Estimation of Free Space on Car Park Using Computer Vision Algorithms

Abstract

A system for monitoring of vacant parking spots can save drivers a lot of time and costs. Other citizens can benefit from a reduction of pollutions too. In our work, we proposed the computer vision system that estimates free space in a car park. The system uses three separate estimation methods based on various approaches to the estimation issue. The free car park area is recognised on a video frame by as the broadest cohesive area, the largest group of pixels with similar colours, and background for parked cars. The raw results of the estimations are aggregated by a Multi-Layer Perceptron to obtain the final estimate. The test on real data from the City of Warsaw showed that the system reaches 95% accuracy. Moreover, the results were compared with the registers from the parking machines to estimate a gap between covered payment and the accurate number of parked cars.