Trajectory Tracking Algorithm And System Of Intelligent Industrial Vehicle

With the development of intelligence and the construction of intelligent factories,the material transportation in the factory has been gradually transformed from the original manual handling to the handling of intelligent vehicles,realizing integration of intelligent production.However,there are varieties of problems in the operation of intelligent carrier vehicles,such as off-track and other problems.At present,manual inspection and observation through monitoring video are the methods that are used a lot to judge the abnormality of the carrier vehicles.The method of manual inspection is not only easy to appear the situation of missing inspection and emergency response is not timely,but also has safety risks in the environment of dangerous chemicals.The method of observation through surveillance video requires 24-hour monitoring,which goes against the original intention of reducing manpower.In recent years,the development of intelligent surveillance video is rapid.The intelligent video surveillance system using computer vision has been widely used in daily life.Through intelligent monitoring video,the trajectory monitoring can be achieved for 24 hours without manual work.This requires that the intelligent surveillance video can detect and track the vehicle,and distinguish the abnormal state of the vehicle.To solve the above problems,this paper designs an industrial vehicle trajectory tracking algorithm and alarm system based on video stream.Surveillance video is used for image acquisition,and deep learning method is used for vehicle detection and tracking.The abnormal identification of the vehicle will be carried out by the track detected.An online monitoring system is built to display real-time vehicle monitoring information and send corresponding alarm for abnormal conditions of vehicles.This paper mainly completes the following work:(1)The latest YOLO＿v5s network is used to detect the target of the vehicle,and KCF algorithm is used to track the vehicle.By selecting the appropriate network structure and configuring the corresponding network depth and width,the detection speed and precision have achieved a excellent effect.It meets the real-time requirement of the system.(2)A new track discrimination method based on historical similarity is proposed,which is realized by judging the IOU of the current frame and the historical experience frame.The method has a fast processing speed and can detect the abnormal conditions such as off-track,overspeed and deceleration in time.At the same time,the method has a certain fault-tolerant range to prevent misjudgment.It can adapt to a variety of different application scenarios.(3)Design and implement an intelligent monitoring system which integrates monitoring and warning.The system can monitor the running condition of the intelligent vehicle in real time,and send an alarm for abnormal situations,so as to be able to timely repair the fault.The construction of the platform facilitates the management of information,provides important data for decision making,and achieves good results.