王侨,刘卉,杨鹏树,孟志军.基于机器视觉的农田地头边界线检测方法[J].农业机械学报,2020,51(5):18-27.
WANG Qiao,LIU Hui,YANG Pengshu,MENG Zhijun.Detection Method of Headland Boundary Line Based on Machine Vision[J].Transactions of the Chinese Society for Agricultural Machinery,2020,51(5):18-27.
摘要点击次数: 2063
全文下载次数: 1015
基于机器视觉的农田地头边界线检测方法   [下载全文]
Detection Method of Headland Boundary Line Based on Machine Vision   [Download Pdf][in English]
投稿时间:2020-02-02  
DOI:10.6041/j.issn.1000-1298.2020.05.002
中文关键词:  机器视觉  农田环境感知  地头自主转弯  地头边界线  地头转向基准线
基金项目:国家重点研发计划项目(2017YFD0700400)、农机北斗导航项目(2017YFD0700402)、北京市博士后工作经费项目(2018-ZZ-061)和中国博士后科学基金项目(2018M641257)
作者单位
王侨 国家农业智能装备工程技术研究中心 
刘卉 首都师范大学 
杨鹏树 国家农业智能装备工程技术研究中心
首都师范大学 
孟志军 国家农业智能装备工程技术研究中心 
中文摘要:在非结构化复杂农田作业环境中,为实现农机在地头处的自主导航转弯,首先需及时、准确地感知地头的空间位置信息,尤其是地头边界位置。本文基于机器视觉技术,首先依据农田内外像素灰度的跳变特征来判断地头是否出现,通过建立正向和负向分布偏差两个度量确定是否存在该灰度跳变特征;随后,将图像沿水平方向平均分成8个子处理区域,针对各子处理区域求取其行灰度平均值分布图,基于局部加权回归法对其进行平滑处理,建立按序离群度参数,通过寻找平滑曲线上首个按序离群程度较大的波峰点或波谷点以及相应的跳前波谷点或波峰点,最终确定跳变特征点的像素坐标,并基于稳健回归法线性拟合跳变特征点,获取实际非规整地头边界的主体延伸方位线;最后,将主体延伸方位线向下平行移动,当其线上像素的灰度平均值接近于田内像素的灰度分布特征时,认为抵达安全位置处,由此获得农机在当前地头处安全转向掉头的边界线。试验结果表明,判断地头出现的准确率不低于96%,地头边界线检测准确率不低于92%。
WANG Qiao  LIU Hui  YANG Pengshu  MENG Zhijun
National Engineering Research Center of Intelligent Equipment for Agriculture;Capital Normal University
Key Words:machine vision  farmland environment sensing  autonomous turning at headland of field  headland boundary line  headland turning baseline
Abstract:In unstructured and complicated filed operation environment, the realization of autonomous navigation turning of agricultural machinery at the headland area of field is one of the key technical bottlenecks for achieving the autonomous navigation walking of agricultural machinery throughout the field. The primary task of realizing the former is to timely and accurately perceive the spatial position information of the headland area, especially the location information of the headland boundary. Based on machine vision technology, whether the headland appeared in the image or not was firstly determined according to the jumping characteristics of the gray values of pixels inside and outside the field. Specifically, two metric values of positive and negative distribution deviations were established to describe the positive and negative dispersion degree between the average pixel gray values of different rows in the image. When one of the two metric values was larger than the judgment threshold, that was, the distribution of the average values was relatively dispersed, it can be considered that the jumping characteristics had occurred and it was judged that the headland was appearing in the image. Subsequently, the image was evenly divided into eight sub-processing regions along the horizontal direction. For each sub-processing region, the distribution curve of the row gray average values was obtained and smoothed by local weighted regression method. The in-order outlier parameter was established, and based on the degree of sequential outlier of the row gray average values corresponding to the peak points or trough points on the smoothed curve, the position coordinates of the jumping peak point and pre-jumping trough point were determined, and accordingly the pixel coordinates of the jumping feature points were determined. Finally, all the jumping feature points were fitted linearly based on the robust regression method to obtain the main-body extended azimuth line of the irregular headland boundary. In the end, the main-body extension azimuth line was moved down in parallel until the average gray value of the pixels on the line was close to the corresponding gray distribution characteristic of the pixels inside the field, which was considered to be shifted to the safe position, thus the boundary line for safe turning of agricultural machinery at the current headland of field was obtained. The test results showed that the accuracy rate of judging whether the headland appeared or not was not less than 96%, the detection accuracy rate of headland boundary line was not less than 92%, and the processing time of single frame image was not higher than 0.52s based on Matlab platform. It can provide a fast, accurate and reliable technical support for agricultural machinery to implement automatic navigation turning safely at the headland of field

Transactions of the Chinese Society for Agriculture Machinery (CSAM), in charged of China Association for Science and Technology (CAST), sponsored by CSAM and Chinese Academy of Agricultural Mechanization Science(CAAMS), started publication in 1957. It is the earliest interdisciplinary journal in Chinese which combines agricultural and engineering. It always closely grasps the development direction of agriculture engineering disciplines and the published papers represent the highest academic level of agriculture engineering in China. Currently, nearly 8,000 papers have been already published. There are around 3,000 papers contributed to the journal each year, but only around 600 of them will be accepted. Transactions of CSAM focuses on a wide range of agricultural machinery, irrigation, electronics, robotics, agro-products engineering, biological energy, agricultural structures and environment and more. Subjects in Transactions of the CSAM have been embodied by many internationally well-known index systems, such as: EI Compendex, CA, CSA, etc.

   下载PDF阅读器