‘SET’ to register the surface. Note: It is a best practice to always use the optimal settings for landmarks.
- The landmark will now register and the Snapshot Editor will show the screen below.
- The landmark is now stored in the Snapshot Gallery and can be referenced in a task. For details on how to use a landmark in the within a task, see the Robot Positioning System page.
Training a Snapshot is required when using the Locator or Inspector features. The Snapshot is the original object image that will be used as the base for the vision recognition. When using the Locator function, a surface will need to be trained using the steps in the section above before the object can be trained.
- Move the arm to the position that will be used to search for the object. The arm should be positioned far enough from the object so that it has a complete view of the object at all potential locations, but not so far as to make the features difficult to discern.
- If using Locator, click ‘OPTIMIZE’ to align the camera principal axis perpendicular to the surface.
- Adjust the camera settings so that the object contrasts with the background. Use of the flash is recommended to provide consistency in varying lighting conditions. In certain cases - especially if working with reflective objects - the flash may cause too much glare and should not be used.
- Resize, move, and rotate the green box to fully outline and match the orientation of the tracked object, including some extra “padding” around the object.
- Note that in many cases it is better to choose a small, detailed, and easily identified portion of the object to track instead of the entire object.
- Click ‘SET’ to train the object, which may take a few seconds to complete. If the object is successfully trained, a green box will appear around it. See the image below fro reference.
- See the section below for details on defining the search area and defaults.
Search Area & Default Settings
The Search Area & Default Node Settings tab is used to define the search area and default settings which will be present when you add either a Vision Inspect or Vision Locate node and track the object being trained.
- Set the similarity to be a value in the range of 20 to 90. Similarity controls how closely the new image of the object must match the initial training of the object for a successful recognition to occur. The higher the similarity setting, the more selective the system will be when looking for objects.
- Set the Start and End Angle to be a value in the range of -180 and 180. Start Angle and End Angle define the highest clockwise and counterclockwise rotation of the object which will still be tracked by Sawyer. If all potential orientations should be recognized, set the angles to be 180 and -180.
- Enabling the symmetrical option makes tracking squares, rectangles, and circles more consistent. When tracking an object which is highly symmetrical, this option should be enabled. Enabling symmetrical automatically overrides the Start Angle and End Angle options.
- Define the search area. A larger green box appears in the camera view after training the Snapshot. Resizing this box defines the search area. Sawyer will only detect objects in this box and ignore the area outside. This feature can be used to reduce time the camera spends searching for objects.
- The image below shows a search area that has been reduced in size and rotated in order to limit the area where the vision system searches for the object.
- Click the SET button to save the settings and finish training the Snapshot. The image below shows the appearance of the Snapshot Editor after the settings are saved.