基于视频检测的储粮害虫死亡评估算法的研究  

Algorithm of Stored Grain Pest Death Assessment Based on Video Detection

作　　者：周亦哲 周慧玲[1] 王威松 Zhou Yizhe;Zhou Huiling;Wang Weisong(College of Automation,Beijing University of Posts and Telecommunications,Beijing 100089)

机构地区：[1]北京邮电大学自动化学院

出　　处：《中国粮油学报》2019年第10期114-119,127,共7页Journal of the Chinese Cereals and Oils Association

摘　　要：在我国,如何对杀虫效果进行评估,以提高防治措施的有效性和经济性,是粮库在害虫防治方面亟待解决的一个问题。目前,我国一些粮库已经安装了储粮害虫图像采集装置,因此本研究提出一种基于视频检测的储粮害虫死亡评估算法,用来检测杀虫过程中害虫死亡的具体数量变化情况。算法的核心是基于深度卷积神经网络的双流法网络,综合图像目标检测算法和两帧差分法进行识别,实现视频数据中害虫的定位与识别。测试结果表明本算法可有效检测储粮害虫的死亡情况,检测平均正确率可以达到89.3%。In China,how to evaluate the insecticidal effect in order to improve the effectiveness and economy of control measures has been an urgent problem for grain depot in pest control.At present,in China,some image acquisition devices have been installed for stored grain pests in some grain depots.Therefore,the paper proposed a video - based assessment algorithm for stored grain pest mortality,used to detect the specific changes in the number of pest deaths during insecticidal process.The core of the algorithm was the double - stream method network based on the deep convolutional neural network,which integrated the image target detection algorithm and the two - frame difference method for identification to realize the locating and identification of pests in video data.The test results showed that the proposed algorithm can effectively detect the death of stored grain pests,and the average detection accuracy could reach 89.3%.

关 键 词：储粮害虫 卷积神经网络 视频目标检测 双流法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]