In order to have the robot pick up the part you would need to calculate the real world position of the object relative to the robot and take into consideration where the center of the object is in the image. Because objects near one side of the image are further away from the robot than the other side the pixel to cm or inch ratio will need to be know. This will depend on your camera setup.
Once you know the true distance from the object to the robot you can actually use that to determine overlapping detections. I.e. if two objects are X cm from the robot with the same shape, then it is a duplicate detection and one of the entries can just be ignored. You can do this in multiple ways but for a small number of objects simply comparing a table of past matches to the current matches will allow you to determine this.
There are multiple ways to do duplicate detection but using an existing table check will allow parts to fail detection multiple times but as long as they are detected once it will work.
Note, you also probably want to filter out any objects that are touching the image borders (see blob filter) to avoid partial matches from causing issues.
If you need additional help, you can post a sample video online somewhere and link to it from this forum so we have an example to work with as each setup is a bit different. Or if you can deduce what you need from our example video we can update that tutorial to include duplicate detection.
STeven.
If you have any questions/issues about RoboRealm or image processing this is your chance to ask a question.
Your question and any responses will be posted to the forum section of the website for others to read and learn from.
Note that any submission will be considered property of RoboRealm. If you do not wish your images to be
displayed as part of the RoboRealm site please do not submit those images.
Any submission may be removed from the website at the discretion of RoboRealm personnel.
