基于多特征融合的红外图像行人检测研究  

Pedestrian detection in infrared image based on multi feature fusion

作　　者：方丽娟 张常友 FANG Lijuan;ZHANG Changyou(Jiangxi University of Engineering,College of Arificial Intelligence,Xinyu Jiangxi 338000,China)

机构地区：[1]江西工程学院人工智能学院,江西新余338000

出　　处：《激光杂志》2022年第3期114-117,共4页Laser Journal

基　　金：江西省教育厅科学技术研究项目(No.181145)。

摘　　要：当前红外图像行人检测方法存在误差大,对外界环境变化鲁棒性差,为了获得理想的红外图像行人检测结果,提出了多特征组合的红外图像行人检测方法。首先对当前红外图像行人检测进展进行分析,找到引起红外图像行人检测效果不佳的因素,然后提取红外图像行人检测的HOG特征和ISS特征,并采用支持向量机进行红外图像行人检测,最后引入证据理论方法对HOG特征和ISS特征的红外图像行人检测进行融合,仿真实验结果表明,光照影响、遮挡影响和复杂背景下,本方法的红外图像行人准确率都超过了99%,明显高于其他红外图像行人检测方法。The current pedestrian detection methods in infrared images have large errors and poor robustness to the changes of external environment.In order to obtain ideal pedestrian detection results in infrared images,a pedestrian detection method based on multi feature combination is proposed.Firstly,the progress of pedestrian detection in infrared image is analyzed to find out the factors that cause poor effect of pedestrian detection in infrared image.Then,HOG feature and ISS feature of pedestrian detection in infrared image are extracted,and support vector machine is used for pedestrian detection in infrared image.Finally,evidence theory is introduced to fuse HOG feature and ISS feature of pedestrian detection in infrared image,and simulation experiment is carried out.The results show that under the influence of illumination,occlusion and complex background,pedestrian detection accuracy of this method is more than 99%,which is significantly higher than other infrared image pedestrian detection methods.

关 键 词：红外技术 图像处理 行人检测 证据理论 仿真测试 

分 类 号：TN216[电子电信—物理电子学]