Vehicle Detection Using Windshield Edge Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle detection systems in Intelligent Traffic Systems face challenges in accurately detecting vehicles in images from low-mounted cameras, especially in heavily occluded scenes or under varying weather conditions, due to the occlusion of vehicle features like windshields.
Innovation Solution
A method that projects a geometrical windshield model onto an image, detects horizontal edges, and determines the probability of each point belonging to a windshield based on edge correlation and color similarity, allowing for direct detection of vehicles even in occluded scenes, with a confidence map generated using Bayesian formulation for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional vehicle detection methods are used in heavily occluded scenes, then detection accuracy deteriorates, but the patent achieves high detection accuracy by focusing on windshield detection
Solution Approach 1:
The patent extracts and focuses on a specific critical feature (windshield) from the complete vehicle structure, rather than attempting to detect all vehicle features. By isolating the windshield as the primary detection target, the system achieves reliable vehicle detection even when other vehicle parts are occluded by traffic density or camera angle
Solution Approach 2:
The patent segments the vehicle detection task into identifying specific geometric features (windshield edges) rather than detecting the entire vehicle at once. This segmentation allows the system to focus computational resources on detecting the most reliable indicator (windshield) that remains visible even when other vehicle features are hidden
2Area of stationary object
If low-mounted cameras are used to capture traffic images, then the field of view is improved, but windshield detection becomes more difficult due to occlusion
Solution Approach 1:
The patent changes the detection parameters by focusing on specific geometric properties of the windshield (horizontal edges, characteristic shapes) rather than relying on overall vehicle appearance. This parameter change enables effective detection despite the challenging viewing angles produced by low-mounted cameras
3Reliability
If comprehensive vehicle feature detection is attempted, then detection robustness deteriorates in occluded scenes, but the patent achieves robust detection by using windshield-specific features
Solution Approach 1:
The patent extracts only the essential windshield features needed for reliable detection, eliminating the complexity of detecting all vehicle features. By taking out just the critical windshield component, the system achieves high reliability with reduced complexity
Solution Approach 2:
Instead of trying to detect the entire vehicle and then identify the windshield, the patent inverts the approach by directly detecting the windshield as the primary target. This inversion simplifies the detection process and improves reliability in occluded scenes
Data Source
AI summary
The invention concerns a traffic surveillance system that is used to detect and track vehicles in video taken of a road from a low mounted camera. The inventors have discovered that even in heavily occluded scenes, due to traffic density or the angle of low mounted cameras capturing the images, at least one horizontal edge of the windshield is least likely to be occluded for each individual vehicle in the image. Thus, it is an advantage of the invention that the direct detection of a windshield on its own can be used to detect a vehicle in a single image. Multiple models are projected (206) onto an image with reference to different points in the image. The probability of each point forming part of a windshield is determined based on a correlation of the horizontal edges in the image with the horizontal edges of the windshield model referenced at that point (220). This probability of neighboring points is used to possible detect a vehicle in the image (224). Aspects of the invention include a method, software and traffic surveillance system.