Vision-based parking lot occupancy evaluation system using 2D separable discrete wavelet transform

A simple system for rough estimation of the occupancy of an ad-hoc organized parking lot is presented. A reasonably simple microprocessor hardware with a low resolution monochrome video camera observing the parking lot from the location high above the parking surface is capable of running the proposed 2-D separable discrete wavelet transform (DWT)-based algorithm, reporting the percentage of the observed parking area occupied by cars. A simple calibration is needed – the mask covering all the areas outside the parking lot must be prepared. The proposed system has been tested on the dedicated FPGA-based hardware in real conditions and proven immune to scene and light changes. As it is discussed in the paper, it can be used in low-power wireless sensor networks.