针对变长夹角链码对曲线近似时的角度信息损失问题,提出了一种变长夹角链码的改进方法,并将该方法应用于码头目标的特征提取。该方法改变了变长夹角链码算法在折线终点选取时的规则,在继承变长夹角链码优点的基础上,保留了曲线曲率较大的拐角,略去了曲线的较小波动,使有较小波动的曲线近似为直线。改进后的方法使折线能更好地逼近曲线,应用折线逼近并表示曲线,有利于提取曲线的角度特征和线特征。应用该方法提取图像特征,用于码头目标的检测,针对水陆分割后的海岸线使用改进的夹角链码提取海岸线的几何特征。实验表明,该方法能够有效地提取码头的直角、平行线特征,再依据码头先验知识,标记出码头区域。
This paper proposed an improved length variable angle chain code algorithm for the angle information loss of curve approximation and used this algorithm in dock detection. The new algorithm changes the rules in the selection of the end point and retains the advantages of length variable angle chain code ; when occupying the same storage quantity, it can achieve better effect on retaining the comer of larger curvature. This method leaves out the small floatation of curve and it has a positive effect on angle feature and linear feature extraction. The authors used the method to extract the feature of image for dock detection, extracted the geometric feature based on the coastline, and marked out the dock areas combining the geometric feature with the existing knowledge. Experiments show this method can effectively extract right angle and parallel features.