基于凸壳式剥离的高黏着度棒材检测计数方法

DETECTION AND COUNTING METHOD FOR BARS WITH HIGH ADHESION DEGREE BASED ON CONVEX-HULL PEELING

高黏着度棒材的精确检测计数是一个难点。提出凸壳式剥离算法:判断画面中是否存在高黏着度棒材。若是,则首先获取棒材端面的边缘轮廓角点和部分中心点,由得到的角点和中心点求出一组能描述棒材端面轮廓的点集,然后依据该点集求出外围棒材端面轮廓的凸壳,从而剥离外围棒材达到准确获取内部棒材中心点的目的;若否,则使用峰值提取法求全部棒材的中心点。实践证明,该方法能有效提升高黏着度棒材识别与计数的精度。

The precise detection and counting of bars with high adhesion degree is a difficult issue. We propose the convex-hull peeling algorithm in the paper. It is to judge whether there are the bars with high adhesion degree in the image. If there are,then the edge contour corners and part of central points on bars＇ end face are obtained first,and from the derived corners and central points the algorithm finds a set of point sets which can describe the end face contour of bars; then according to the point sets it finds the convex hull of end face contour of external bars,so as to peel the external bars off to achieve the goal of precisely obtaining the central points of internal bars. If there are not,then the peak extraction method will be employed to detect the central points of all bars. Practice proves that the method can effectively enhance the accuracy of recognition and counting on bars with high adhesion degree.