Design And Tests Of Autonomous Following Method For The Master-slave Vehicles System Based On Stereo Vision

As the main carrier,the agricultural vehicle has been involved in every aspect of agricultural production.The practical processes of agricultural production need master-slave vehicles to complete related tasks.In the production process of medium-sized farms,it is necessary and important to apply the master-slave vehicles to ensure the quality and effect of production.The labor shortages and labor costs has increased in current situation of rural countryside.Hence,how to improve operational efficiency and productivity of equipment has become an urgent problem to be solved.Due to the current limit level of intelligent technology,the security and intelligence of autonomous navigation of agricultural vehicles can't be effectively guaranteed.There is still a large gap between operating effects of agricultural vehicles and actual production requirements of agriculture.Based on the above description and consideration,The master-slave operating mode with few people has been seriously discussed and technically observed to realize multi-car collaboration by many researchers.Therefore,it has been regarded as the important research direction to realize the intelligence and automation of agriculture by many agricultural engineering scholars.Master-slave operation with many vehicles should solve the follow problem between the master-vehicle and slave-vehicle.Then a follow method of master-slave vehicles based on stereo vision system has been proposed in this paper.Autonomous follow test of master-slave vehicle in the orchard has been observed with two electric vehicle platform to make a valid attempt to realize master-slave operating of vehicles.Research contents and results are summarized as below:1.A follow method of master-slave vehicles based on typical feature extraction has been proposed.The checkerboard feature board is the typical calibration object of the stereo vision system's parameter.It is choosed as the guide feature of the master-vehicle to simplify the feature description of the master-vehicle which is conducive to the implementation of the algorithm in this paper.Checkerboard feature board has a strong adaptability for background interference and light environment.Stereo vision can be technically applied to effectively extract three-dimensional information of checkerboard feature board's each corner point.By constructing the kinematic model of master-slave vehicles,three-dimensional information of checkerboard feature board's each corner point can be soundly converted to relative pose parameters between master-slave vehicles to achieve the autonomous follow of slave vehicle.2.The stereo vision system based on PC platform has been developed:in order to quickly verify the feasibility of the method,the stereo vision system based on PC platform was developed firstly which took VS2010 as the software development environment and took the open source computer vision library-OpenCV as a primary source of image processing technology.The upper-position computer software system which was a set of image acquisition and processing,calibration,matching,3D reconstruction and other functions was built,with which the 3D information of inner-corner points in the checkerboard feature board was extracted in real time and the parameters of the relative position and oriention were calculated by these 3D information.The entire system data update frequency was 4Hz and sent the parameters of the relative position and orientation to the vehicle control system by serial port to control the corresponding actuator3.The stereo vision system based on embedded system platform has been developed:in order to realize application that this system was mounted on vehicles,reduce system cost and improve the system's applicability,the related processing algorithms and functions are transformed and transplanted to the embedded system based on the development of PC platform.This stereo vision system included the embedded development board produced by SEED and two universal miniature surveillance cameras which had a good environmental adaptability.The OpenCV was transplanted for embedded system.The development of upper-position computer interfaces took QT as the software development environment.This stereo vision system realized the real-time 3D information extraction of inner-corner points in the checkerboard feature board and transmitting the parameters of the relative position and orientation which is calculated by 3D information of inner-corner points to the vehicles control system by serial port.4.The performance of stereo vision system based on PC platform and embedded system platform was tested by static tests:the stereo vision systems based on two platforms could generate stable and accurate parameters of relative position and orientation and have good adaptability for different light intensity.The average output frequency of stereo vision based on PC platform was 4Hz,while the average output frequency of stereo vision system based on embedded system could reach 0.77 Hz.The above tests showed that the method of extracting the vehicle posture information based on stereo vision proposed is feasible.If the hardware system has enough computing speed,the algorithm can be used in embedded integration applications.5.The information collection and control system for the master-slave electric vehicles has been developed:in order to apply stereo vision system to actual vehicles operation,a group of master-slave electric vehicles platform was selected as the test vehicle and the information collection and control system was developed.This system was based on extensible CAN-bus structure.Each CAN node took 32-bit micro controller-STM32F103VET6 as the main controller,and extended multiple input and output interfaces used as PWM outputting,pulse acquisition,AD input,DA output,serial port and CAN bus interface.This system has good applicability and reprehensibility which can collect speed,steering and other information of vehicle operation and control speed,steering and other implementing agencies through communicating with the stereo vision system.6.The vehicle safety redundancy system based on wireless communication has been developed.The wireless transmission modules were used to exchange operating state information between two vehicles.Vehicle start command,stop command,emergency stop command,cooperative control communication command were defined for preventing the stereo system failure and other emergency situations.At the same time,vehicle control strategy is proposed when the target is lost during the following movement,which provides redundancy protection for the safe operation of the vehicle.7.The follow system of master-slave vehicles was tested in actual orchard environment.During the process of tests,the output frequency of relative position and orientation parameters was 4Hz,the vehicle speed was set to 2km/h,the target follow distance was 1.6m,the target lateral offset was Omm,target heading angle was 0°,the length of test path in orchard is 25m.Throughout the testing process,the average deviation of the lateral offset was 3.68cm,the average deviation of the distance was 8.24cm,the average deviation of heading angle was 2.320.GPS track record showed that the traveling path of two vehicles was substantially coincident.The results showed that the proposed method could guide slave-vehicle to obtain the relative position and orientation information effectively which could guide the automatic follow movement in master-slave vehicles system.