A convolutional neural network approach to parking monitoring in urban radar sensing

We propose a method for occupancy detection of parking space using background estimation and convolutional neural networks. To reduce the impact of ongoing traffic on the MIMO radar image, the moving objects in the scene are labeled as foreground and separated from static background. A convolutional neural network is trained to classify the images based on the occupancy of the parking spaces in the static images. Experimental measurements of real scenarios are presented showing very good performance after removing the foreground objects.