典型猪舍光环境下机器视觉图像量化评价及筛选  

Quantitative assessment and screening of images in lighting environment of typical piggery

作　　者：杜晓冬[1] 滕光辉 李卓[1] 石晨[1] Du Xiaodong Teng Guanghui Li Zhuo Shi Chen(College of Water Conservancy and Civil Engineering, China Agricultural University, Key Laboratory of Agricultural Engineering in Structure and Environment, Beijing 100083, China)

机构地区：[1]中国农业大学水利与土木工程学院,农业部设施农业工程重点实验室,北京100083

出　　处：《农业工程学报》2017年第2期213-219,共7页Transactions of the Chinese Society of Agricultural Engineering

基　　金：“十二五”国家科技支撑项目(2014BAD08B05)

摘　　要：随着数字化农业进程的发展,利用机器视觉技术进行生猪养殖方面研究已成为许多研究人员关注的热点之一。机器视觉中外界光环境对采集图像的质量具有较大影响,但是较多猪体质量监测方面研究的焦点集中在图像算法方面,忽视外界光环境的重要性,缺乏量化评价机器视觉光环境优劣的方法。该文基于猪体质量检测平台,利用LabVIEW软件编程分析实际光环境条件下拍摄图像质量。经现场试验分析,理想图像、曝光过度图像、自然光图像、阴阳图像4类图像在曝光参数和照度均匀度参数有明显区别:理想图像和曝光过度图像的照度均匀度参数均满足照明工程标准要求,理想图像曝光正常比率较高;自然光图像和阴影图像的照度均匀度参数不满足照明工程标准要求,且曝光正常率偏低。该文采用曝光正常,最小照度与最大照度比值U1>0.7和最小照度与平均照度比值U2>0.8判断实现理想图像筛选,便于研究者前期图像预处理工作。With the advancement of digital agriculture, researching on pig weight estimation by using machine vision technology has become one of hotspots. The outside light environment of machine vision has a great influence on the quality of captured images. Ignoring the significance of lighting system, majorities of studies on pig weight estimation focused on algorithms and image processing without quantitative assessment methods of light environment. This research focus on the light environment for machine vision in typical piggery and the purpose is to explore a method to achieve effectively screening images taken by monitoring platform to filter out a large amount of invalid images interfered by outside light environment. This paper, based on pig weight monitoring platform, LabVIEW software was used to analyze the image quality of actual light environment condition. Researches consisted of field test and laboratory test. Light measurement of region of interest was mainly carried out in field test during different breeding periods as well as with different heights of pig body. Laboratory test was divided into two parts. One was the calibration of light sensor and the other was image processing for comparing the difference between the algorithm value and the real value. Experiments were carried out in the experimental station of Shangzhuang of China Agricultural University and the test objects were 5 heads of castration landrace. AS813 illuminance meter of SMART SENSOR Company was used to conduct research in test spot. In order to ensure the data accuracy, it was conducted two times of sensor calibration. The analysis of uniformity of illumination in measurement area was referred to lighting engineering standards for evaluating the intensity of illumination evenness. U1 and U2 parameters were used to evaluate evenness index of illumination intensity. U1 is the ratio of the minimum illuminance and the maximum illuminance. U2 is the ratio of the minimum illuminance and the average illuminance. By means of computer softwar

关 键 词：机器视觉 照明 监测 猪体质量 光环境 LABVIEW 照度均匀度 

分 类 号：S821.46[农业科学—畜牧学]