摘要
对复杂混合部位姿态图像进行识别,在人员定位及识别方面有着重要的应用价值。由于复杂混合部位姿态图像相对普通姿态图像特征要复杂,使得局部动态变化也会给复杂混合部位姿态识别带来很大麻烦。传统的识别算法对于复杂混合部位姿态图像特征只能通过外部轮廓进行简单的识别,对于细致的姿态特征需要经过大量的迭代计算进行识别,导致识别准确性低、效率差。提出基于支持向量机的复杂混合部位姿态图像识别方法。通过高斯模型和背景差方法相结合对复杂混合部位的姿势图像进行姿态分离,引入支持向量机对分离出来的人物姿态图像进行分类,并将分类结果与多级二叉树的形式相结合,构建复杂混合部位姿态图像分类器,完成对复杂混合部位的姿态图像识别。仿真实验结果表明,改进识别算法的准确率和效率均优于传统的识别算法,具有一定的实用性。
In order to improve the recognition accuracy and efficiency, a recognition method for gesture image of complex mixed parts based on the support vector machine is proposed. Firstly, the Ganssian model is integrated with the background subtraction method to make gesture separation for gesture image of complex mixed parts. Then, the support vector machine is introduced to classify the separated figure gesture image. Finally, the classification results are integrated with the multilevel binary tree form to build the gesture image classifier of complex mixed parts, and the gesture image recognition of complex mixed parts is completed. The simulation results show that the accuracy and efficiency of modified algorithm is superior to traditional recognition algorithm.
出处
《计算机仿真》
CSCD
北大核心
2016年第11期365-368,共4页
Computer Simulation
关键词
复杂混合部位
姿态
图像识别
Complex mixed part
Gesture
Image recognition
