智能全景视觉传感网络视频多目标跟踪仿真  

Simulation of Video Multi-Target Tracking Based on Intelligent Panoramic Vision Sensor Network

作　　者：高智勇 乔姝函 GAO Zhi-yong;QIAO Shu-han(College of Information Science and Engineering,Shandong Tai'an 271018,China)

机构地区：[1]山东农业大学信息科学与工程学院,山东泰安271018

出　　处：《计算机仿真》2023年第1期223-226,238,共5页Computer Simulation

摘　　要：传统视频多目标方法忽略了对网络视频多目标状态的核验,导致出现效率偏低、误差较大等问题。提出智能全景视觉传感网络视频多目标跟踪方法。利用卷积神经网络,分割网络视频图像多目标跟踪信息,通过选择网络视频图像多目标特征,得到网络视频图像的投影矩阵。基于网络视频图像中多目标特征的状态,核验多网络视频多目标状态,以特征函数的状态作为当前网络视频图像的目标特征,将传感网络视频多目标特征作为状态向量,计算并归一化处理状态向量的权值。通过智能全景视觉传感网络中每个视频对应的图像,检测出多目标元素,关联状态向量权值和目标位置,完成网络视频多目标的跟踪。实验结果表明,所提多目标跟踪方法具有较高的多目标跟踪成功率和正确率,增强了多目标跟踪效果。Traditionally, some methods ignore the verification for the multi-target state of network video, resulting in low efficiency and large error. Therefore, this paper presented a multi-target tracking method for intelligent panoramic vision sensor network video. At first, the convolution neural network was used to segment the multi-target tracking information of video images. After selecting the multi-target features, the projection matrix of the network video image was obtained. Then, the multi-target state of the multi-network video was verified according to the state of multi-target features in network video images. Moreover, the state of the feature function was taken as the target feature of the current image, and the multi-target feature of the sensing network video was used as the state vector. Furthermore, the weight of the state vector was calculated and normalized. Based on the image corresponding to the video in the intelligent panoramic vision sensor network, the multi-target elements could be detected. Finally, the state vector weight was associated with the target position. Thus, the multi-target tracking of network video was completed. Experimental results show that the proposed method has a high success rate and accuracy during multi-target tracking, and enhances the tracking effect.

关 键 词：全景视觉 目标跟踪 目标区域 特征提取 传感网络 视频图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]