Research On Outdoor Parking Status Recognition Algorithm Based On Image Processing

Intelligent Transportation System(ITS)integrates computer technology,electronic sensing,communication and other technologies to realize intelligent traffic management.It plays an important role in solving parking difficulties,alleviating traffic congestion and maintaining traffic safety.The researchers proposed sensor-based parking status recognition algorithm.At first,they used infrared,ultrasonic,radar and other sensors to measure the distance to the target vehicle to determine the status of parking.However,above these methods not only sensitive to external environmental factors,but also has the disadvantages of high cost and difficult maintenance.Then,researchers utilize image processing and machine vision technology,proposed parking status recognition algorithm based on image processing.How recognition parking space status? How to determine the specific location of parking spaces in the parking lot image? In order to address the above problems,we focus on parking marking and parking status recognition.On the one hand,we design a parking marking method for the irregular parking lot.We introduce the AdaBoost classifier algorithm to identify the location of the vehicle in the parking lot,complete the rough marking of the parking spaces.In order to reduce the statistical error,an improved candidate box restraining method is proposed to statistic the parking regular of vehicles for a certain period.Moreover,we increase the fault tolerance mechanism to realize the accurate positioning of the parking spaces.Parking marking provide parking spaces of interest of region for parking status recognition.On the other hand,we propose parking status recognition based on a multi-scale feature and serial feature fusion.Through introduce LBP and LPQ operators to extract features of the parking space image,these features can better describe the rotation invariant feature and the texture feature of the low resolution image.Color is one of the most intuitive property,we utilize LUV color space to get a feature vector,which may cause dimension disaster.Therefore,this paper improves the feature calculation method: computing the sum of multi-size rectangular cross-sectional area pixel.By analyzing the advantages and disadvantages of the current state-of-the-art parking recognition algorithm,the serial feature fusion method is used to combine the texture features and the color features,obtain a new feature vector,then use SVM classifier to achieve the parking status classification.During the experiment,we test about 750,000 parking spaces images,including five parking areas under different view,and verify the feasibility and robustness of proposed algorithm.Experimental results show that the proposed algorithm is more effective and has better generalization performance.