基于计算机视觉的动物跛脚行为识别  

Lameness Recognition of Animals Based on Computer Vision

作　　者：钱建轩 朱伟兴[1] QIAN Jian xuan;ZHU Wei xing(School of Electrical and Information Engineering,Jiangsu University,Zhenjiang 212013,China)

机构地区：[1]江苏大学电气信息工程学院,江苏镇江212013

出　　处：《软件导刊》2018年第10期10-13,共4页Software Guide

基　　金：国家自然科学基金项目(31172243);江苏高校优势学科建设工程项目(苏政办发[2011]6号)

摘　　要：为了识别出动物正常行走和跛脚行走,利用步态能量图特征描述和降维思想,提出一种基于计算机视觉的动物跛脚行为识别方法。以猪为研究对象,首先通过图像预处理获得目标猪二值图像,并进行步态周期检测和图像归一化处理。然后合成猪的步态能量图(PGEI),利用二维主成分分析(2DPCA)方法对其降维。最后使用最近邻分类器识别出猪的跛脚行为。通过猪的步态数据库进行试验,最终识别率达到93.25%,说明该方法可以有效识别出猪的跛脚行为。该研究为采用计算机视觉技术识别动物的异常步态和跛脚行为提供了一种新思路。In order to identify the normal walking and lame walking of animals, using the idea of gait energy image description and feature reduction, we proposed a method based on computer vision. Pigs were taken as the research object. Firstly, the binary image of the target pig was obtained by preprocessing, and the gait cycle detection and image normalization were conducted. Then the pig gait energy image （PGEI） was calculated,and two dimensional principal component analysis （2DPCA） was used to reduce the dimension of it. Finally, the nearest neighbor classifier was used to recognize the lameness of pigs. The samples in the pigs＇ gait database were tested using the above method. The recognition rate is 93.25%. It shows that this method can effectively identify the lame behavior of pigs. This study provides a new idea for identifying animals with abnormal gait and lame ness by using computer vision technology.

关 键 词：计算机视觉 步态能量图 二维主成分分析 动物跛脚行为 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]