Research On The Measurement Algorithms Of The Outline Size Of Small Parts Based On Machine Vision

Parts size measurement technic has a very important position in the industrial manufacturing process,it can distinguish unqualified products in time and improve production efficiency.Traditional measurement methods have low efficiency and poor accuracy,while advanced equipment is expensive and complicated to operate.Therefore,it is necessary to develop equipment and corresponding algorithms that meet the needs of high-speed and largescale measurement.In this thesis,a measurement method is designed for the parameters to be measured for common small mechanical parts,and corresponding algorithms are proposed to improve the efficiency of part size measurement under the premise of ensuring accuracy.This thesis analyzes the needs and difficulties of part size measurement,proposes a measurement scheme based on machine vision,and researches the measurement methods of related parameters and the specific flow of the algorithms.Edge extraction is an important issue in image measurement.This thesis proposes an improved Canny algorithm,which replaces the original Gaussian filter with a hybrid filter that combines the characteristics of adaptive median filtering and bilateral filtering,adaptively determines high and low thresholds based on the principle of maximum entropy,to extract part edges and mark the neighborhood of the results.Then the sub-pixel edges are searched in the region of interest(ROI),and the parameters in the calculation process are analyzed and judged to efficiently and accurately obtain the sub-pixel edge coordinate data of the part.The parameters to be measured are classified into polyline,arc,roundness,and cylindricity,and the corresponding measurement algorithms are proposed.The contours of the parts are polyline edges in the image.The Douglas-Pucker algorithm(DP)is used to divide them into line segments,and the results are calculated by the least squares straight line fitting algorithm based on the vertical distance.A Particle Swarm Optimization algorithm based on chaos optimization(CPSO)is proposed to calculate the roundness and cylindricity errors,which significantly improves the search efficiency and accuracy of the original algorithm.A visual measurement platform for size measurement of small parts was built to verify the algorithms proposed in this thesis.The results are analyzed and compared with the results from coordinate measuring machine to evaluate the performance of the vision system and algorithms.The results show that all algorithms can meet the measurement accuracy requirements.