Computer Vision News - February 2018

Another type of machine vision tasks is OCR for robotics : robots complete their vision with textual data appearing in their visual field. Machine vision should be highly accurate, especially when the robot faces a changing environment, like an event. When conditions change (lighting, viewing angles, moving or changing objects) accuracy of machine vision is of crucial importance, to make sure that the robot can carry its task: navigating robots, like in a warehouse or in retail , need to move between people and/or other robots. Objects may not be oriented or located in the right position and since the robot is using its arms to pick and place, the need for precise guidance of the arm into the location is key for the success of the task. All the more so for robots doing surgery or other medical tasks . This imposes a great challenge, because classic machine vision algorithms cannot be programmed to handle such a wide range of possible events: changing combination of parts, size and orientation, changing light conditions and many more. Deep Learning techniques , like CNN , are used to train the machine vision of the robots and prepare it to handle this wide range of events. Engineers at RSIP Vision are experts in both machine vision and deep learning : we have already carried successful projects involving detecting and classifying object under a wide range of occurrences. As we understand, robotics is not just the hardware and its mechanical parts: once you have the robot with its requirements, we can assist you in training your robot to do the task. We will use video or image feeds as a training set and we will program the required procedure, adapting the most proper deep learning scheme for the operation at hand. Read more on our website about Machine Vision for Robotics . Computer Vision News Project Computer Vision News 9 Machine Vision for Robotics “Accuracy of Machine Vision is of crucial importance, to make sure that the robot can carry its task ”