Design And Implementation Of Parallel Strip Edge Detection System Based On GPU

Steel strip detection system is a kind of detecting system having high precision, high accuracy, high real time capability which combine the transmission of light sensor technology, communication technology,computer technology and industrial control technology. Strip occupy a larger part of iron and steel production. The quality is more and more influence to all ways of life and all areas.The strip edge is judgment decide whether the steel can be used or not accord to a standard of quality. In the process of strip steel production, steel production lines affected by producing rocker still has a certain amount of weight, so the timely detection of the edge of the strip steel shape in video is deserving of research.By comparing the shape of strip edges, we found that the edges of the rocker and other defects usually can be used as a very important feature points of image processing.By comparing the image feature matching effect and edge effect, it is concluded that a kind of image feature image matching algorithm based on Phase correlation image matching in order to overlap the matching fusion region strip image. According to the overlap of strip edge we can know the coordinates; According to the strip edge texture features, we got a improved Canny edge detection algorithm and get the final edge according to the strip property and the prior knowledge; Then I realize the parallelized algorithm on the GPU, at the same time I realize the detection of edge based on Canny in GPU.The architecture of the system are discussed in detail, the core module and the corresponding software implementation.The core of the module code of the shape of strip edge detection based on video system was tested respectively. I also test the system for all kinds of strip steel edge shape for recognition and detection and make a comparison with serial program at the same time.The experiments proved that the system based on the parallel implementation of plate edge shape in the video detection greatly improve the processing speed and improve the real-time performance of plate edge process in the video.