地球信息科学学报 ›› 2021, Vol. 23 ›› Issue (5): 773-784.doi: 10.12082/dqxxkx.2021.200709

• 地球信息科学理论与方法 • 上一篇    下一篇

动态权重约束下的农村机耕路提取方法

戴激光1,2(), 王晓桐1,2,*(), 智新宇1,2, 马榕辰1,2, 张依蕾1,2   

  1. 1.辽宁工程技术大学测绘与地理科学学院,阜新 123000
    2.辽宁工程技术大学交通时空大数据研究中心,阜新,123000
  • 收稿日期:2020-11-23 修回日期:2021-01-26 出版日期:2021-05-25 发布日期:2021-07-25
  • 通讯作者: 王晓桐
  • 作者简介:戴激光(1978— ),男,黑龙江双鸭山人,博士,副教授,主要研究方向为遥感影像信息提取。E-mail:daijg03@163.com
  • 基金资助:
    国家自然科学基金项目(42071428);国家自然科学基金项目(42071343);辽宁省教育厅服务地方项目(LJ2019FL008);自然资源部测绘科学与地球空间信息技术重点实验室经费资助项目(2020-3-5)

An Extraction Method of Rural Mechanically Cultivated Road Under Dynamic Weight Constraint

DAI Jiguang1,2(), WANG Xiaotong1,2,*(), ZHI Xinyu1,2, MA Rongchen1,2, ZHANG Yilei1,2   

  1. 1. School of Geomatics, Liaoning Technical University, Fuxin 123000, China
    2. Institute of spatiotemporal transportation data, Liaoning Technical University, Fuxin 123000, China
  • Received:2020-11-23 Revised:2021-01-26 Online:2021-05-25 Published:2021-07-25
  • Contact: WANG Xiaotong
  • Supported by:
    National Natural Science Foundation of China(42071428);National Natural Science Foundation of China(42071343);Liaoning Provincial Department of Education Project Services Local Projects(LJ2019FL008);Key Laboratory of Surveying and Mapping Science and Geospatial Information Technology of Ministry of Natural Resources(2020-3-5)

摘要:

宽度窄、路面材质与农田差异性低是农村机耕路的特点,也是导致现有模板匹配方法自动化程度低的主要因素。本文针对这一问题,提出动态权重约束下的农村机耕路提取方法。该方法首先通过改进多尺度线段方向直方图(Multi-Scale Line Segment Orientation Histogram, MLSOH)模型,对机耕路局部道路方向进行预测,能够降低由于田埂干扰所导致的道路方向错误预测几率;其次为明确表征机耕路线性特征,对影像进行线段提取。并将局部区域线段长度作为权重动态分配的主要因子,对不同道路预测方向进行动态权重分配,以此解决路面宽度窄导致的匹配准确度下降问题。最后,将HSL色彩空间相似性分析模型与动态权重因子进行结合,构成HSL动态匹配模型,以此提高机耕路与农田之间的对比性。本文以3幅不同地区、不同类型的高分辨率影像为实验数据,与其他多个模板匹配算法进行对比分析,结果表明,本文方法道路提取的完整度、正确率以及质量均在95%以上。同时相对于其他方法,在保证机耕路提取精度的基础上,本文提出的方法具有自动化程度高的优点,并且本文的方法也适用于其他农村区域道路。

关键词: 机耕路, 权重分配, 动态约束, HSL色彩空间, 欧式距离, 模板匹配, 道路提取

Abstract:

The agricultural machinery field work has developed rapidly. There is an urgent need for more accurate Mechanically Cultivated Road (MCR) network data in agricultural production scheduling. Thus, it is necessary to obtain accurate and effective rural MCR information. However, compared with other types of roads, the narrow pavement width and the small difference between pavement material and farmland are the typical characteristics of rural MCR, which are the main factors leading to the low degree of automation in existing template matching methods. In order to solve the problems mentioned above and improve the accuracy of the MCR extraction, the solutions are proposed as follows: Firstly, by improving the Multi-Scale Line Segment Orientation Histogram (MLSOH) model, we can not only predict the local road direction of MCR, but also reduce the probability of wrong prediction of road direction due to the interference of ridges. Secondly, the line segments of the image are extracted, which can clearly characterize the linear characteristics of MCR. The length of line segments in the local area is taken as the main factor of the dynamic weight distribution. The dynamic weight distribution is carried out for different road prediction directions, so as to solve the problem of the decrease in matching accuracy due to the narrow width of the MCR. Finally, the similarity analysis model of HSL color space is combined with the dynamic weight factor to form the HSL dynamic matching model to improve the contrast between the MCR and the farmland, so as to increase the accuracy of the MCR extraction. In this paper, in order to verify the effectiveness of the proposed algorithm, three high-resolution remote sensing images of different regions and data types are acquired. Two GF-2 images, with a spatial resolution of 0.8 m, covered areas in Tongliao City, Inner Mongolia, and areas in Enshi City, Hubei Province, respectively. One Geo-Eye image, with a spatial resolution of 0.5 m, covered the town of Hobart, Australia. Through qualitative and quantitative analysis of the proposed and comparison algorithms, the conclusions are as follows: the road extraction integrity, accuracy, and quality of the proposed algorithm are all above 95 %. The proposed algorithm has the advantage of high automation while ensuring the extraction accuracy of MCR. It can also be extended to other rural roads.

Key words: Mechanically Cultivated Road (MCR), weight allocation, dynamic constraint, HSL color space, euclidean distance, template matching, road extraction