A belt longitudinal tear line detection system based on machine vision is proposed in this study.By fixing the industry cameras on both sides of the conveyor belt, while the anti-explosion lightsource is placed on the surface of the conveyor belt, it can achieve the fullest extent of the imageacquisition. Industrial camera captures the image of the conveyor belt and transmit them to theperson computer (PC),in which the pattern recognition methods are used to detect longitudinaltear on the conveyor belt, the methods were used as follows: the first is the use of the conveyorbelt tear image preprocessing, including histograms equalization, edge detection and improvedpulse coupled neural network (PCNN-Pulse Coupled Neural Network) algorithm, imagepreprocessing highlights the image features, and better noise suppression; Secondly, choose acertain number of the conveyor belt after the image preprocessing methods above, produce theimage after the pretreatment of samples, which contains conveyor tear and normal image.Extracting the feature of the samples, then we use a combination of genetic algorithm supportvector machine training samples. Samples can be classified by the use of trained support vector.We can ultimately achieve intelligent and accurate identification results. |