The Algorithm Research And Design Of Ceramic Substrate Detection Based On CMOS Image

With the rapid development of computer technology and photoelectricity technology, the machine measurement technology has been developed rapidly and applied widely. As a non-contact detection instrument, it is getting increasing regard. Using machine vision measurement for industrial inspection can not only overcome the disadvantages of manual detecting action, but also can improve the detection accuracy and detection efficiency. Also, it can reduce the production costs. With the machine vision theory studying deeply, machine vision detection gradually developes towards to online real-time and high accuracy detection. It must be widely applied in various fields.Ceramic substrate is a ceramic plate. It is produced by pressing and sintering. It has a degree of uncertainty which affects its applications when it is pressed, because its subsrate material is relatively soft when pressed. The detection of the ceramic substrates quality mainly detects its basic characteristics(composition, status, colored, etc.), appearance quality(cracks, spots, defects, etc.), dimensional tolerances(shape tolerance, perpendicularity, parallelism, etc.) and product specifications(thicknesser, shape). At present, lots of substrate manufacturing enterprises mainly rely on manual inspection to detect the substrate quality. This way has a low detection speed, and also results in test error because of the visual fatigue. With hardware devices developing and pixel segmentation technology emerging, it is possible that detect the ceramic substrate quality with machine vision technology.The CMOS-based ceramic substrate detection system has been mainly studied in this paper. The system obtained the two-dimensional image data of the ceramic substrate by CMOS-camera using optical methods. Firstly, it processed the image that was obtained by CMOS-camera with the image processing software to eliminate all kinds of image noise. And then rough positioning and precise positioning with edge detection algorithms and sub-pixel subdivision algorithm based on spatial moments were done. Finally, the ceramic substrate sizes were computed according to the pixel equivalents which were obtained by the system calibration algorithm.In this paper, the composition of machine vision measurement system, the elimination of image noises, the location methods of edge detection and system calibration technique were studied. The CMOS image sensor technique were summarized. An edge thinning algorithm based on Roberts was proposed. Also, an calibration algorithm which was sub-pixel level accuracy was designed. Finally, the accuracies of detection results were analyzed combined with the sub-pixel subdivision algrithm based on spacia moment and discussed the error sources which proposed the theoretical basis for improving the deteciton system.