基于计算机视觉的鹦鹉行为分析关键技术的研究

Key Technology Research of Parrot Behavior Analysis Based on Computer Vision

由于某些鸟类对环境比较的敏感可作为观察环境某种变化的生物指示器,通过观察和分析鸟类的行为可以感知到环境发生的一些变化.为了能够自动的识别和分析鸟的行为,选取了鹦鹉作为视觉观察对象,设计了一个基于计算机视觉的系统,能够智能的识别鹦鹉形态和跳动的行为.首先,采用改进型的背景差分算法和阴影去除和抖动的消除等方法,来精准地提取鹦鹉前景对象;然后,结合朴素贝叶斯的方法来识别鹦鹉的各种姿态;最后,根据鹦鹉的形态、位移等特征的变化信息,判断鹦鹉的跳动的行为.实验研究结果表明,提出的识别方法能有效地提取出观察对象、识别出鹦鹉的姿态和跳动的行为,且具有较高的识别率及鲁棒性,为下一步深入的研究鹦鹉行为等研究打下了坚实的基础.

Due to some birds are sensitive to the environment,so we can use the birds as a biological indicator,by observing and analyzing of the behavior of the birds we can sense some changes in the environment.In order to be able to automatically identify and analyze the behavior of birds,we select the parrot as the observation object,and designed a system based on computer vision to recognize parrot＇s shape and beating behavior intelligently.Firstly,using the improved background subtraction algorithm and the shadowremoval and jitter elimination methods,to extract the parrot foreground object more accurately;then combined with the naive Bayesian method to identify the parrot various postures;finally,according the parrot＇s shape and displacement information to judge the parrot＇ s beating behavior.Experimental results showthat the proposed identification method can effectively extract the observation object and identify the parrot＇s all sorts of postures and beating behavior,and has a high recognition rate and robustness.This work also has laid a solid foundation for the next step research.