一种改进的应用于高速目标角点检测的AFJHA算法

A Developed AFJHA Algorithm Applied to Detect Corners on High-speed Target

针对经典Harris角点检测算法对于全局像素点求取角点响应函数所带来的计算复杂度高,从而不适合实时检测的弊端,提出了一种新的快速角点检测算法。鉴于角点数量只占整个图像像素点很小一部分,首先利用四方向亮度导数来进行候选角点预判断,阈值选取为图像像素点像素值的标准差的倍数。经过预判断后,再考虑8-邻域相似点个数,根据相似点个数对应的情况对候选角点进行再判断,将剩下的候选角点进行Harris算法检测。仿真结果表明,此改进算法能够有效地提高角点检测效果并减少检测时间。

Because the classic Harris comer detection algorithm calculates the corner response functions of whole pixels,it has the disadvantages of bringing about high computational complexity and is not suitable to real-time detection, a new fast comer detection algorithm was introduced to solve this problem. Given the quantity of comer makes up only a small part of the whole pixels, use four-direction brightness derivative to preliminary judge candidate comers, the threshold is the multiples of all pixels＇ standard deviation. After preliminarily judge, consider 8-neighborhood similarity numbers. Then, according to the numbers to re-judge candidate comers, use Harris algorithm to detect the rest of candidate comers. The experimental results showed that, the developed algorithm can improve the efficiency of its system and shorten running time of corners detection.