摘要
基于经典Hough变换提出了一种改进的随机Hough变换算法,该算法可以大大提高识别速度和检测精度,减小存储空间,降低运算量,且仍具有经典Hough变换对噪声不敏感,随机Hough变换随机抽样和多对一收敛映射的特点,实现了一种逐次提取曲线的方法,将曲线按照由长到短的顺序依次提取出来,并给出了曲线的起始点和终止点的计算方法,实验证明了方法的有效性。在此基础上,还对形状不变性特征进行了分析,得出图形平移、旋转、缩放后,Hough变换峰值的个数没有改变,对角度归一化后,峰值的初始位置有所变化,但峰值间的间隔没有变化。
A improvement stochastic Hough transformation algorithm is proposed based on the classics Hough transformation. This algorithm may greatly enhance the recognition speed and the examination precision, reduce the storage space, reduce the operand, and still has characteristics of the classics Hough transform, such as not to be insensitive to noise, stochastic Hough transformation random sampling and multitude to one mapping. One kind of method gradually to withdraw the curve is realized from long to short withdrew the curve in turn. The computational method of the curve initial and terminal station are given, the experiment has proved the method validity. On the basis of experiment the shape invariable characteristic is analysed after the graph translation, revolves, scaling, the Hough transformation peak value quantity do not change, after the angle normalization, the peak value initial point changes, but the peak value gap does not change.
出处
《传感器与微系统》
CSCD
北大核心
2007年第5期86-89,共4页
Transducer and Microsystem Technologies
关键词
随机HOUGH变换
曲线识别
形状检测
模式识别
randomized Hough transformation
curve recognition
shape detection
pattern recognition