文中针对如何快速检测圆形目标展开研究,在随机霍夫变换的基础上提出了两步筛选法和区域约束法,对可能存在的错误采样进行了分析,利用图像的梯度信息和圆的几何特性筛选采样点,避免了大量的无效累积;同时通过设计检验窗口,减少了真实性检验的计算量,有效提高了检测速度和精度。该算法应用于圆形目标实时检测系统,在实时性与准确性上取得了良好的效果。
This thesis researches on how to detect the circle target fast,and based on Randomized Hough Transform it puts forward two-step selecting method and region-restraining method.The methods analyze the possible errors and select sampling points by gradient messages of the image and geometrical property of the circle so that the methods avoid a large amount of useless accumulation.Meanwhile,the methods decrease the calculation of factuality test by the design of inspection window and thereby improve the detecting speed and accuracy.The methods are used in the real-time detecting system of the circle target and get good results in real-time and accuracy.