The Research And Implementation Of Parking Detection Algorithm Based On Magnetic Sensors

The rapid increase in the number of motor vehicles in urban areas poses a challenge for the more timely and accurate management of public transportation and large parking lots.Parking detection technology plays a crucial role in this process.Due to the advantages of low cost,easy installation,and high sensitivity,many researchers have conducted research on parking detection algorithms based on magnetic sensors.However,the magnetic signal is subject to various magnetic interference substances,which affects the detection accuracy.The classical adaptive threshold detection algorithm solves the problem of baseline drift.However,its detection state machine does not consider special factors such as weakly magnetic vehicles,and cannot cover all scenarios in practical applications,resulting in missed or false detections.Moreover,each parking space corresponds to a sensor,which increases the deployment and maintenance costs in large parking lots.Each sensor is also more prone to interference from adjacent vehicles.Since most parking lots are equipped with intelligent parking management systems,the cost is one of the most important considerations in engineering design.However,few studies in the existing literature have addressed how to reduce the system cost.Furthermore,most of the studies extract features manually,and the results are given by the potential difference of magnetic signal which is susceptible to environmental interference.Once the application scenario is changed,the original parameters will no longer apply.This thesis proposes a multi-transitory parking detection algorithm based on variance sequences and adopts the translational-invariant wavelet denoising method to pre-process the magnetic signal,which can effectively reduce the noise impact and meet the detection requirements.A multi-transitory state machine based on variance signal is designed for parking detection,considering the challenges of weak magnetic vehicles in practical applications.To reduce the application cost of parking detection algorithms in large parking lots,a W-shape magnetic wireless sensor array is proposed,which significantly reduces the number of sensors and thus reduces the system costs accordingly.Based on this deployment,a parking detection algorithm using a deep learning model is designed and implemented,which does not rely on manual feature extraction and then judge by the change value of the magnetic signal.Instead,it fuses data from multiple sensors for automatic decision-making,which effectively improves the applicability of the algorithm in different scenarios.In this thesis,the performance of the proposed algorithms is evaluated in a standard parking lot.The magnetic sensor nodes are designed and implemented,and a parking detection system is built.Compared with the adaptive threshold detection algorithm,the proposed multi-transitory parking detection algorithm based on variance sequences has higher accuracy,with an overall detection accuracy of 98.82%,and it can also correctly detect weakly magnetic vehicles.The experimental results demonstrate that the algorithm can achieve high accuracy in various situations while being simple and easy to implement.Moreover,in the same scenario,the number of sensors required by the proposed detection algorithm based on a W-shape magnetic is only half compared to the traditional detection algorithm that uses one-to-one deployment.Test in the actual scene shows that the algorithm can significantly reduce costs and improve adaptability while ensuring high detection accuracy.Thus,it is easier to implement in large parking lots.