Vehicle Pedestrian Detection Using Luminance Mask Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle surroundings monitoring systems inaccurately determine the type of objects, particularly pedestrians, when multiple pedestrians are close to each other, due to incorrect extraction of image features from infrared camera images.
Innovation Solution
A vehicle surroundings monitoring apparatus that extracts a first image portion likely to be a pedestrian's head and searches for additional image portions such as legs within a lower search area, using reference and comparative mask areas, to accurately determine the object type based on luminance and distance criteria, thereby preventing incorrect recognition of pedestrians as other objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only a first image portion (head) is extracted for pedestrian detection, then the detection process is simple and fast, but the accuracy deteriorates when multiple pedestrians are close together
Solution Approach 1:
The patent divides the pedestrian detection task into multiple segments: extracting a first image portion (head region) and a second image portion (body/leg region), then independently analyzing each segment. This segmentation allows the system to use multiple extraction conditions (luminance, area, shape) for each portion without overwhelming the processing system, while collectively achieving high detection accuracy even when pedestrians are close together.
2Reliability
If multiple image portions are extracted and analyzed, then the object type determination accuracy is improved, but the processing time and complexity increase
Solution Approach 1:
The patent performs preliminary extraction of the first image portion (head) and determines its attributes (luminance, area, shape) before proceeding to extract the second image portion (body). This preliminary action allows the system to quickly filter out non-pedestrian objects early in the process, reducing the need for extensive processing of all objects and thereby minimizing overall processing time while maintaining high reliability.
3Productivity
If determination is based only on first image portion features, then the processing is fast, but false positives increase when pedestrians are close together
Solution Approach 1:
The patent introduces the second image portion (body/leg region) as an intermediary element that mediates between the first image portion (head) and the final determination. By requiring both portions to satisfy their respective extraction conditions, the system acts as a filter that eliminates false positives (such as animals or objects that may resemble heads) while maintaining detection speed through efficient conditional checking.
Data Source
AI summary
A vehicle surroundings monitoring apparatus capable of recognizing the type of object existing around a vehicle with accuracy. The vehicle surroundings monitoring apparatus includes a process unit, which extracts a first image portion HP—1 likely to be the head of an object and sets a reference mask area MASK_C including HP—1, a left-hand mask area MASK_L near the left side of the reference mask area MASK_C, and a right-hand mask area MASK_R near the right side of the reference mask area MASK_C, and a process unit which searches for a second image portion presumed to be a leg image within a lower search area AREA—3 when a difference between an average luminance AVE_C of the reference mask area MASK_C and an average luminance AVE_L of the left-hand mask area MASK_L or a difference between the average luminance AVE_C and an average luminance AVE_R of the right-hand mask area MASK_R is equal to or greater than a predetermined level and which recognizes that the object type is “pedestrian” in the case where a plurality of second image portions are detected.