Side Window Detection via Spatial Probability Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for automated HOV/HOT lane enforcement face challenges in accurately detecting side windows due to variations in vehicle types, leading to lower performance compared to windshield detection, especially in cluttered backgrounds and with high intra-class variation in side window shapes and sizes.
Innovation Solution
A two-stage approach is implemented for side window detection, involving the calculation of a spatial probability map (heat map) to weight detection scores, using methods such as deformable parts models and region-convolutional neural networks, and performing image classification to determine passenger presence in the rear seat.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a standard object detector is used for side window detection, then the system can detect windows across the entire image, but the detection accuracy is low due to large variations in side window shapes and sizes for different vehicle types
Solution Approach 1:
The patent applies preliminary action by first detecting the driver side window and windshield before detecting the passenger side window. The detected driver side window location is used to define a search region, and the windshield detection results are used to create a spatial probability map. This preliminary detection and region definition improves passenger side window detection accuracy by constraining the search space, while avoiding the need for complex full-image detection of all window variations.
2Reliability
If the detection search covers the entire image, then all possible window locations are examined, but computational resources are wasted examining regions where windows are unlikely to be located
Solution Approach 1:
The patent applies local quality by creating a spatial probability map that assigns different likelihood scores to different regions of the image based on windshield detection results and geometric relationships. The passenger side window detector is then applied selectively to regions with high probability scores rather than uniformly across the entire image. This concentrates computational resources on likely window locations while maintaining detection reliability.
3Measurement precision
If the system uses simple detection methods, then the processing is faster and simpler, but false positives and missed detections increase due to cluttered backgrounds and vehicle variations
Solution Approach 1:
The patent performs preliminary actions of detecting the driver side window and windshield first, then uses these results to define a constrained search region for the passenger side window. This multi-stage preliminary detection approach improves precision by using geometric relationships and probability maps to guide the final detection, while maintaining productivity by avoiding exhaustive search of the entire image.
Solution Approach 2:
The patent introduces an intermediary spatial probability map that mediates between the simple detector and the complex variations in window appearances. The probability map acts as an intermediary layer that guides the detector to focus on likely window locations based on windshield position and geometric relationships, improving precision without requiring the detector to handle all possible window variations directly.
Data Source
AI summary
Methods, systems, and processor-readable media for detecting the side window of a vehicle. A spatial probability map can be calculated, which includes data indicative of likely side window locations of a vehicle in an image. A side window detector can be run with respect to the image of the vehicle to determine detection scores. The detection scores can be weighted based on the spatial probability map. A detected region of interest can be extracted from the image as extracted image patch. An image classification can then be performed with respect to the extracted patch to provide a classification that indicates whether or not a passenger is in the vehicle or no-passenger is in the vehicle.