基于视觉伺服的联合收割机群协同导航从机定位方法

Slave positioning method for cooperative navigation of combine harvester group based on visual servo

针对收割机群协同导航从机实际定位需求,在分析主机与从机间相对位置关系的基础上,提出了一种采用视觉系统求解靶标投影矩形区域形心坐标,并以此作为导引信息,引导相对位姿测量装置测量从机相对主机位姿,实现从机精确定位的方法。重点对从机定位方法中需解决的运动靶标跟踪和对靶控制2个关键问题进行了研究,提出了一种基于双重波门的运动靶标跟踪方法和一种基于变尺度变论域模糊控制的对靶控制方法。试验结果表明,该文所提定位方法基本不受收割机前进速度的影响,不同速度最大偏差的标准差为0.003 5 m,平均偏差的标准差为0.002 3 m。

With the improvement of the farm mechanization level,the more corporate＇ style farming emerges.For example,more than one harvester collaborated with each other completes the task of harvesting operations.The new corporate style farming creates several new challenges for the agricultural machinery navigation,and the group navigation is the trend of agricultural machinery navigation technology.Group positioning and collaborative navigation control are the 2 critical technical problems to be resolved.In order to solve the slave positioning problem,a slave positioning method based on visual servo was proposed.Firstly,the combine group based on mater-slave structure was introduced,and the slave positioning scheme was proposed.In order to fulfill the automatic measurement of the slave position,the relative positioning device was designed.The structure and measurement process of the positioning device were introduced in detail.The positioning device was composed of laser ranging sensor,PTZ（Pan Tilt Zoom） controller and monocular camera.The monocular camera was used to guide the laser ranging sensor to aim at the target.The laser ranging sensor was used to measure the distance between the master and the slave.The laser ranging sensor guaranteed high precision of the measurement data.The position measurement process could be divided into 2 steps.The first step was to measure the relative position posture between the master and the slave.According to the deviation between the target＇s center and the image center,the system adjusted the platform＇s direction to change the posture of the camera,so that the 2 centers in the target and the image could coincide approximately.After aiming at the target,the laser sensor was triggered to measure the distance between the master and the slave.The second step was to calculate the slave positon.A global localization model was built up to show the relative location relationship between the master and the slave,and the slave＇s coordinate under global coordinate system could be calculated.Secondly,in order to solve the 2 key problems in the process of the position measurement,2 methods were proposed.A method of automatic identification for motion target based on dual windows was proposed to reduce the visual feedback delay.This target identify method filtered out most of the identified area by setting dual windows,and searched the target point only in a small area of the image.A variable-scale variable-universe fuzzy control method was proposed to improve the target control accuracy.This method introduced the contraction-expansion factor to improve the granularity of fuzzy rules without increasing the number of control rules,and introduced the variable scale factor to improve the adaptability to the change in the distance between the master and the slave.Finally,in order to verify the effectiveness of the proposed model and method in this paper,the slave positioning experiments were carried out.When the slave ran at the speed of 0.8,1.0 and 1.2 m/s,the average positioning error was 0.127 9,0.128 6 and 0.132 1 m respectively.The experimental results show that the slave positioning accuracy is independent of the forward speed,and can meet the slave positioning requirement.