Research On Empty Parking Space Recognition Algorithm Based On Roadside Equipment

With the vigorous development of our country’s automobile industry and the increase of national income,the number of personal automobiles in our country has increased year by year.According to the data of the Ministry of Public Security of China,by the end of 2021,the number of automobiles in China will be 302 million.The construction of parking spaces in big cities is lagging behind,and the parking spaces in a large number of big cities are relatively insufficient.Therefore,time is often wasted in searching for parking spaces during daily travel,causing a series of traffic problems.Today,with the development of technologies such as V2 X,AI,and cloud computing,the realization of modern transportation is inseparable from the coordination of vehicles,roads,and clouds.Share the information interaction resources of the vehicle,the road and the cloud to build a collaborative intelligent transportation system.In the context of new infrastructure construction,parking space recognition also plays an important role and has strong practical significance.Many traditional outdoor park use a variety of sensors such as geomagnetic coils sensors,electromagnetic waves sensors,ultrasonic waves sensors,and geomagnetic sensors to detect the state of the parking space,but the installation and maintenance costs of the sensors are expensive.Therefore,compared with using sensors,the method of using camera images to detect the parking space status has lower cost and higher detection efficiency.At the same time,the camera data can be applied to other scene services,so it has a more economical and broader development prospect.And traditional machine learning algorithms require manual design of features and classifiers,it is also difficult to achieve end-to-end.Therefore,this paper proposes a convolutional neural network based on residual network structure.Accuracy rates of99.1% and 94.6% can be achieved on the two datasets,which are both better than the m Alex Net.In this paper,we construct a new convolutional neural network structure.The residual network structure is used to optimize the problem of feature loss and preserve richer features.We also use batch normalization algorithm,adaptive average pooling and Kaiming parameter initialization to optimize the convolutional neural network model,so that the convolutional neural network model parameters can be used in more complex scenarios such as snow environments.It still achieves good results,surpassing the existing m Alex Net.In view of the problem of the snow environment that is not included in the existing parking space datasets,the parking space dataset under snow weather is collected.Finally,the algorithm model proposed in this paper is deployed on the roadside edge computing device,which improves the operation efficiency of the convolutional neural network model.V2 X technology is used to realize road-vehicle communication,Io T technology is used to realize road-cloud communication,and the algorithm model is used for parking spaces recognition.Based on these,we completed the application of the algorithm used in the actual scene.