Vehicle Identification Using Multi-View Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computer vision object detectors are inefficient and inaccurate in tracking vehicles, especially when visibility is limited, such as from overhead cameras or when license plates are occluded by other cars, leading to poor results in automated license plate recognition.
Innovation Solution
A system that uses one or more processors to detect vehicles with cameras, capture multiple images, determine vehicle features such as shape and color from various viewpoints, and generate a unique vehicle identification based on these features, enabling efficient and accurate tracking of vehicles even when license plates are not visible.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional computer vision object detectors are used to track vehicles, then the system can perform automated tracking, but the tracking becomes difficult and inaccurate when visibility is limited (overhead cameras or occluded license plates)
Solution Approach 1:
The system transitions from relying solely on 2D license plate recognition to extracting and matching 3D vehicle features (length, width, height, shape) from multiple camera viewpoints. This dimensional transformation enables accurate vehicle identification even when license plates are occluded or cameras are positioned overhead, directly resolving the visibility limitation problem.
2Area of stationary object
If overhead cameras are used for vehicle detection, then the system can capture vehicle images from above, but license plate recognition becomes difficult due to lack of visibility
Solution Approach 1:
The system segments the vehicle identification task into two independent components: (1) capturing vehicle imagery from overhead cameras for broad coverage, and (2) extracting specific vehicle features (dimensions, shape, color) from those images for identification. This segmentation allows the system to maintain wide area coverage while compensating for license plate invisibility through alternative feature extraction.
3Productivity
If conventional techniques are used for vehicle tracking, then the system can perform basic tracking, but the results are poor when objects are occluded by other cars
Solution Approach 1:
The system performs preliminary extraction and storage of vehicle features (length, width, height, shape, color) from images captured by overhead cameras before tracking begins. When vehicles are occluded during tracking, the system can rely on these pre-extracted features for identification, maintaining both tracking efficiency and reliability even when direct visibility is blocked.
Data Source
AI summary
Implementations generally relate to identification of a vehicle for smart patrolling. In some implementations, a method includes detecting a vehicle with a camera. The method further includes capturing a plurality of images of the vehicle. The method further includes determining a plurality of vehicle features from each image of the plurality of images. The method further includes generating a vehicle identification for the vehicle based at least in part on some of the plurality of vehicle features.