基于机器视觉的农田机械导航线提取算法研究

Research of Farm Machinery Navigation Algorithm Based on Machine Vision

随着科学技术的发展,精细农业已成为现代农业发展的主导方向。农业机械的自动导航技术是精细农业的关键技术之一,也是实施精准农业的基础。机器视觉由于其广泛实用性,已成为农田机械导航线提取的重要方法。目前,机器视觉自动导航线提取易受自然环境干扰,且在实时处理速度上有待提高。为此,研究了一种导航线提取算法,旨在简化图像处理,提高通用性。首先对CCD获取的彩色农田图像,使用改进的过绿色算法进行灰度化,得到目标区分较好的图像;然后使用改进的OTSU算法对图像进行分割,得到二值图像,再采用滤波、腐蚀、膨胀相结合的算法去除图像噪声;最后提取作物行骨架,拟合作物行直线并进行方向校正,计算相机偏差,为实时校正航向提供反馈信息。试验结果表明,该算法处理一幅图像所用时间在200ms左右,可满足农田机械实时导航的要求。

With the development of science and technology, precision agricultural has become the dominant direction of modem agriculture development. Agricultural machinery automatic navigation technology is one of the key technology of precision agriculture, it＇ s the basis for the implementation of precision agriculture. Machine vision has become an impor- tant method to farm machinery navigation line extraction due to its wide practicability. Now, a lot of machine vision auto- matic navigation line extraction algorithms are susceptible to interference, the real-time processing speed needs to be im- proved. This article mainly is to simplify the image processing algorithm and improve its generality. Firstly, the color im- ages captured by CCD are grizzled with the improved algorithm, the target is distinguished well. Then the improved OTSU algorithm is used to get the binary image. Combine filtering, corrosion and expansion algorithm to remove the noise. Then the skeleton of crop line is extracted. , the straight crop line is fitted to adjust the direction, the camera deviation is calculated to provide feedback information for the real-time correction. Experimental results show that the proposed algo- rithm use around 200ms every image, it can satisfy the requirement of real-time navigation.