Object detection and robotic sorting system in complex industrial environment

In complex industrial environment, there are many interference objects when detect the objects on the production line. The interference objects are very similar with the objects to be sorted in terms of color, shape and size. The existing detection method like edge detection and object segmentation is difficult to recognize the objects when it comes to complex industrial environment. In the complex industrial environment, objects' position and posture appear randomly, and it is difficult to calculate the position and posture at same time. This work uses the CNN network to deal with the object image acquired by vision sensor. And we design two decoupling regression layers to calculate the position and posture separately. It gets a high performance on objects detection in the complex environment. The robots followed by the vision sensor will sort the objects after objects detection. Because the image processing takes a relatively long time in the control system, this work solves the large delay control problem by designing a robots scheduling module. Then the sorting job running on a high speed. The experiment results show that our system can achieves 91.6% precision on the Production Dataset, Which takes 220ms per image running on Nvidia GPU TitanX. More important, the system can substitute humans in smelly and dirty environment and reducing labor costs by 67% for factory, where factory need some dedicated people to do the sorting jobs before.