广义Hough变换算法的分析改进

为了改进广义Hough(generalized Hough transform,GHT)变换的虚假投票问题,先分析广义Hough变换的投票机制,给出理想的投票机制模型,用该模型对现有的传统GHT的投票机制进行了评价,指出在检测不规则多边形图形时产生虚假投票的原因,即由于建立参考表R是以梯度为索引项,使得在不同梯度值间的条目数分布不均,在向参数空间映射时会造成大量的投票冗余.针对此分析结果提出一个改进的GHT算法.改进的GHT算法在建立参考表时采用了几何特征角来代替梯度,使得建立的R表中各梯度索引项的条目数分布更均匀,减少了参数空间的投票冗余,提高了算法的效率.实验证明该改进方法是有效的.

Analysis and improvement on generalized Hough transformation algorithm

In order to solve the problem of ambiguous votes of GHT(generalized Hough transform),the principle of voting in GHT was analyzed and a new model of vote was presented.With this model,the voting mechanism of the traditional GHT was appraised and the reason for ambiguous votes was identified.Because gradient is the index item in creating the reference table,the numbers of items with concrete gradient values are not uniform.These results in the large vote redundancy in the parameter space.Based on this,an improved GHT algorithm was presented.The new algorithm used a geometrical feature angle to replace gradient,which makes the numbers of items distribute rather evenly,reduces vote redundancy in the parameter space and improves the algorithm'S efficiency.Experiments show that this modified method is effective.