Vehicle Detection Device Adaptive Luminance Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle detection methods using on-vehicle cameras face challenges in accurately detecting vehicles due to low luminance differences between vehicles and backgrounds, leading to false extractions of shadows and other non-vehicle regions, especially when vehicles have similar colors or are overlapped, or when shadows of trees or buildings are mistaken for vehicle shadows.
Innovation Solution
A vehicle detection device that creates a brightness distribution of the captured image, estimates an upper luminance threshold for the vehicle bottom side, and uses pattern matching to determine if the assumed region corresponds to a vehicle, focusing on the lower illuminance of the road surface and tire area to differentiate it from shadows and other regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern matching is applied to the entire image region, then vehicle detection accuracy is improved, but calculation time increases and real-time detection becomes difficult
Solution Approach 1:
The image is divided into multiple regions of interest (ROIs) based on detected edges or features. Pattern matching is then applied only to these segmented regions rather than the entire image, reducing calculation time while maintaining detection accuracy for vehicles within these regions.
Solution Approach 2:
Instead of applying pattern matching to the entire image, the method applies it partially to only those regions where vehicles are likely to be present (assumed regions). This partial action reduces computational load while still achieving effective vehicle detection.
2Productivity
If algorithms extract black regions or high luminance variation regions as assumed vehicle regions, then detection speed is improved, but false detections increase due to shadows and other non-vehicle regions
Solution Approach 1:
The patent applies different extraction criteria to different regions of the image. For example, it uses luminance variation criteria for regions where vehicles are expected but applies shadow detection criteria only in regions where shadows are likely to occur. This localized approach reduces false detections while maintaining detection speed.
Solution Approach 2:
The patent dynamically adjusts extraction parameters such as luminance thresholds and variance values based on local image characteristics. By changing these parameters adaptively across different regions, the system can distinguish between actual vehicles and shadows or other high-contrast regions, improving reliability without sacrificing detection speed.
3Productivity
If vertical edge extraction is used to detect vehicle regions, then detection speed is improved, but detection accuracy decreases when luminance difference between vehicle and background is small
Solution Approach 1:
The patent dynamically switches between different extraction methods (vertical edge extraction, black region extraction, luminance variation extraction) based on local image conditions. When luminance difference is sufficient, it uses vertical edge extraction for speed; when luminance difference is small, it switches to more accurate but computationally intensive methods, thus maintaining both speed and accuracy adaptively.
Data Source
AI summary
Attention is paid on the phenomenon that substantially no direct sun light or no reflected light from surrounding objects is applied, to a vehicle bottom side and tire travel surfaces and the vehicle bottom side and the tire travel surfaces have lower luminance values than all the other portions including the shades of trees or buildings along the road in an on-vehicle camera image. Luminance threshold specifying unit (13) analyzes an image luminance histogram and specifies the upper limit of the pixel luminance which can be assumed as a region of the vehicle bottom side. Vehicle assumed region extracting unit (15) assumes that the pixel position having a luminance value not greater than the upper limit is the region of the vehicle bottom side and sets a vehicle assumed region, thereby verifying presence/absence of a vehicle in the vicinity.