Research On Detection Of Parking Space Status In Open-air Parking Lot Based On UAV Video Image

Finding a parking space in a relatively large outdoor parking lot is a waste of time and energy.This is mainly because the driver neither has the personnel to guide in time nor know the specific location of the empty parking space.In order to reduce the difficulty for drivers to find empty parking spaces and improve the management of parking lots,the key is to detect the status of parking spaces and find empty parking spaces.The parking space detection method based on video image is based on the video image collected by the camera,and uses image processing,machine vision and other technologies to complete the automatic parking space detection without a large amount of hardware equipment.In the past,most of the parking space detection methods based on video images were obtained by monitoring cameras or setting up cameras by themselves.The angle between the camera and the ground is about 450.In addition,the height of the camera is very limited.Finally,the field of view obtained is very limited,and it is difficult to put the entire parking lot into one image.Moreover,the vehicle may be blocked by other vehicles in the obtained video image.To solve these problems,this paper studies the parking space state detection method based on UAV video image.Obtaining the overall image of the parking lot by using drones to carry cameras overhead view of the parking lot has changed the fact that it is difficult for one camera to completely cover the entire parking lot in previous studies.The research content of this paper mainly includes two parts: the location of parking space in the image and the detection of parking space status.In the positioning part of the parking space,the coordinates of two points on the diagonal of a single parking space are used to indicate its position in the image.For how to obtain the coordinates of the two points on the diagonal,this article chooses the cumulative probability Huo Husband transform algorithm.Since the cumulative Hough transform algorithm requires the input image to be an edge binary image,after the image is preprocessed,the image must be edge detected.In this paper,an improved Canny operator is used to perform edge detection on the image to improve parking space positioning.The degree of accuracy.In the part of parking space status recognition,this paper introduces machine learning and deep learning to detect parking space status.In the machine learning method,this paper analyzes the classification effect of SVM on parking space images under different feature vectors,and finds that different feature vectors have a significant impact on the SVM classification effect;in the deep learning method,this paper designs automatic detection The program first cuts the parking space images according to the parking space coordinates,then sends them to the classic convolutional neural network VGG-16 to classify these parking space images,and finally outputs the number of the vacant parking space.Through experiments and the evaluation and analysis of the experimental results,it is shown that the open-air parking space detection method based on UAV video images designed in this paper can quickly detect the status of all parking spaces in the entire parking lot and output the positions of vacant parking spaces.The detection accuracy of vacant parking spaces is over 95%.