Safety Warning And Obstacles Avoidance System Of Tractor Using Multi-information

China has put forward higher requirements for reducing the labor intensity of agricultural operations and improving the efficiency of agricultural production as China begins to speed up the process of agricultural development.As a consequence,we must vigorously promote agricultural mechanization,intelligence,and improve the level of agricultural machinery automation.Safety early warning and obstacle avoidance strategies are also very important for intelligent agricultural machinery in order to improve the safety of agricultural machinery in the process of driving.At present,the two problems are that the information of obstacles obtained by a single sensor is incomplete and there are few studies on the safety warning model suitable for agricultural machinery.Based on the Oubao 4040 tractor equipped with lidar,camera,Beidou navigation equipment,inertial measurement unit and angle sensor,with the help of multi-sensor information fusion technology,the research on obstacle detection,state determination,tractor safety early warning and obstacle avoidance strategy in different states was carried out in the unstructured farmland environment.The main contents and results of the study are as follows:(1)The tractor information collection and control platform is constructed.Oubao 4040 tractor is selected as the construction platform,and the environmental information sensing system with lidar and camera is developed;The tractor information acquisition system with Beidou navigation equipment and front wheel angle sensor is built;The safety warning system with Arduino microcontroller,relay and LED alarm light are constructed.The construction of tractor information collection and control platform provides an implementation platform for verifying the obstacle detection method and safety early warning level division model.(2)A method of obstacle detection and state determination is proposed.The DBSCAN clustering method is selected to cluster the point cloud data of the lidar to realize the detection of obstacles;Obstacles are tracked by Kalman filtering and minimum cost maximum flow algorithm to obtain the information of the obstacle,and the tractor state information acquisition system is used to obtain the information of the tractor.By analyzing the relative motion relationship between the obstacle and the tractor,the obstacle state is judged.When the difference between the speed of the obstacle relative to the tractor and the speed of the tractor is less than the threshold k,the obstacle is judged to be static;On the contrary,the obstacle is judged to be dynamic.The field test results show that the accuracy of the system for judging obstacles in different states is more than 82.9%,indicating that the method is feasible.(3)A classification model of tractor safety early warning level is proposed.The corresponding security early warning level division models are proposed in this model according to the obstacles in different states.For the static obstacle,taking whether the tractor can turn around the obstacle at the maximum angle as the critical condition,a total of three warning levels of level 1,level 2 and level 3 are established.For the dynamic obstacle,the hazard assessment index δ based on the relative motion relationship between the obstacle and the tractor and the safe braking distance of the tractor is proposed,and the three-level warning level is determined.Several groups of field tests are carried out for static and dynamic obstacles respectively,and the accuracy of the test results of the two states of obstacles is more than 82.1%,which basically meets the test requirements,thus verifying the effectiveness of the security early warning level division model proposed in this paper.Finally,it provides a basic guarantee for the proposed obstacle avoidance strategy.(4)The data fusion of lidar and camera is completed and an obstacle avoidance strategy considering multiple factors is proposed.YOLOv3 algorithm is selected to train the target recognition model suitable for the field environment;The information fusion of lidar and camera data are completed by using the projection method,The categories of single and multiple obstacles and the corresponding distance and position information can be obtained;An obstacle avoidance strategy considering the driving scene of the tractor,the state of the obstacle,the early warning level of the obstacle and the category of the obstacle is proposed.The fusion system is tested for both single and multiple obstacles,and the results show that the error is less than 3.7%.(5)The obstacle state judgment and tractor safety early warning system are integrated and field tests are carried out.The obstacle detection system and tractor safety early warning system are integrated and debug,and the performance of the system is tested through field tests.As a result,it verifies the effectiveness of the obstacle state judgment and tractor safety early warning system proposed in this paper.