Research On Machine Vision Based Part Geometrical Measuring Method And Development Of Measuring System

In the recent years, on one hand, with the development of manufacturing technology, some parts appear which are thin or have features hard to measure by contact type measurement; on the other hand, people have higher requirement for quality of product, so producers need to detect as many products as possible to assure quality. Traditional contact type measurement and its equipment is unable to meet detection requirements of some products due to need to contact with product, long measuring intervals of time, large artificial error and other disadvantages. This paper focuses on the research of image denoising, edge detection, image mosaic and geometrical dimensional measurement technique among machine vision measuring key technologies, and this paper also develops a geometrical dimensional measurement system to realize the non-contact measurement of distance, diameter, angle, straightness, parallelism and roundness.For image denoising, to filter the hybrid noise of part image, an adaptive filtering method for hybrid noise is proposed. This method firstly uses rule to judge the noise type, and then select corresponding filter to filter noise which has good filtering performance for single noise according to noise type.For image mosaic, to solve the problem that part image is hard to mosaic, an accurate image mosaic method for part with rare feature information is proposed. Using a part with enough feature information, this method captures serial images of the part when imaging system moves along X axis and Y axis respectively and does pretreatment, and then solves registration parameters using phase correlation algorithm on two adjacency images. With the support of good repeatability of closed loop motion control system, the solved image registration parameter can be used as system registration parameter calibrated, namely, is registration parameter of the image mosaic of part with rare feature information.For edge detection, to detect single pixel edge, Otsu threshold method is introduced to adaptively calculate double threshold for edge connection, and Canny edge detection method is used to detect part edge which is not single pixel but is close to. Afterwards, a rule is designed to thin the edge to single pixel edge. To get sub-pixel edge, based on single pixel edge, conics fitting method is adopted to position the edge to sub-pixel.For geometrical dimensional measurement, the research is focused on structuring method of geometrical elements and geometrical dimensional measurement principle of contact type measurement. According to characteristic of part image, least squares technique is selected to structure the line edge and circular arc edge. Geometrical dimensional measurement is carried out according to corresponding contact type measurement principle.Taking gauge block, ring gauge and angular gauge block which are standard component with high precision as measuring objects, experiments verify the validity and the measurement precision of theoretical research and measuring system. Experiments firstly capture the single image or mosaic image of measuring objects, and then do geometrical dimensional measurements using measuring software developed by this paper, afterwards compare measuring result with that of standard component. Experimental results show, (1) measuring precision:when part can be put in area of 64mm X 48mm, namely, when the measurement is done on the single image, measuring precision in length, straightness, parallelism, diameter, roundness is 6μm,10μm,10μm,3μm, 3μm, respectively; when part is beyond area of 64mm X 48mm, namely, when the measurement is done on the mosaic image, measuring precision in length, straightness, parallelism, diameter, roundness is 10μm,10μm,7μm,13μm,10μm, respectively; Measuring precision in angle is 2’. (2) Measurement efficiency:the measuring time for gauge block including single image capturing time and time of measuring length, straightness and parallelism and the time for ring gauge including single image capturing time and time of measuring diameter and roundness are 68s and 35s respectively; the measuring time for gauge block including mosaic image capturing time and time of measuring length, straightness and parallelism and the time for ring gauge including mosaic image capturing time and time of measuring diameter and roundness are 84s and 53s respectively. Measuring system developed in this paper realizes high-precision, rapid and non-contact part geometrical measurement based on machine vision.