Browse Prior Art Database

Efficient Use of Visually Trained Virtual Directional Loops for Parking Stall Management

IP.com Disclosure Number: IPCOM000237923D
Publication Date: 2014-Jul-21
Document File: 8 page(s) / 258K

Publishing Venue

The IP.com Prior Art Database

Abstract

Parking management systems are being proposed that provide real-time parking occupancy data to drivers to reduce fuel consumption and traffic congestion. In the context of parking stall occupancy determination; there are various levels of performance metrics that need to be met depending on the applications: total # of occupancy of the entire lot, total # of occupancy of sub-regions within a lot (e.g. each aisle), occupancy state of each individual stall, etc. Among them, total # of occupancy of the entire lot is the most common and useful information for parking lot applications. This metric is what this idea aims to solve. Prior video-based methods focused on video-processing (motion detection and optional tracking) which are prone to noise due to the low-contrast scenario, moving shadows of side traffic, or a group of walking people, etc. A vision-based method alone is computationally expensive, and its classification task may be challenging due to different poses and entering/exiting positions of vehicles passing the entrance of a lot. This idea proposes a system and method to count vehicles using virtual video-based loops. Robust counting is achieved by using a combination of tracking (with direction) and vehicle detection classifiers. The system is applied to determining parking occupancy in lots or garages.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 14% of the total text.

Efficient Use of Visually Trained Virtual Directional Loops for Parking Stall Management

Parking management systems are being proposed that provide real-time parking occupancy data to drivers to reduce fuel consumption and traffic congestion. In the context of parking stall occupancy determination; there are various levels of performance metrics that need to be met depending on the applications: total # of occupancy of the entire lot, total # of occupancy of sub-regions within a lot (e.g. each aisle), occupancy state of each individual stall, etc. Among them, total # of occupancy of the entire lot is the most common and useful information for parking lot applications. This metric is what this idea aims to solve. Prior video-based methods focused on video-processing (motion detection and optional tracking) which are prone to noise due to the low-contrast scenario, moving shadows of side traffic, or a group of walking people, etc. A vision-based method alone is computationally expensive, and its classification task may be challenging due to different poses and entering/exiting positions of vehicles passing the entrance of a lot.  This idea proposes a system and method to count vehicles using virtual video-based loops. Robust counting is achieved by using a combination of tracking (with direction) and vehicle detection classifiers. The system is applied to determining parking occupancy in lots or garages.

Background

Parking management systems are being proposed that provide real-time parking occupancy data to drivers to reduce fuel consumption and traffic congestion. In the context of parking stall occupancy determination; there are various levels of performance metrics that need to be met depending on the applications. For example, one level of performance metric is measured by the accuracy of the total number of spaces available in a parking lot over time. This total number of spaces available can be considered as the lowest (most achievable) level of information for this application but is also the most common and useful information. Another level of information can be the total number of spaces available for each floor in a parking building (indoors) or the total number of spaces available for each aisle/aisle-pair (outdoors). This can be useful for providing efficient navigation to a parker entering a large parking lot. The highest level of information can be the state of each parking stall (where are all those available spaces) in the parking lot. If accurate information can be achieved in this level, it opens up several additional applications such as mining parking patterns for better management and configuration, managing/monitoring unexpected parking capacity reduction due to poorly parked vehicles or poor weather conditions (snow piled up) etc. Furthermore, high-level information can be easily aggregated to yield the lower level information by a simple summation. Given these reasons, one would argue why not only develop methods that...