Robot 3D Pickinng Technology Of Scattered Components Based On Binocular Vision

With the development of robotics technology,production and processing methods are increasingly tending to be diverse and flexible.In an automated production line,how to sort the scattered workpieces in the material frame becomes a problem that hinders the development of automated production lines.At present,the main use in the actual production is the mechanical vibration sorting device for loading and unloading scattered workpieces.However,vibration sorting devices often have the disadvantages of large noise,poor flexibility,and complicated design.With the maturity of visual inspection technology,the use of visual inspections in conjunction with robots to disorganize the stacking of workpieces can effectively overcome the above shortcomings.This dissertation focuses on how to use three-dimensional vision technology to sort workpieces that are in a stacking state.This paper studies the three-dimensional reconstruction and three-dimensional spatially dispersed workpiece pose matching.Combining with the three-dimensional reconstruction requirements of the application scene,through the analysis of the advantages and disadvantages of different threedimensional reconstruction methods,the binocular three-dimensional vision based on structured light is applied.For the reconstruction scope requirements of the scene and the positioning accuracy requirements of the target workpiece,the internal parameters of the binocular measurement system are determined.In order to solve the problem of difficulty in matching the left and right cameras in the three-dimensional reconstruction of binocular vision,the spatial encoding method is studied.Aiming at the problem of how to determine the pose of the target workpiece from the three-dimensional point cloud scene of scattered workpieces,a three-dimensional positioning method for scattered workpieces is presented by analyzing the advantages and disadvantages of existing three-dimensional positioning methods and the application limitations.By using the scattered image and gray image in the scene of the workpiece,a method of segmenting the 3D point cloud using a 2D image is provided.For the problem of how to register the target point set and the model point set,a registration method of the target point set and the model point set is proposed,and a registration evaluation coefficient is proposed to screen the results.Based on the above research content,the system's hardware and software are designed to build a test platform for testing.Using the binocular visual three-dimensional sorting platform for binocular vision,a three-dimensional sorting and comprehensive experiment was performed on the workpieces in the material box.Through white plastic bottles,disposable cups,and alloy tee castings,the data was recorded during the test and six-joint robots were used to perform three-dimensional workpiece sorting.