Research On Visual Positioning Technology Of De-Palletizing Robot Based On Complex Industrial Scene

The stacker robots destack the bags via visual positioning system in detecting and recognizing the shape of bags in the automatic production line.Thus,stacker robots are able to carry out tasks of autonomous navigation and execution,which improves the perception ability and intelligence of the robots.This thesis aims at resolving the visual requirements of the factories in complex destacking environment with unstable illumination and strong background interference,then a visual recognition of numerous stacking bags and the positioning of packaging bags are realized.The research content of this thesis is listed as follows:(1)The overall solution of the vision positioning system of the destacking robot is designed.This thesis promotes an overall solution for destacking positioning under the environment of unstable natural light,diversified stack types,workpiece appearance deformation and a large number of background interference after investigating and analyzing the destacking environment in complex industrial scenes.The main hardware selection analysis is completed,and the camera internal and external parameters are calibrated according to Zhang's calibration method.(2)The algorithm of visual positioning system is designed.A multi type experimental image database is established,which is based on the characteristics of destacking environment and packaging bag.A general vision processing scheme is proposed,which extracts the whole stack and then locates the single workpiece.The relationship model between the gray value of the image and the optimal threshold is established.This model aims at the segmentation of the whole target stack with considering the influence of the unstable illumination.A region growing method based on the dynamic threshold is proposed to realize the accurate recognition and segmentation of the target stack under the complex background interference.(3)The PPHT algorithm is improved based on the proposed optimal three parameter method.This algorithm realizes the accurate positioning of a single workpiece in the unstable natural light environment.(4)Previously proposed algorithm is verified by experiments.On the basis of building the hardware and software platform of the visual positioning system and developing the man-machine interface,the versatility and robustness of the visual positioning system proposed in this thesis to adapt to different stack types and light changes are verified through quantitative experimental analysis.This thesis provides theoretical basis and effective solutions for the popularization and application of destacking robot in complex industrial scenes,and further improves the automation and intelligence of robot.