Capacitance Screen Defect Recognition Method Based On Machine Vision Research

Touch-screen, as a new human-computer interaction techniques, are widely used in many fields, such as communications equipment, public information inquiry services, appliances, etc. Because of its high transmittance, durability and supporting for multi-touch technology characteristics, projected capacitive touch-screen has grown to become the market trend. Therefore, in the production process of the capacitance plate, its quality testing is particularly important, in the past, manual inspection was commonly used, however, there are several drawbacks, such as low efficiency, large differences, etc. this article works over segmentation and recognition of capacitance plate defects by the method of machine vision. The details are as follows:(1) According to the defect sample library provided by the manufacturer of capacitance plate, the software algorithm is researched, which includes Image Preprocessing, Image Segmentation and Post-Processing. Feature Extraction and Image Recognition.(2) The defect image has necessarily presence of noise due to the production of the environmental factors at the scene, therefore, it’s necessary to be pretreated prior to image recognition.(3) The laws of defect are analyzed by comparing the standard image and the defect image, a suitable method of segmentation is selected by comparing with each other, after that image is handled by filtering.(4) Boundary tracking algorithm is applyed to process the segmented, then the features is extracted, after comparing the extracted feature parameters. Hu invariant moments is regarded as the feature vectors of image recognition.(5) After a few steps above, we can identify the capacitance plate defect image. This article is based on the way of BP network classifier.Experimental results show that the training of the mean square error eventually is able to get a good convergence. Then select a certain amount of testing samples for identification. The average recognition rate reaches95%that basically conforms the requirements of the capacitance plate defect recognition.