警用无人机侦查取证的视频图像特征提取方法  

Video image feature extraction method for police drones investigation and evidence collection

作　　者：陈思信[1] 何凤伟 CHEN Sixin;HE Fengwei(Police Tactics Department,Chongqing Police College,Chongqing 401331,China)

机构地区：[1]重庆警察学院警务战术系,重庆401331

出　　处：《西安工程大学学报》2024年第6期98-104,共7页Journal of Xi’an Polytechnic University

基　　金：重庆市科学技术研究项目(KQNJ202301506);重庆市教育委员会人文社会科学研究项目(24SKGH336);重庆市高等教育教学改革研究项目(243278)。

摘　　要：警用无人机在侦查取证的实际应用中,不仅需要在夜间、雾霾、地形复杂等不利条件下稳定作业,确保侦查取证工作的连续性和可靠性;还需具备良好的兼容性,能够与其他警务系统无缝对接,实现信息共享与协同作战。针对这些要求,给出一种新的视频图像特征提取方法。首先,对视频图像中目标对象的HSV色彩进行详细的分析。然后,基于对HSV色彩空间的分析结果,引入注意力机制进行视频图像特征提取,分别构建引入注意力机制的光谱多尺度特征提取网络和空间多尺度特征提取网络。最后,将光谱与空间网络合并为一个整体,同时进行训练与融合,得到光谱-空间联合特征提取网络,实现警用无人机侦查取证的视频图像特征提取。实验结果表明:文中方法的特征偏离度较小,特征提取精度最高为97%,总体的每秒帧数(frame per second,FPS)值在50~75,整体保持在较高水平,具有较好的稳定性和可靠性。这不仅满足了警用无人机对高效、准确特征提取的需求,还为其在复杂侦查环境中的广泛应用提供了技术支撑。In the practical application of police drones in investigation and evidence collection,it is not only necessary to operate stably under unfavorable conditions such as night,haze,and complex terrain,but also to ensure the operation continuity and reliability.The drones needs to have good compatibility and be able to seamlessly integrate with other police systems to achieve information sharing and collaborative operations.In response to these requirements,a new video image feature extraction method was proposed in the article.Firstly,we made a detailed analysis of the HSV color of the target object in the video image.Then,based on the results of the HSV color space,attention mechanism was introduced for video image feature extraction,and constructed spectral multi-scale feature extraction network and spatial multi-scale feature extraction network with attention mechanism.Finally,the spectrum and spatial network were merged into a whole,and trained and fused to obtain a spectral spatial joint feature extraction network,which achieved the extraction for police unmanned aerial vehicle investigation and evidence collection.The experimental results show that the feature deviation of the method in the article is relatively small,and the highest feature extraction accuracy is 97%.The overall frame per second(FPS)value is between 50~75,which remains at a high level and has good stability and reliability.This not only meets the demand for efficient and accurate feature extraction of police drones,but also provides technical support for their widespread application in complex investigation environments.

关 键 词：警用无人机 视频图像特征提取 注意力机制 HSV色彩空间 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]