基于改进Faster R-CNN识别深度视频图像哺乳母猪姿态  

Lactating sow postures recognition from depth image of videos based on improved Faster R-CNN

作　　者：薛月菊 朱勋沐 郑婵 毛亮 杨阿庆 涂淑琴[4] 黄宁 杨晓帆 陈鹏飞 张南峰[5] Xue Yueju;Zhu Xunmu;Zheng Chan;Mao Liang;Yang Aqing;Tu Shuqin;Huang Ning;Yang Xiaofan;Chen Pengfei;Zhang Nanfeng(College of Electronic Engineering, South China Agricultural University, Guangzhou 510642, China;Guangdong Engineering Research Center for Datamation of Modern Pig Production, Guangzhou 510642, China;Guangdong Engineering Research Center for Information Monitoring in Agriculture, Guangzhou 510642, China;College of Mathematics and Informatics, South China Agricultural University, Guangzhou 510642, China;Guangzhou Entry-Exit Inspection and Quarantine Bureau, Guangzhou 510623, China)

机构地区：[1]华南农业大学电子工程学院,广州510642 [2]广东省现代养猪数据化工程技术研究中心,广州510642 [3]广东省农情信息监测工程技术研究中心,广州510642 [4]华南农业大学数学与信息学院,广州510642 [5]广州出入境检验检疫局,广州510623

出　　处：《农业工程学报》2018年第9期189-196,共8页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家科技支撑计划(2015BAD06B03-3);广东省科技计划项目(2015A020209148);广东省应用型科技研发项目(2015B010135007);广州市科技计划项目(201605030013);广州市科技计划项目(201604016122)

摘　　要：猪舍场景下,昼夜交替光线变化、热灯光照影响,及仔猪与母猪的粘连等因素,给全天候哺乳母猪姿态自动识别带来很大困难。该文以深度视频图像为数据源,提出基于改进Faster R-CNN的哺乳母猪姿态识别算法。将残差结构引入ZF网络,设计ZF-D2R网络,以提高识别精度并保持实时性;将Center Loss监督信号引入Faster R-CNN训练中,以增强类内特征的内聚性,提升识别精度。对28栏猪的视频图像抽取站立、坐立、俯卧、腹卧和侧卧5类姿态共计7 541张图像作为训练集,另取5类姿态的5 000张图像作为测试集。该文提出的改进模型在测试集上对哺乳母猪的站立、坐立、俯卧、腹卧和侧卧5类姿态的识别平均准确率分别达到96.73%、94.62%、86.28%、89.57%和99.04%,5类姿态的平均准确率均值达到93.25%。在识别精度上,比ZF网络和层数更深的VGG16网络的平均准确率均值分别提高了3.86和1.24个百分点。识别速度为0.058 s/帧,比VGG16网络速度提高了0.034 s。该文方法在提高识别精度的同时保证了实时性,可为全天候母猪行为识别提供技术参考。The maternal behaviors reflect the health and welfare of the sows, which directly affect the economic benefit of the pig farm. Computer vision provides an effective, low-cost and non-contact method for monitoring the behavior of animal for precision farming. Under the scene of piggery, it is a challenge for 24-hour automatic recognition of lactating sow postures due to the daily illumination variations, influence of heat lamp, and adhesion between piglets and sows. This paper proposed an automatic recognition algorithm of lactating sow postures based on improved Faster R-CNN（convolutional neural network） using depth video images. To improve the recognition accuracy and satisfy the real-time need, we designed a ZF-D2 R（ZF with deeper layers and 2 residual learning frameworks） network by introducing residual learning frameworks into ZF network. First, 3 convolutional layers were added in the ZF network to design ZF-D（ZF with deeper layers）. Then, in ZF-D network, shortcut connections were used to form 2 residual learning frameworks. The whole network made up the ZF-D2 R network. Moreover, the Center Loss was introduced to Fast R-CNN detector to construct a joint classification loss function. With the joint supervision signals of F-Softmax Loss and Center Loss in Fast R-CNN detector, a robust model was trained to obtain the deep feature representations with the 2 key learning objectives, which led to intra-class compactness and inter-class dispersion as much as possible. So, the joint supervision of F-Softmax Loss and Center Loss could reduce recognition errors caused by the similar features between different postures. By taking ZF-D2 R as basic net and adding the Center Loss to Fast R-CNN detector, the improved Faster R-CNN was built. Experiments to obtain the actual data set of lactating sow posture from the depth video of sows in the 28 pens were performed. The data set included 2 451 standing images, 2 461 sitting images, 2 488 sternal recumbency images, 2 519 ventral recumbency images and 2658 later

关 键 词：图像识别 算法 模型 FASTER R-CNN 残差结构 CENTER LOSS 哺乳母猪 姿态识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]