基于热红外成像与骨架树模型的奶牛眼温自动检测
Automatic Detection of Dairy Cow’s Eye Temperature Based on Thermal Infrared Imaging Technology and Skeleton Tree Model
-
摘要: 针对现有方法无法实现奶牛热红外图像中眼温信息自动获取的问题,为实现奶牛眼温无接触、自动、高精确检测,提出了一种基于热红外成像技术与骨架树模型的奶牛眼温自动检测方法。首先,在获得奶牛侧面热红外图像的基础上,利用基于差距度量的阈值分割方法提取奶牛目标,对奶牛骨架进行精确提取,并构建了奶牛骨架树模型,在该模型上对奶牛头部区域进行准确定位;然后,根据头部轮廓的形状特征与眼睛几何位置特征,对奶牛眼睛区域中心点进行准确定位;最后,以眼睛中心点为圆心,以半径为20像素区域内的最高温度作为眼睛温度,对奶牛热红外图像中眼温进行自动检测。为验证本文方法的有效性,随机选取来自50头奶牛的100幅侧视热红外图像进行了试验,结果表明,采用本文方法检测结果的平均绝对误差为0.35℃、平均相对误差为0.38%,具有较高的精度。本研究可为奶牛体温非接触、自动化、高精度检测提供技术支撑。Abstract: The maximum temperature of the dairy cow’s eye area is highly correlated with the widely used rectal temperature. The existing methods have not been able to automatically extract the eye temperature from the thermal infrared image. In order to achieve non-contact,automatic and highprecision detection of dairy cow’s eye temperature,a method for automatic detection of dairy cow’s eye temperature based on thermal infrared imaging technology and skeleton tree model was studied and proposed. On the basis of the thermal infrared image of the side of dairy cow,the threshold segmentation method based on the gap measurement was used to extract the dairy cow target,and the precise extraction of the dairy cow skeleton was realized and the dairy cow skeleton tree model was constructed. In this model,the head area of dairy cattle was accurately located, and then according to the shape characteristics of the head outline and the geometric position characteristics of the eyes,the center point of the eye area of dairy cattle was accurately located. Finally,the eye temperature was automatically detected in the thermal infrared image of dairy cow with the center point of the eye as the center and the highest temperature in the 20 pixels radius area as the eye temperature. In order to verify the effectiveness of this method,totally 100 thermal infrared images from 50 dairy cows were randomly selected for the test. The results showed that the average absolute error of eye temperature was 0. 35℃,and the average relative error was 0. 38%. The method had high accuracy and can provide technical support for the noncontact,automatic and high-precision detection of dairy cow temperature.