The Research And Development Of The Bar Counting Based On Image Processing

Automatic counting bar steel has been the problems which steel bars production enterprises do not resolve well. At present, companies mainly rely on manual measurement to complete the task of bar steel counting. Sometimes a group of bars on the need for regular counting and verification, this is heavy workload. And long time counting work will cause worker physical fatigue and bring large error. So the development of the automatic counting bar system is an urgent need to be resolved. It mainly study on the bar identification and count system from following parts in this paper:image pre-processing, image edge detection, identification of the target image. In the image pre-processing methods, it introduced processing methods based on mathematical morphology:removing image background noise,image enhancement,image smoothing,image sharpening,image filtering,image segmentation. It makes the image pre-processing effect better. It also make the identification of target easy. At the same time, it also resolved the uneven image binarization caused by uneven illumination in the images collection process, which image enhancement and image sharpening are main problems in study. Through the use of the image histogram equalization, it make the region of original image gray-scale concentrated tensile or gray-scale distribution. And it also make the pixel gray-scale dynamic range increasing. So the image contrast become increasing and the details of image become clear.It sharpening the bar image using spatial domain Laplacian operator for round bar with a real-time processing. It strengthen the bar-end part of the border. Then it give the binarization processing to the image using image threshold segmentation method. Since under the conditions of uniform illumination, the bar's gray value of the surface is similar, and it has a significant differences in background. So it extracts bars from the background using the image gray value as a threshold segmentation method. It analyzes the bar image using several typical edge. Through the experimental comparison, Canny operator performs better. Finally, it tests circular targets. Basing on some recognition algorithms, such as the type of round heart,the circular edge and radius. it meets certain requirements of the independent regional as class circle. As a result of uneven distribution of bar, it will have bar adhesion. It gives larger errors to identify work. So it need give image segmentation to partial image in order to improve the correct recognition rate of bar.Take bar-end image processing as an example, it gave a bar identification method with the relevant experiments. It have a high ability to identify. It laid the foundation for the bar of rapid extraction and recognition count.