【作者】 刘国海； 王振武； 刘慧； 沈跃； 崔业民；

【Author】 Guohai Liu;Zhenwu Wang;Hui Liu;Yue Shen;Yemin Cui;Department of Electrical and Information Engineering,Jiangsu University;Nantong Guangyi Electromechanical co.LTD;

【机构】 江苏大学电气信息工程学院； 南通广益机电有限责任公司；

【摘要】 为了实现农田喷药机器人自主导航,本文提出一种改进的并行细化算法来提取稻田路径。首先把稻田图像的RGB色彩空间变换成HSI色彩空间,分析HSI三个通道图像的特征。对H通道进行阈值分割,得到二值图像,然后对二值图像进行膨胀运算,并运用形态学方法去除噪声。然后对去噪后的路径区域进行凸包变换,最后运用改进的Zhang并行细化算法提取路径。实验结果表明,与传统的最小二乘法相比,本文算法提取出的导航路径准确率达96%,单幅图像平均耗时95ms,实时性更高、鲁棒性更好。适用于多行窄长路径的提取,能够满足喷药机器人在多种环境下实时作业的需求。更多还原

【Abstract】 In order to realize the autonomous navigation of the field spraying robot, this paper proposes an improved algorithm to extract rice paddy fields. First through the on-board monocular camera real-time acquisition paddy images, and then put the rice image of RGB color space to HSI color space, and analysis of H, S, I three channels path characteristics of the image. The H channel image is segmented by a fixed threshold, and the binary image of the rice field path is obtained, and its expansion operation is performed, and the noise is removed by morphological method. Because the path of rice field has narrow and long characteristics, the paper makes convex hull transformation in the path area after denoising, and finally uses the improved Zhang parallel refinement algorithm to extract the center path. The experimental results show that compared with the traditional least square method, this algorithm to extract the navigation path accuracy of 96%, the average processing time for single image 95 ms, higher real-time performance and higher robustness, and is suitable for multiple lines long narrow path extraction, can satisfy the spraying robot real-time operation in a variety of environmental requirements.更多还原