低成本受控环境下的果蝇识别算法研究  

Study on Drosophila Recognition Algorithm in Controlled Environment of Low Costs

作　　者：车向前[1] 谷源涛[1] 

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《计算机仿真》2016年第11期199-202,223,共5页Computer Simulation

摘　　要：用计算机技术对实验动物进行自动识别与分析可以极大地提高动物行为学的研究效率。针对传统算法仅适用于高成本、低噪声环境的缺点,提出了一种能够在低成本受控环境下对果蝇个体进行识别,并自动分析其行为的算法。算法通过对容器和果蝇的建模,使用自适应目标大小的背景模型识别果蝇个体,并通过姿态匹配算法确定果蝇的姿态,进而提取特征,用分类算法分析果蝇的行为。仿真结果表明,上述算法可以正确地识别不同姿态的果蝇个体,并能在此基础上对果蝇的打架行为进行分析。上述算法可以直接应用于动物行为学的实验分析,也可以拓展到针对其它目标的识别领域。Automatically recognizing and analyzing experimental animals by computer technology can significantly improve the efficiency in research of ethology. Aiming at the shortages of traditional algorithms which can only be used in the environment of high costs and low noises, we propose an efficient algorithm which can automatically recognize flies and analyze their behaviors in controlled environment of low costs. The proposed algorithm uses size adaptive background model to recognize flies by modeling the container and flies themselves. The posture of fly is determined by posture matching algorithm and its features are extracted to classify behaviors. Simulation results show that the proposed algorithm can recognize flies with different postures correctly and can analyze fight in drosophila. The proposed algorithm can be applied to the analysis of ethology experiments or to the recognition of other objects.

关 键 词：背景模型 果蝇识别 行为识别 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP391.9[自动化与计算机技术—计算机科学与技术]