一种基于背景模型的果蝇轮廓提取算法  

ADrosophila Contour Extraction Algorithm Based on Background Model

作　　者：侯奇 谷源涛[1] HOU Qi;GU Yuan- tao(Department of Electronic Engineering, Tsinghua University, Beijing 100084, China)

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《计算机仿真》2018年第6期296-299,共4页Computer Simulation

摘　　要：果蝇轮廓提取是果蝇行为自动识别的重要步骤,其结果会影响果蝇特征的提取和行为识别的准确率。提出一种基于背景模型的果蝇轮廓提取算法,可以从视频中提取果蝇的身体和翅膀。算法采用单高斯模型作为背景模型,通过亮度畸变区分果蝇翅膀和身体。初步建模后分离果蝇身体和活动台背景;然后在对活动台单独进建模的基础上,得到精确的背景模型用于轮廓提取。实验表明,算法对不同拍摄场景的视频表现良好,具有较高的准确性和鲁棒性。上述算法可以推广应用于动物行为学研究对其它动物的实验分析。Drosophila contour extraction plays an important role in the automatic recognition of drosophila behavior, which affects the accuracy of behavioral analysis. In this paper, we propose a contour extraction algorithm using background model to extract the body and wings of drosophila from a video. A single Gaussian distribution algorithm was used as the background model, and the drosophila wings and body were distinguished through luminance distortion. After the preliminary background modeling, we can separate the drosophila body from the containers. Then we carried out background modeling on the container only, which makes a better background model for drosophila contour extraction. Experiments show that the algorithm works well at different scenes with high robustness. The above algorithm can be further applied to experimental analysis of other species in animal behavior studies.

关 键 词：背景模型 轮廓提取 果蝇识别 

分 类 号：P301.6[天文地球—地球物理学] TP391.9[自动化与计算机技术—计算机应用技术]