Key Technology Research Of Precision Parts Size Detection Based On Machine Vision

Part size measurement based on machine vision is based on Image processing,andresearch parts size measurement algorithms is mainly in software as a means to solvevisual measurement accuracy and speed problems. Machine vision measurementtechnology has capabilities of real-time detection and analysis. In order to improve themeasurement accuracy of part size, edge detection and sub-pixel edge detection andother key technologies are studied.When image processing and analysis is going to be done, edge detection is oftenfirstly used. Edge detection is mainly discussed in the study of classical edge detectionoperator, and some common pixel edge detection methods have been researched. Byusing Matlab programming tools for these types of algorithms experimental comparison,are compared with their strengths and weaknesses and scopes. Finally, an improvedbetter multi-directional wavelet transform edge detection algorithm is proposed.Several commonly used sub-pixel edge detection methods were studied, includinggraymoment, spatialmoment method, zernikemoment method and some other sub-pixeledge detection algorithms. By using Matlab programming tools to achieve these typesof commonly used sub-pixel edge detection algorithms, respectively, two differenttarget images thread and gear are being simulation experimented, comparing theadvantages and disadvantages of each method.An improved quadratic polynomial interpolation and least-squares fitting ofirregular workpiece parts were measured to compare the relative error of the ideal of theworkpiece. This method greatly reduces the fitting sub-pixel positioning errors, improvethe measurement accuracy.In the measurement of glass bottles in the collection of the gray-scale imageprocessing, noise reduction implemented gray image threshold segmentation and edgedetection using improved quadratic polynomial interpolation of the sub-pixel positioning, analysis of the positioning accuracy. Sub-pixel edge detection combinedmoments, least squares fitting, quadratic polynomial interpolation, some of the Houghtransform algorithm for sectional radius parameter measurements made cylindrical glassbottle, the measurement results show that the image processing algorithms andeffective measure to meet the accuracy requirements.