基于几何特征和贝叶斯的运动目标分类识别方法

Moving objects classification and recognition method based on geometrical features and Bayes

针对传统基于几何特征的运动目标分类识别方法在模式类预定义、特征提取利用和分类器判定策略上的一些细节处理不足,提出一种改进方法。依据目标外轮廓形态差异程度,在模式类下定义子模式类;利用提取出目标的高维度几何特征向量,通过伪划分方式分组得到若干子特征向量,多方面描述目标;通过分类器和子特征向量组计算的结果,利用综合判定机制,得到最终的分类识别结果。利用目标类别的平均识别率指标进行实验,实验结果表明,该方法对预定义的4个模式类有较好效果。

Aiming at some shortcomings about pattern class pre-definition,feature extraction utilization and classifier decision strategy using the traditional methods of moving object classification and recognition based on geometrical features,an improved method was proposed.The sub pattern classes were defined from pattern classes depending on external contours morphological differences of objects.High dimensional geometry feature vector of object and some sub feature vectors were obtained through pseudo division,thus,the object was described in many ways.The final result of classification and recognition was figured out through the results of classifier and sub feature vectors and comprehensive decision rules.The average recognition rates of object classes were used to do the experiment.Results show that this method has excellent recognition rates in four predefined patterns.