基于图像处理的生猪耳部颜色异常检测技术

Pig Ear Abnormal Color Detection on Image Processing Techniques

为研究规模化生猪养殖场中蓝耳病疫情预警监测方法,提出一种非接触式耳部颜色自动检测方法。该方法将生猪热红外图像和可见光图像相结合寻找最优尺度因子,确定可见光图像中生猪耳根部特征区域;采用主动形状模型方法,选取34个生猪耳部轮廓特征点,并将搜索范围限定在生猪头部区域,用以提取生猪耳部轮廓;将提取的耳部轮廓进行颜色对比,判断该生猪是否患有蓝耳病疫情。结果表明,由于限定生猪头部区域搜索范围,能快速准确地提取生猪耳部轮廓。对生猪耳部颜色检测准确率达到77%以上。

In order to study blue ear epidemic early warning and monitoring method in large-scale pig farms,the non-contact ear automatic color detection method was proposed.Pig＇s thermal infrared images and visual images were collected at the same time and the same viewing angle.Two pictures were used together to find the optimal scale factor of matching.By the optimal scale factor,pig ear root section can be found in visible image,and then the ear root central point can be confirmed.According to active shape model method,pig ear root central point was selected as the first feature point,the ear tip as 18 th feature points,and both middle point of outlines as 9th and 26 th feature points,and 34 pig ears outline feature points in all were selected by human-computer interaction.ASM search scope was defined in pig head region,thus pig ear outline could be extracted correctly.Then the extracted pig ear color was compared with the color of normal pig ears,the ear color detection accuracy could be above 77%,and it could be easily found whether there was the risk of blueear pig disease.The results showed that due to the limited search scope method,the pig ear contour could be extracted accurately,and it could be applied to auto ear color detection in swine house.