Research On The Key Technologies Of Image Processing In Vision Measurement For Automobile Crossbeam

Machine vision measurement method has been widely used in many industrial fields because it can achieve fast, non-contact, automated measurement and with low cost. But due to the limitation of the camera view, how to improve measurement accuracy for large workpiece like automobile crossbeam has always been technical difficulty of visual measurement technology. This paper aims to study some key technical problems of image stitching for large workpiece measurement and provides technical support for automotive crossbeam vision measurement system development.Firstly, the reason of image noise and the method for image filtering have been studied. According to structural features of the car chassis beam, the median filter and a predetermined image of image preprocessing program are presented through a series of comparative experiments. Moreover, by analyzing the limitations of traditional Hough algorithm, an improved algorithm based on Hough circle extraction has been studied.And combination the method and sub-pixel positioning method can meet the needs of the circle extraction and also improve the visual accuracy.Secondly, By comparing two common mosaic models, establishing the image mosaic method is established based on geometric features, the coordinate normalization of two adjacent images is investigated and the image mosaic is realized. And the nonlinear problems of two-dimensional similar transformation model are mainly solved,the influence of local image measurement precision and the public number of coordinate transformation precision are analyzed. It can be found that the local measurement precision of the coordinate transformation and public numbers affect the accuracy of coordinate transformation. The experimental results show that hole stitching precision can reach 0.3mm in any two adjacent pieces of automobile beam image.Thirdly, the precision control method of several image sequence mosaic and the overall accuracy has been discussed. And the method to eliminate the accumulated error of the mosaic method based on geometric features and control field has been studied and confirmed by experimental measurement. The experimental results show that the splicing model with control field shows good adaptability in automobile beam fitted holes measurement, and it can eliminate the effect of cumulative error and improve themeasurement precision greatly. Additionally, it can overcome the disadvantages of image mosaic due to the rarely geometric characteristics.Finally, the calibration method and the camera imaging model are studied based on the measuring system, and the method with Zhang Zhengyou is selected to calibrate the camera. In the experimental condition,we choose a steel plate with 1200 mm length as the sample. The experimental results show that the aperture error of the system is less than ±0.2mm, and the hole position error was about ±0.5mm, which can meet the production requirements of enterprises and achieve anticipated effect.