中文摘要：

为了解决采摘机器人在自然环境中对扰动状态荔枝的视觉精确定位问题，该文分析机械手采摘过程中荔枝产生扰动的因素，设计制造了模拟荔枝振动的试验平台，该试验平台通过改变方向、振频、振幅等振动条件来模拟采摘过程中的扰动环境；结合振动平台运动参数，提出了双目立体视觉系统采集扰动状态的荔枝图像方法，在HSI颜色空间中对预处理后荔枝图像利用模糊C均值聚类法（FCM，fuzzyC-means）分割荔枝果实和果梗，然后利用Hough变换算法进行直线拟合确定有效的果梗采摘区域和采摘点，对多帧图像中采摘点坐标取平均值，然后进行三维重建确定空间采摘点坐标。荔枝扰动状态的视觉定位试验结果表明，空间定位深度值误差小于6cm，荔枝采摘机械手能实现有效采摘，该研究为机械手实际作业提供指导。

英文摘要：

In order to solve the visual accurate positioning problem for litchi in disturbed state in the natural environment, the litchi disturbance factors in manipulator picking process were analyzed, and a simulation experiment platform of litchi vibration was designed and manufactured, which can simulate the vibration environment in the process of being picked based on changing vibration conditions including direction, vibration frequency, amplitude. The collection method of turbulent litchi image in the binocular stereo vision system was put forward using vibration platform motion parameters. Fuzzy C-means clustering method （FCM） was used to the pretreatment litchi image to segment litchi fruit and fruit stems in the HSI color space, and then the Hough transform algorithm was used to linear fitting determine the effective fruit picking area and picking point. Average values for picking point coordinates of multiple frame images were taken and then the space coordinate of picking point was determined using 3D reconstruction. The turbulent litchi visual position experimental results showed that the space position depth data error was less than 6 cm, and the litchi picking manipulator can realize effective picking. The study provides guidance for manipulator practical operation.