白芦笋采收机器人视觉定位与采收路径优化方法

Visual Positioning and Harvesting Path Optimization of White Asparagus Harvesting Robot

依据笋芽出土状态的选择性收获是目前白芦笋公认的最佳收获方式。针对采收过程中机器视觉识别笋尖存在笋尖与垄面纹理和颜色相近等识别难题,本研究提出了一种变尺度感兴趣区域(ROI)检测方法,融合图像色域变换、直方图均值化、形态学和纹理滤波等技术,研究了笋尖识别与精准定位方法;在定位多笋尖坐标基础上,提出了多笋芽的采收路径优化方法,解决了因采收路径不合理导致的采收效率低的问题。首先,通过机器人视觉系统实时采集采收区域图像并进行RGB三通道高斯滤波,采用HSV色域变换并进行直方图均值化处理。在此基础上,对笋尖、土壤进行特征聚类分析,根据笋芽抽发程度研究变尺度ROI检测方法,对采集图像中笋尖的形态学以及笋尖和土壤的纹理进行统计学分析,设定笋尖的似圆度阈值,并参考纹理特征参数,判定笋尖位置,计算其几何中心,获得笋尖轮廓中心坐标。其次,为实现白芦笋的高效采收,根据多目标点与集箱点的位置分布,本研究设计了一种基于多叉树遍历的采收路径优化算法,以获得多个目标笋尖的最优采收路径。最后,搭建采收机器人试验平台开展了笋尖定位与采收验证性试验。结果表明,视觉系统对白芦笋的识别率可达98.04%,笋尖轮廓中心坐标的定位最大误差X方向为0.879 mm,Y方向为0.882 mm,采收笋的个数在不同情况下,采用路径优化后的末端执行器运动距离平均可节省43.89%,末端执行器定位成功率达到100%,在实验室环境下的白芦笋采收率达到88.13%,验证了采用视觉定位的白芦笋采收机器人选择性采收的可行性。

For white asparagus selective harvesting is the best harvesting method determined by its growth characteristics.Focusing on the difficulties that the texture and the color of shoot tips are similar with ridge surface under machine vision,the recognition method of asparagus shoots and precise positioning were studied in this research.A changeable scale ROI detection method was proposed,with the fusion of color transformation,histogram averaging,morphology and texture filtering.After that,a harvesting path optimization method of multiple asparaguses was proposed,which solved the problem of harvesting efficiency reduction caused by unreasonable harvesting paths.Firstly,real-time acquisition of the image and individual RGB channel Gaussian filtering were implemented.Based on the HSV color transformation and histogram averaging processing,the asparagus shoot and soil feature clustering analysis were carried out.According to the sprout degrees of asparaguses,the changeable scale ROI detection method was studied.The morphology and the texture of the shoot,and soil were statistically analyzed.According to the texture feature parameters,the position of shoot was determined and its geometric center was calculated.Secondly,in order to improve harvesting efficiency,a path optimization algorithm based on multiple asparaguses was designed according to the locations of the asparaguses and the bins to obtain the optimal harvesting path.Finally,in order to verify the reliability of the proposed methods,asparagus shoot location and harvest verification tests were carried out on the established harvesting test platform.The results showed that the recognition rate of white asparagus in the visual system was more than 98.04%,the maximum positioning error of the center coordinate of the white asparagus shoot was 0.879 mm in X direction and 0.882 mm in Y direction,and the average reduction of end-effector motion distance could be 43.89%after path optimization under different circumstances,the success rate of end-effector localization was 100%and the harvest rate of white asparagus in the laboratory test was 88.13%.The research verified the feasibility of the visual positioning and harvesting path optimization of the white asparagus selective harvesting robot.