双窗目标特征提取算法研究

为了得到更加稳定性和准确性的目标特征,提出了一种双窗目标特征提取算法.首先在目标周围设置两个初始窗,其中小窗作为提出的改进活动轮廓模型的初始曲线;其次根据活动轮廓模型运动规则进行迭代得到目标的轮廓,作为特征提取的内窗.运用加权比例直方图法提取内、外窗的日标特征,根据设定的规则计算目标特征.最后通过实验证明了上述算法相对传统目标特征提取算法能够得到更佳准确的结果.

A Feature Extraction Algorithm Based on Double Window

A feature extraction algorithm using double window is proposed for getting stable and accurate object feature. Two windows are set around the object. The small window is regarded as the initialized curve of the proposed improved snake model. The boundary of the object is fitted by an iterative algorithm based on the moving rule of the snake model, which is the inner window. The features in the inner window and outer window are extracted using the weighted histogram. The object feature is gotten based on the rule. The experiment proved that the algorithm can get better result than traditional feature extraction algorithm.