多特征融合和深度信念网络的红外弱小目标跟踪方法  

Infrared dim small target tracking method based on multi feature fusion and deep belief network

作　　者：段继光[1] 周玉宏[2] DUAN Jiguang;ZHOU Yuhong(Institute for Nationalities Attached to Hebei Normal University,Shijiazhuang 050091,China;College of Mechanical and Electrical Engineering,Hebei Agricultural University,Baoding 071001,China)

机构地区：[1]河北师范大学附属民族学院,石家庄050091 [2]河北农业大学机电工程学院,河北保定071001

出　　处：《激光杂志》2023年第7期143-148,共6页Laser Journal

基　　金：河北省高等学校科学技术研究重点项目(No.ZD2020405);河北省高等学校科学技术研究指导项目(No.Z2017105)。

摘　　要：研究基于多特征融合和深度信念网络的红外弱小目标跟踪方法,实时掌握目标运动状况,为有效处置目标突发情况提供可靠依据与参考。利用FPGA技术采集红外弱小目标图像,并通过FPGA完成采集图像降噪处理,采用双矩形窗口提取图像特征,并使用基于协方差矩阵的多特征融合策略获取融合结果并将其输入到深度信念网络模型中,在此基础上对红外弱小目标进行识别与检测,并通过Camshift算法明确质心所处方位,最终完成红外弱小目标跟踪。实验结果表明:误差结果约为0.1~0.15,处理图像帧数为150帧,说明本红外弱小目标检测、跟踪效果较好、误差小、效率高。Research on the infrared weak and small target tracking method based on multi-feature fusion and deep belief network,grasp the target movement status in real time,and provide a reliable basis and reference for effectively dealing with target emergencies.Using FPGA technology to collect infrared weak and small target images,and complete the noise reduction processing of the collected images through FPGA,use double rectangular windows to extract image features,and use the multi-feature fusion strategy based on covariance matrix to obtain the fusion results and input them into the deep belief network model On this basis,the infrared weak and small targets are identified and detected,and the position of the centroid is clarified through the Camshift algorithm,and the infrared weak and small target tracking is finally completed.The experimental results show that the error results are about 0.1~0.15,and the number of processed image frames is 150 frames,indicating that the infrared weak and small target detection and tracking effect in this paper is good,the error is small,and the efficiency is high.

关 键 词：多特征融合 深度信念网络 红外弱小目标 目标跟踪 FPGA技术 协方差矩阵 

分 类 号：TN919.81[电子电信—通信与信息系统]