樱桃小番茄腋芽摘除机器人关键技术研究

发布时间：2018-03-29 07:54

本文选题：小番茄　切入点：腋芽摘除　出处：《浙江大学》2016年博士论文

【摘要】：樱桃小番茄腋芽摘除是提高产量的关键,由于腋芽颜色特征和周围枝条、叶片相近,且容易被遮挡,本文以小番茄植株为对象,以成功摘除腋芽为目标,研究了图像采集、去除点定位,侧枝水平投影角测量等关键问题。主要研究内容与成果：(1)研究了农艺技术,使用半剪的方式摘除樱桃小番茄的腋芽。测量了定植后的樱桃小番茄植株几何特性,还测定了其腋芽、侧枝的力学特性,获得了樱桃小番茄的株高、主茎直径、节距、侧枝朝向、倾角等特征参数。提出在温室大棚中应标准化种植,按照固定的行距、株距、畦距等要求,为机器人自动摘除腋芽提供一定基础。(2)研制了蓝色LED光源染色系统,照射小番茄植株的侧枝基部,将其用蓝光染色,方便进行图像分割。建立了蓝色LED可调式染色系统,通过光照度传感器检测周围环境的照度,调节灯珠的亮度,防止发生高光反射或者染色不足。对RGB颜色空间的B分量进行图像分割,得到目标部位清晰完整的图像。(3)通过快速傅里叶变换(FFT),使用低通滤波器去除毛刺和噪声,保留基本轮廓特征；由形态学膨胀算法突出腋芽两侧特征点,通过Shi-Tomasi角点检测算法,找到目标图像角点,再经过特征点判别算法,找到特征点,由此判别腋芽存在与否,定位腋芽去除点,最后摘除腋芽。实验结果表明,腋芽识别的成功率为93.9%,腋芽摘除成功率为88.9%,能满足自动去除的要求。(4)设计了光电式自动旋转测向机构。采用闭合环形机构环绕樱桃小番茄主茎,并通过旋转方式使固定在活动环上的光电传感器能水平360。扫描侧枝,得到侧枝的水平投影方向角。分析活动环的运动轨迹,得到其线速度与角速度的关系及比例系数。经试验表明：比例系数为1.5,角速度为1.875π/s,线速度为25mm/s时,检测成功率为97%,角度标准差为±8.6。。高度补偿设定为20mm,成功率为91%。对42株樱桃小番茄的153个侧枝进行测试,在1.8m高度内的成功率为88.2%。(5)标定了工业摄像头,标定结果误差在2mm之内。分析了单目摄像头测距的不足,使用超声波测距传感器与摄像头组成定位系统,并对超声波测距传感器进行标定,根据其相对摄像机坐标系的波束向量。(6)完成了腋芽摘除机器人末端执行机构的整体设计和装配。确定了基于导轨的机械导航方式,并使用二次定位法,使机器人准确定位樱桃小番茄植株位置。分析了机械臂的正逆运动。进行摘芽整体试验,总成功率为83%。
[Abstract]:Axillary bud extraction is the key to increase the yield of cherry tomato. Because of the color characteristics of axillary bud and the surrounding branches, the leaves are similar and easily blocked, this paper studies the image collection of the small tomato plant as the object and successfully picking the axillary bud as the target. The main contents and results of this study were as follows: 1) the agronomic technique was studied, and the axillary buds of cherry tomato were removed by semi-cutting. The geometric characteristics of the plants were measured. The results were as follows: (1) the main contents and results were as follows: (1) the agronomic technique was used to remove the axillary buds of the cherry tomato, and the geometric characteristics of the plants were measured. The mechanical properties of axillary bud and lateral branch were also determined, and the plant height, diameter of main stem, pitch, lateral branch orientation and inclination angle of cherry tomato were obtained. The blue LED light source dyeing system was developed to illuminate the lateral branch base of tomato plant and dye it with blue light. A blue LED adjustable dyeing system is established to detect the illumination of the surrounding environment and adjust the brightness of the lamp beads. The B component of the RGB color space is segmented to get a clear and complete image of the target part. FFT3 is used to remove burr and noise through fast Fourier transform (FFT), low pass filter (LPF) is used to remove burr and noise. The morphological expansion algorithm is used to highlight the characteristic points on both sides of the axillary bud, and the corner point of the target image is found by the Shi-Tomasi corner detection algorithm, and then the feature point is found through the feature point discrimination algorithm, and then the axillary bud is judged to exist or not. Location of axillary bud removal point, and finally removal of axillary bud. The successful rate of axillary bud recognition was 93.9, and the successful rate of axillary bud removal was 88.9.The photoelectric automatic rotation direction finding mechanism was designed to meet the requirement of automatic removal. The closed ring mechanism was used to surround the main stem of cherry tomato. And the photoelectric sensor fixed on the active ring can be horizontal 360 by rotation. The horizontal projection direction angle of the lateral branch is obtained by scanning the lateral branch, and the moving track of the active ring is analyzed. The relationship between the linear velocity and the angular velocity and the proportional coefficient are obtained. The experimental results show that the ratio coefficient is 1.5, the angular velocity is 1.875 蟺 / s, and the linear velocity is 25mm/s. The successful rate of detection was 97 and the standard deviation of angle was 卤8.6.The height compensation was set to 20mm and the success rate was 91. The industrial camera was calibrated by testing 153 branches of 42 cherry tomatoes at a height of 1.8m. The error of calibration results is within 2mm. The shortcomings of monocular camera ranging are analyzed. The ultrasonic ranging sensor and camera are used to form a positioning system, and the ultrasonic ranging sensor is calibrated. According to the beam vector of relative camera coordinate system, the whole design and assembly of the terminal actuator of axillary bud removal robot are completed. The mechanical navigation mode based on guide rail is determined, and the secondary positioning method is used. The robot is able to locate the position of cherry tomato plant accurately. The forward and inverse movement of the robot arm is analyzed. The whole experiment of picking buds shows that the total power is 83%.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S225.92;TP242

【相似文献】