VISION-BASED ON-STREET PARKED VEHICLE DETECTION VIA NORMALIZED-VIEW CLASSIFIERS AND TEMPORAL FILTERING

摘要

A system for estimating parking occupancy includes a vehicle-detection device including an adjustable mast supporting an image capture device at a select height. The image capture device acquires video of a current parking area. A computer processor in communication with the image capture device is configured to receive the video data and define a region of interest in the video data. The processor is further configured to perform a spatial transform on the ROI to transform the ROI to a normalized geometric space. The processor is further configured to apply features of a detected object in the normalized geometric space to a vehicle classifier—previously trained with samples acquired from a normalized camera perspective similar to the normalized geometric space—and determine the occupancy of the current parking area using an output of the classifier.

主权项

1. A method for estimating parking occupancy in a current parking area, the method including:
deploying an image capture device in the current parking area; defining a current region of interest (ROI) within a camera field of view of the current parking area; acquiring a sequence of frames captured by the image capture device; performing a spatial transform on the current ROI to transform the current ROI in the camera field of view to a normalized geometric space; detecting at least one object in the normalized geometric space; selecting a vehicle classifier previously trained with samples acquired from a normalized camera perspective similar to the normalized geometric space; and, determining an occupancy of the current parking area by applying extracted features of the detected object to the classifier.