Research On Robot Autonomous Polishing System Based On Binocular Visio

Sanding is an essential process in the manufacturing process.At present,the traditional sanding process mainly relies on experienced manual operation,with slow sanding speed and low consistency of sanding accuracy.The robot is widely used in the grinding industry to replace the traditional manual grinding method with its intelligent and automatic features,which can avoid the health hazards caused by the workers’ longterm monotonous labor,but the planning of the robot grinding path directly affects the grinding quality of the workpiece.This paper researches and designs a 3D point cloudguided robotic autonomous grinding system based on binocular structured light for the surface grinding of castings,which can effectively solve the problems of low efficiency and inconsistent accuracy caused by manual operation,and reasonably plan the grinding path and position according to the real shape of the polished workpiece to improve the grinding accuracy and quality.First,for the measurement of reflections on the surface of metal castings,this paper proposes a three-dimensional measurement method combining complementary phase encoding and multi-frequency outlier.The outlier phase of the main value of the streak image phase is obtained by phase outlier,the streak level is obtained by using the complementary phase-encoded streaks,and the phase expansion is performed according to the outlier phase and streak level,and the high-precision point cloud data of polished castings is obtained by combining the sub-pixel binocular polar line constraint method.This method can avoid 3D measurement errors caused by phase jumping and phase superposition,and is faster,more robust and reliable when measuring reflective objects,black objects and scenes with large contrast between light and dark.Then,the unification of coordinate system is achieved by hand-eye calibration,and path planning is carried out using real casting point cloud data.In this paper,we first calculate the adaptive row spacing for grinding based on the dimensional model of the grinding wheel,and use the tangent plane method to obtain the grinding route according to the row spacing for the image projected onto the 2D plane from the point cloud model.Secondly,the nodes to be selected for each grinding trajectory are obtained by the equidistant step method,and the maximum residual height of the node neighborhood is interpolated to obtain the exact path points for grinding.Then the normal vector of each path point is calculated by using the weighted principal component analysis method,and the normal vector is optimized by improving the point cloud smoothing filtering algorithm,and the normal vector is transformed into the robot’s positional expression to complete the grinding path and positional planning.Finally,a software system is designed to realize the vision-guided robot to complete accurate,flexible and intelligent autonomous grinding processing.The experimental results show that the binocular vision measurement system can achieve ±0.015 mm 3D reconstruction accuracy in the measurement range of320mm×280mm,and ±0.5mm accuracy for robot autonomous grinding,and improve the efficiency of grinding to meet the industrial requirements of casting grinding.This system can also be applied to intelligent manufacturing fields such as vision-guided robot intelligent welding and vision-guided robot intelligent gripping.