Person Detection Area Adjustment for Obstructed Camera Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The accuracy of person detection in image data obtained from a camera mounted on a moving body, such as a vehicle, is affected by the area in which the detection is performed and the position of the person within that area, leading to reduced effectiveness in identifying individuals, particularly when obstructed by other objects.
Innovation Solution
An image processing device that adjusts the center-of-gravity area-width of detected obstacles in image data to improve person detection accuracy by comparing post-adjustment areas with dictionary data, using HOG features to classify whether the obstacle is a person.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fixed area is used for person detection in the image data, then the detection process is simple, but the detection accuracy decreases when the person is partially obstructed or positioned at different locations
Solution Approach 1:
The patent applies dynamics by making the detection area adaptive rather than fixed. The area width is dynamically adjusted based on the center of gravity position of detected obstacles, allowing the detection region to expand or contract according to the actual spatial distribution of objects in the image. This resolves the contradiction by enabling accurate detection of partially obstructed persons without requiring a complex predetermined multi-zone configuration.
Solution Approach 2:
The patent changes the parameter of area width in the detection process. By modifying the width parameter based on center of gravity calculations, the system adapts the detection area to match the actual spatial extent of obstacles. This parameter adjustment allows the same detection algorithm to maintain high accuracy across different scenarios without increasing overall system complexity.
2Adaptability or versatility
If the detection area width is increased to capture more potential persons, then detection coverage improves, but false detection rate increases due to inclusion of non-person objects
Solution Approach 1:
The patent applies local quality by creating a non-uniform detection area that is wider in regions where persons are likely to be located (based on center of gravity analysis) and narrower or excluded in regions where false detections are more probable. This localized adjustment of detection sensitivity maintains high coverage in critical areas while reducing false positives in less relevant areas.
Solution Approach 2:
The patent performs preliminary action by calculating the center of gravity of detected obstacles before finalizing the detection area. This preliminary analysis allows the system to anticipate the spatial distribution of persons and pre-adjust the detection area boundaries, thereby improving coverage while maintaining reliability by excluding regions unlikely to contain valid targets.
3Measurement precision
If person detection is performed without area adjustment, then processing speed is fast, but detection accuracy decreases for obstructed or off-center persons
Solution Approach 1:
The patent extracts the essential spatial information (center of gravity position) from the detected obstacles and uses only this key parameter to adjust the detection area. By taking out only the necessary information rather than performing complex full-image analysis, the system achieves improved accuracy for obstructed persons while maintaining relatively fast processing speeds.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
An image processing device (41) of a person detection system (30) mounted on a moving body (10) is configured to: detect, in image data obtained from a camera (31), an area (B1, B2) of the image data in which an obstacle appears; perform a center-of-gravity area-width adjustment process in which a position of center of gravity (G1) of the obstacle in the area (B2) of the image data is estimated and a width (W1) of the area (B2) of the image data is adjusted based on the position of center of gravity (G1); and determine whether the obstacle in the area (B2) is a person by comparing a post-adjustment area (B23) obtained after the center-of-gravity area-width adjustment process is performed with dictionary data (D2).