| The applications for stacked stainless steel pipes automatic counting is in the blank stage now. Because the stainless steel pipe has mirror like surface, for saving the material the pipe wall always very thin so the edge in cross section is also very thin and difficult to detect in image. General in a real working field the illumination is not uniform and steel pipes place in irregularly. Addition all about, the stainless steel pipes automatic counting is very difficult.Image processing algorithm and system implementation for stainless steel pipes auto-count is researched in this paper. Achieving the counting purpose through the recognition of the steel pipe’ section image, the main research content of algorithm includes image pretreatment, image division and pipe number counting. An automatic counting system for Stainless Steel Pipes Auto-count is developed based on the algorithm. And the paper has launched research work in the following aspects:1. The RGB color image is converted to grayscale image by using the weighted average method, and the image which has the non-uniform light is calibrated. Due to the incomplete and thin edge in the image after detecting edge, a novel convolution template is prompted for enhancing the steel pipe edge.2. The image of stainless steel pipes section is divided into two images by establishing threshold. The erosion operation can make each adhesion region contract to small regions separated from each other, and then non-adhesion regions can be got which are restored from those small regions by using the watershed method. Merge segmentation results of the two images.3. Mark all connected domain and get the number, and the basic characteristics of properties connected domain can be calculated at same time. Target areas can be recognized under getting minimum circumscribed circle area of the connected domain. Count the target areas and achieve the stainless steel pipes count at last.4. Build the hardware and design the software, then the stainless steel pipes counting system is implemented by programming in MATLAB.Using the system to count the total number in a real environment, the accuracy reaches94%. The system has application value. |