A Research And Implement Of The Computer Vision Real-time Measurement Technology

Computer vision real-time measurement technology which gets over the imprecision and inefficient of the traditional measurement technology has become the mainstay industrial measurement technology today and to measure hardware geometrical size by means of computer vision real-time measurement technology which has a brilliant practical future. The image segmentation algorithms is mainly used and make further research on the accuracy edge detection algorithms by which to real-time measure the width of stripe in the coronary artery bracket to monitor its polishing procedure.In the image preprocessing phase the present image enhancement technology is studied and then improve the median point filter algorithm based on the characteristic of the gray-level histogram of the image form the spot and the geometrical traits of the bracket itself which denoise the random pulse noise in the precondition of protecting the edge information. In order to find out the location of the coronary artery bracket the optimized threshold segmentation and iterative threshold segmentation algorithms are discussed in this dissertation as a base then improve the iterative threshold algorithm with the multi-scale idea which reduce the runtime of the classic algorithm and it is very suit for real-time measurement.In order to measure the size of the stripe in the bracket we must locate the single stripe area. We take full advantage of the upright and horizontal projections of the bracket image and the configurable character of the stripe combined together with statistical histogram and difference to accurately locate and dynamic choose one stripe which is most fit for measurement among the stripes in the bracket. Then specify one rectangle area which contains the stripe. One improved adaptive filter algorithm is used to treat texture areas that contain the edges and smooth areas with different filters.During the studying of measuring the size of the stripe in the bracket we find out that detecting the edges of the stripe by multi-scale wavelet transform have the advantage of thetraditional edge detection operator in accuracy and anti-noise after making further research on the application of the multi-scale wavelet analysis in edge detection based on the former researches. In this dissertation the 3-order based spline multi-scale wavelet edge detection algorithm are applied to detect edges which out and away improve the measurement accuracy of the stripe size. As accepted that the wavelet transform is not fit for the realtime detection for its heavy computing. We propose a simplified multi-scale wavelet transform edge detection algorithm based on the characteristic of the stripe and because of the small segmentation area the algorithm can complete in 0.24 seconds which can absolutely meet the need of the real-time monitoring. Finally the measurement error can be controlled within + 7 millimeters.