Research On The Detection And Utilization Of Three-dimensional Information Of Parts Based On Laser Scanning

In the field of high-speed online inspection, machine/robot vision, quality control,automatic processing, physical profiling, etc, three-dimensional information′s detectionand application is increasingly important. The development of the three-dimensionaldigital information detection system is of tremendous significance, which can simplyextract parameter and be high accuracy, real-time, and good versatility.In this paper, the laser triangulation method is applied to set up three-dimensionaldigital measurement system. It can measure parts without physical contact and easilyget access to data, in addition, it has simple structure and can be real-time measuring.Swing prism issued by the laser spot scan a cross-section. According to the pixellocation of the spot on the CCD sensor, the points′coordinates in the scanned cross-section can be measured point by point. Sensor can be moved along the normaldirection for scanning cross section, the3D point cloud data of the measured surfacecan be obtained throuth this method.Motion-controled chip LM629and motor driver IC LMD18200are applied todesign a closed-loop circuit with the software of Altium Designer6.0. PID algorithmand speed ladder are designed in Microcontroller ATmega16in order to run the motorsmoothly.For the extraction of the spot center, the means of fast median filtering based on thecomparison matrix is put forward. It effectively filter out the noise at the center of thespot after median filtering. Apart from this, extraction algorithm of the spot center isresearched, the extreme value method, the threshold, the improved center of gravitymethod, the adaptive threshold method were analyzed separately. Spot center isextracted by the extreme value method at normal exposure, and extracted by theadaptive threshold method when overexposed. They both achieved good results.The main factors affecting the measurement accuracy include the accuracy of thelaser spot center′s extraction, the accuracy of the relationship between depth and pixelnumber, the error of angle′s measurement. According to this, extraction algorithm ofthe spot center is researched, the output A and B signals of encoder are refined toimprove the accuracy of calculation of angle, and a simple and practical sensorcalibration method is proposed. The relationship between the depth and pixel number isestablished, and the error angle of laser scanning has been corrected.For the srecognition of parts, the selection of characteristic has been carried out.There are recognitions for the basic shape of the circle, polygon, triangle, etc. For theparts like boss, shaft, bearing and block, the feature of diversification of coordinates andthe angle between edges are choosed to identify the part. The algorithm of extraction ofV-shaped weld feature is designed to extract the depth and cross-sectional area of weld. The research of the identification of parts in this paper laids a solid foundation for theresearch of the pattern recognition of more complex parts and the strategy of robotcontrol.