Person Identification via Segmented Candidate Area Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately associating individuals across images, especially when partial body parts are not captured due to camera installation conditions, leading to restricted feature extraction and difficulty in precise comparison.
Innovation Solution
An image processing apparatus that divides candidate areas into smaller sections, performs comparisons between corresponding parts, and uses certainty factors to predict missing feature information, allowing for precise identification even if partial areas are not suitable for similarity degree calculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If feature extraction is performed using images captured by cameras with specific installation conditions, then the comparison can be performed using available features, but the identification precision deteriorates when partial body parts are not captured
Solution Approach 1:
The patent segments the person's body into multiple candidate areas (head, torso, limbs) and processes each area independently. This segmentation allows the system to identify which body parts are actually captured in the image and focus comparison on those available parts, rather than requiring complete body information. The segmentation enables partial feature matching to contribute to overall identification accuracy.
Solution Approach 2:
The patent performs preliminary determination of certainty factors for each candidate area before conducting the full comparison process. By pre-assessing the reliability of each body part region (considering factors like camera angle, occlusion, and image quality), the system can weight the importance of different features appropriately. This preliminary action allows the comparison to focus on high-certainty regions while compensating for missing or low-quality features in other areas.
2Measurement precision
If the entire body image is used for association, then more features are available for comparison, but the processing complexity increases when dividing and comparing multiple parts
Solution Approach 1:
The patent divides the person's body into multiple candidate areas and processes each area independently through the comparison system. This segmentation transforms a complex whole-body comparison into multiple simpler regional comparisons, making the processing more manageable while maintaining high precision through focused analysis of each body part's specific features.
Solution Approach 2:
The patent applies different certainty factor assessments and comparison strategies to different candidate areas based on their local characteristics. Each body part region is evaluated according to its specific quality metrics (visibility, image clarity, relevance to identification), allowing the system to optimize processing for each local region rather than applying a uniform complex process to the entire body.
3Productivity
If candidate areas with low certainty factors are excluded from comparison, then the processing speed increases, but the identification accuracy may deteriorate due to loss of useful information
Solution Approach 1:
The patent dynamically adjusts the certainty factor threshold parameter based on the specific comparison context and available candidate areas. Rather than using a fixed threshold that would exclude potentially useful low-certainty regions, the system adapts the threshold to balance processing efficiency with information utilization, allowing more candidate areas to participate in the comparison when appropriate while maintaining speed.
Solution Approach 2:
The patent performs comparison processing on more candidate areas than a strict threshold would require, including some areas with lower certainty factors. This excessive action ensures that potentially useful information is not discarded, and the system can compensate for missing or low-quality features in high-certainty areas by utilizing additional partial information from lower-certainty regions, thereby maintaining both speed and accuracy.
Data Source
AI summary
An apparatus for identifying a candidate area in a first image corresponding to an object in a second image, includes a memory and a processor to divide the plurality of candidate areas into a plurality of small candidate areas, divide an image area of the object into a plurality of small areas, perform first comparison processing for a first part, when there is a first candidate area lacking image information of the small candidate area corresponding to the first part, perform second comparison processing for a second part, predict missing result on the small candidate area corresponding to the first part in the first candidate area based on a result of the first comparison processing on a candidate area other than the first candidate area, and a result of the second comparison processing on the plurality of candidate areas, and identify the candidate area based on a prediction.