Aiming at the problem that in robotic graphical teaching programming system

complex background will affect on detection of teaching object

a spatial target detection and tracking algorithm based on fully convolutional neural networks is proposed. The algorithm trains the network to identify the target

segment the image

and noise points is eliminated by binary processing and morphology processing.After doing these

the centroid of the largest connected region is obtained

and the target position is determined by using visual space coordinate transformation.Experimental results show that

compared to the general algorithm

the algorithm can identify some specific targets

to some extent

expand the range of teaching space

reduce the difficulty of teaching complex trajectory

and eventually improve the control accuracy of the robot.