Research On Parking Space Detection Algorithm Based On Machine Vision

In recent years,with the development of modern intelligent transportation system,parking space recognition and location information acquisition has become a hot research direction of intelligent transportation,and also one of the key technologies and difficulties of intelligent parking.In recent years,with the rapid development of society,the number of vehicles is rising rapidly.The available parking spaces can no longer meet the needs of the present people.There are a lot of parking area congestion problems in the city.In this paper,through the detection of parking space,the parking information is pushed for the parking people,which can alleviate the parking difficulties and the parking area congestion to a certain extent.This paper mainly the problem of parking space detection from two aspects.On the one hand,for the traditional parking space detection based on small scene,a new method of parking line detection based on HSV color space and improved Otsu algorithm is proposed.Firstly,RGB image is transformed into HSV space,and the information of non parking line is filtered by using the different properties of H and s components,and the improved Otsu algorithm is used to filter the information of non parking line in turn,and the parking line is divided initially.Then,the two-value morphological processing is used to preprocess the gap and crack of the initially divided parking line,and then the effective recognition and location acquisition of the parking space area are carried out by the method of connecting analysis,backbone extraction,Hough linear detection and contour detection.The improved Otsu algorithm introduces particle swarm optimization algorithm,which greatly shortens the segmentation time compared with the traditional Otsu algorithm.The results show that the algorithm can effectively segment the parking line area,and the recognition effect of human eye is good and the segmentation time is greatly reduced,which has strong practicability.On the other hand,a deep learning method is proposed for the detection of parking space based on large scene.The existing network model is used to train,and the distribution of vehicles in the parking area is visualized and the parking space is judged.The visualization and orderly marking of the empty position are designed and realized.After the data and self-sampling data preprocessing are used,the practicability and reliability of the visualization of vehicle distribution,the discrimination of parking space and the orderly marking of parking position in the training module are verified.Taking the outdoor parking lot with small scene and large scene as the research platform,a parking space detection algorithm based on machine vision is proposed,which makes the parking process more convenient,reliable and safe,and provides a good platform for intelligent management of parking lot,and also provides a strong support for the accurate parking of intelligent parking system.