摘要
针对油菜直播机视觉导航路径识别效果受天气、稻茬噪声等影响的难题,提出一种结合小波变换和改进随机抽样一致性(RANSAC)的导航路径识别方法。首先,对原始图像灰度化后进行小波变换,在大尺度低分辨率下凸显导航路径宏观轮廓;然后利用直播机作业区与未作业区图像对比度大的特点获取导航路径上的特征点集合;最后针对获取的特征点集合运用结合预检验和后处理校正的改进随机抽样一致性算法区分内外点,并对内点集运用最小二乘法进行导航路径直线拟合,从而获取导航路径参数。田间图像测试表明,该方法可以稳定、准确地检测出导航路径,正确率达到96.7%,同时每帧图像的处理时间在31 ms以内,能为油菜直播机的视觉导航提供技术支撑。
In practical applications, the complexity of the farmland environment including weather and rice stubble noise has resulted in bad effect of vision navigation of rapeseed direct seeder.To solve this problem, a method of identifying the vision navigation of rapeseed direct seeder based on wavelet transform and improving random sample consensus （RANSAC） was proposed.The original image grayed was wavelet transformed to highlight the macro profile of the navigation path.The feature of work area and non-work area with high-contrast in the work images of rapeseed direct seeder was used to find the feature points of navigation path through ROI sliding windows.The improved RANSAC with correction of pre-test and post-process was used to distinguish inner points and outer points in the set of navigation path feature points, and the navigation path parameters was calculated by least squares in the set of inner points.Results of field experimental tests showed that the algorithm could detect stably and accurately the navigation path with the correct rate of 96.7 %.The processing time for each frame was less than 31.0 ins.It will provide technical support for identifying the vision navigation of rapeseed direct seeder.
出处
《华中农业大学学报》
CAS
CSCD
北大核心
2016年第3期128-133,共6页
Journal of Huazhong Agricultural University
基金
国家油菜产业技术体系专项(CARS-13)
国家公益性行业(农业)科研专项(201503116-6)
武汉市高新技术产业科技创新团队项目(2014070504020240)
关键词
油菜直播机
视觉导航
小波变换
随机抽样一致性
最小二乘法
rapeseed direct seeder
vision navigation
wavelet transform
random sampling consis-tency
least squares
作者简介
周雅文，硕士研究生．研究方向：图像识别和机器视觉导航．E-mail：zyw19900927@163．com
通信作者：丁幼春，博士，副教授．研究方向：油菜机械化生产装备智能化技术．E-mail：kingbug163@163．com
