Video-Based Vehicle Detection Using Irrelevant Window Elimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting vehicle occupancy in parking spaces, such as puck-style sensors and ultrasonic sensors, are costly and inefficient, especially in multi-space parking configurations, where they struggle to accurately monitor and maintain availability due to the lack of defined boundaries.
Innovation Solution
A video-based method using a fixed camera that performs a sliding window-based search within a Region of Interest (ROI) to detect vehicles, incorporating a classifier to score candidate windows and applying Non-Maximal Suppression (NMS) to eliminate overlapping and irrelevant detections, thereby improving detection accuracy and reducing partial vehicle detection errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor-based methods (puck-style sensors, ultrasonic sensors) are used for vehicle detection, then detection capability is provided, but installation cost and maintenance cost increase significantly
Solution Approach 1:
The patent replaces mechanical sensor-based detection systems (puck-style sensors, ultrasonic sensors) with a video-based detection system using cameras and image processing algorithms. This substitution eliminates the need for expensive physical sensor installation while maintaining vehicle detection capability through computer vision techniques.
Solution Approach 2:
The patent uses visual information captured by cameras as a copy or representation of the physical parking environment to infer vehicle presence. Instead of direct physical sensing, the system creates digital representations (video frames) and processes them to detect vehicles, thereby avoiding costly hardware installation.
2Reliability
If multiple in-ground/ultrasonic sensors are installed to monitor parking spaces, then detection coverage is improved, but maintenance disruption to street traffic increases
Solution Approach 1:
The patent replaces multiple in-ground sensors with a single video camera system that uses image processing to detect vehicles. This eliminates the need for physical sensor maintenance in the street, thereby reducing disruption to traffic flow while maintaining monitoring accuracy.
Solution Approach 2:
The patent introduces video images as an intermediary medium between the camera and the detection algorithm. The video frames serve as a intermediary representation that allows the system to monitor parking spaces without direct physical contact with the street environment, thereby avoiding maintenance disruptions.
3Adaptability or versatility
If video-based solution is used for multi-space parking configurations, then cost is reduced and versatility is improved, but detection precision in multi-space configurations becomes more challenging
Solution Approach 1:
The patent segments the video frame into multiple regions of interest (ROIs) corresponding to different parking spaces. By dividing the overall scene into discrete monitoring zones, the system can apply detection algorithms to each segment independently, thereby maintaining precision even in complex multi-space configurations.
Solution Approach 2:
The patent applies different processing characteristics to different regions of the video frame based on their specific parking space characteristics. Each ROI can be processed with optimized parameters suited to its location and size, thereby maintaining high detection precision across diverse multi-space parking configurations.
4Reliability
If sliding window-based search is performed to detect vehicles, then detection coverage is improved, but computational complexity increases
Solution Approach 1:
The patent divides the video frame into multiple overlapping sliding windows that systematically scan the entire image. This segmentation approach ensures comprehensive coverage while allowing parallel processing of different window regions, thereby managing computational complexity through structured division of work.
Solution Approach 2:
The patent performs detection on multiple sliding windows that may overlap or extend beyond the actual vehicle boundaries. This excessive action ensures that no potential vehicle detections are missed, and the system can then filter results to eliminate false positives, thereby achieving reliable coverage with manageable computation.
Data Source
AI summary
This disclosure provides vehicle detection methods and systems including irrelevant search window elimination and/or window score degradation. According to one exemplary embodiment, provided is a method of detecting one or more parked vehicles in a video frame, wherein candidate search windows are limited to one or more predefined window shapes. According to another exemplary embodiment, the method includes degrading a classification score of a candidate search window based on aspect ratio, window overlap area and/or a global maximal classification.