摘 要：目前，尽管采用无人机巡检，航电枢纽库区非法捕捞现象依然严重。库区管理人员无法对非法捕捞人员取证。为了解决这一问题，本文提出将视频图像目标检测技术应用到无人机库区巡检中，对无人机采集的图像进行处理，获得证据。首先，估计初始化背景图像；其次，采用色度相似性估计的方法将图像背景与前景运动目标更清晰分割；最后，为了抑制背景噪声，采用双阈值法判别前景和背景，从而提取运动目标，锁定证据。通过在两个无人机库区巡检采集的视频中试验，该算法能够实现目标分割。目标检测技术可以很好地应用在无人机库区巡检中，配合管理人员取证。
关键词：信息技术；非法捕捞取证；目标检测；视频图像；巡检
中图分类号：TP391.41；U268.6 文献标识码：A 文章编号：2096-4706（2018）01-0024-03
The Application of Video Image Target Detection Technology in the Inspection of the Reservoir Area
WANG Junwen，AN Xiaogang，CHENG Weiping
（China Waterborne Transport Research Institute，Beijing 100088，China）
Abstract：At present, despite the use of unmanned aerial vehicle inspection，navigation and hydropower reservoir area illegal fishing phenomenon is still serious. The manager of the reservoir area was unable to obtain evidence of illegal fishing personnel. In order to solve this problem，the video image target detection technology is applied to the UAV reservoir area inspection，and the images collected from UAV are processed to get evidence. First，estimate to initialize the background image； secondly，using the colorimetric method of the similarity estimation of target image background and foreground segmentation more clearly； finally，in order to suppress the background noise，distinguishing foreground and background by using the dual threshold method to extract the moving target locking evidence. The algorithm can achieve target segmentation by testing the video captured in two unmanned aerial vehicles （UAV） area. The target detection technology can be well applied to the inspection of the UAV reservoir area，and can be used with the managers to obtain evidence.
Keywords：information technology；illegal fishing evidence；target detection；video image；inspection
参考文献：
[1] 程为平，安小刚，王俊文. 基于大顶子山库区无人机库区巡航应用方案研究 [J]. 科技资讯，2017，15（1）：11.
[2] Felzenszwalb P F，Girshick R B，Mcallester D，et al. Object detection with discriminatively trained part-based models[J].Computer，2014，47（2）：6-7.
[3] Girshick R，Donahue J，Darrell T，et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation [J]. 2013：580-587.
[4] Huang H，Li S，Tang K，et al. Accurate segmentation of moving objects and their shadows via brightness ratios and movement patterns [C]// Intelligent Vehicles Symposium Proceedings. IEEE，2014：820-826.
[5] Cucchiara R，Grana C，Piccardi M，et al. Detecting objects，shadows and ghosts in video streams by exploiting color and motion information [C]// International Conference on Image Analysis and Processing，2001. Proceedings. IEEE，2001：360-365.
[6] Kyungnam Kim，Thanarat H. Chalidabhongse，D a v i d Ha r w o o d . R e a l - t ime f o r e g r o u n d - b a c k g r o u n d segmentation using codebook model [J].Real-Time Imaging，2005，11（3）：172-185.
作者简介：王俊文（1990 －），男，河南方城人，研究方向：港口、航电枢纽工程中的图像检测与识别。