Video Monitoring Behaviors of Captive-farmed Porcupines
-
摘要: 为了更好地了解豪猪的习性，提高豪猪人工养殖技术水平，本文设计了基于一种视频图像分析的圈养豪猪检测及基本行为识别方案。首先通过混合高斯模型背景建模法，对圈养豪猪养殖环境进行背景建模，标记出场景中的豪猪及其他运动物体轮廓，采用分类算法对识别出的轮廓进行分类，对豪猪的识别准确率达到86.34%；为了进一步提高准确率，引入图像局部特征ORB关键点作为分类属性，使豪猪的识别准确率提升到93.23%；在此基础上，以饲养池结构及豪猪活动实际情况为判断依据建立圈养豪猪行为识别模型，实现了对豪猪静息、进食、饮水、排泄、啃咬铁门及水槽等7种基本行为的识别。Abstract: To understand the living habits for remotely managing the breeding of captive-farmed porcupines, this study applied video to monitor and establish a recognition model with the aid of computation for the behaviors of the animals. Firstly, the mixed Gaussian background modeling was used to build a movement contour model of the porcupines in the pan. Using 3 chosen classifiers, the marked scenes of porcupine activities were categorized with an accuracy of 86.34%. Subsequently, ORB key points were introduced as an additional attribute for the classification which raised the accuracy to 93.23%. The resulting model could now recognize 7 basic behaviors, including resting, eating, drinking, excretion, and chewing an iron gate or a water trough, of porcupines in captivity.
-
图 6 4种特征点检测结果对比
注：A为Shi-Tomasi，B为SURF，C为SIFT，D为ORB。
Figure 6. Results of 4 key point detection algorithms
图 7 ORB关键点检测及分组
A为输入图像，B为ORB特征点检测，C为特征点提取，D为特征点分组。
Figure 7. ORB key point detection and classification
表 1 4种特征点算法运行时间
Table 1. Running time of 4 key point detection algorithms
算法 Shi-Tomas SIFT SURF ORB 时间/ms 31416.3 602.262 126.081 20.7 
下载: 导出CSV
表 2 几种分类算法的识别准确率比较
Table 2. Comparison of recognition accuracy among classifiers
(单位/%) 分类算法 1只豪猪 2只豪猪 3只豪猪 平均值 J48 86.10 89.40 72.90 86.43 J48C 92.30 96.00 91.40 93.23 BAYESNET 81.20 89.60 81.40 83.39 BAYESNETC 83.90 94.50 94.30 87.07 C-SVM 85.50 88.10 4.30 82.93 C-SVMC 88.40 84.10 80.00 86.95
下载: 导出CSV
-
[1] 陆明洲, 沈明霞, 丁永前等.畜牧信息智能监测研究进展[J].中国农业科学, 2012, 45(14):2939-2947. doi: 10.3864/j.issn.0578-1752.2012.14.017 [2] OCZAK M, ISMAYILOVA G, COSTA A, et al. Analysis of aggressive behaviours of pigs by automatic video recordings[J]. Computers & Electronics in Agriculture, 2013, 99(C):209-217. [3] OTT S, MOONS C P H, KASHIHA M A, et al. Automated video analysis of pig activity at pen level highly correlates to human observations of behavioural activities[J]. Livestock Science, 2014, 160(1):132-137. [4] 朱燕. 基于时空兴趣点的猪的跛脚行为识别[D]. 镇江: 江苏大学, 2016. [5] 劳凤丹, 滕光辉, 李军, 等.机器视觉识别单只蛋鸡行为的方法[J].农业工程学报, 2012, 28(24):157-163. http://d.wanfangdata.com.cn/Periodical/nygcxb201224024 [6] 何东健, 孟凡昌, 赵凯旋, 等.基于视频分析的犊牛基本行为识别[J].农业机械学报, 2016, 47(9):294-300. doi: 10.6041/j.issn.1000-1298.2016.09.040 [7] 洪留荣.应用轮廓变化信息的实验鼠行为识别[J].计算机工程, 2014, 40(3):213-217, 223. http://d.wanfangdata.com.cn/Periodical/jsjgc201403045 [8] 赵利强. 基于移动轨迹分析的大鼠行为识别研究[D]. 杭州: 浙江大学, 2016. [9] 陈吉宏, 俞守华, 区晶莹.猪舍智能监控系统中猪只识别算法研究[J].广东农业科学, 2011, 38(10):151-153. doi: 10.3969/j.issn.1004-874X.2011.10.056 [10] 陈显周, 俞守华, 区晶莹.异常挖掘在猪只行为数据分析上的应用[J].农业现代化研究, 2011, 32(S1):52-55. [11] 许丹纯, 俞守华, 区晶莹等.可拓分析法在猪场环境安全预警中的应用[J].广东农业科学, 2011, 38(23):160-163. doi: 10.3969/j.issn.1004-874X.2011.23.054 [12] 周勇钧, 俞守华, 区晶莹.多特征Camshift和Kalman滤波结合的猪只智能跟踪[J].广东农业科学, 2013, 40(9):174-177, 188. http://d.wanfangdata.com.cn/Periodical/gdnykx201309049 [13] YU S. Tracking Algorithm Based on Multi-feature Detection and Target Association of Pigs on Large-scale Pig Farms[J]. Journal of Information & Computational Science, 2015, 12(10):3837-3844. [14] SHI J. Good features to track[D].New York:Cornell University, 1993. [15] BAY H, ESS A, TUYTELAARS T, et al. Speeded-Up Robust Features[J]. Computer Vision & Image Understanding, 2008, 110(3):404-417. http://d.wanfangdata.com.cn/Periodical/jsjyyyj201303072 [16] LINDEBERG T. Scale Invariant Feature Transform[J]. Scholarpedia, 2012, 7(5):2012-2021. http://d.wanfangdata.com.cn/Periodical/xtgcydzjs-e201506022 [17] RUBLEE E, RABAUD V, KONOLIGE K, et al. ORB:An efficient alternative to SIFT or SURF[C]//IEEE International Conference on Computer Vision. IEEE, 2011:2564-2571. [18] 陶新民, 曹盼东, 宋少宇, 等.基于半监督高斯混合模型核的支持向量机分类算法[J].信息与控制, 2013, 42(1):18-26. http://d.wanfangdata.com.cn/Periodical/xxykz201301004 -
点击查看大图
计量
- 文章访问数: 1248
- HTML全文浏览量: 134
- PDF下载量: 132
- 被引次数: 0
