Image Search Apparatus Evaluation Value Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image search systems face malfunctions when detecting a large number of human images in a single image, leading to depth of field issues and processing inefficiencies.
Innovation Solution
An image search apparatus that includes a detection device for finding target images based on evaluation values, a storage device for storing these images, and an erasing mechanism that prioritizes removing images with lower evaluation values to maintain a prescribed number, thereby preventing system overload.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all detected face image portions are stored and processed, then detection completeness is improved, but system reliability deteriorates due to malfunctions when handling very large numbers of individuals
Solution Approach 1:
The patent changes the parameter of detection-target likeliness evaluation value to prioritize and rank detected face images. By erasing images with lower evaluation values when the number exceeds the preset threshold, the system maintains reliability while preserving the most relevant detection results, thus resolving the contradiction between system reliability and detection completeness.
2Productivity
If the number of detection-target image portions is increased to capture all individuals, then detection coverage is improved, but device complexity increases leading to processing inefficiency
Solution Approach 1:
The patent introduces an evaluation value parameter to assess detection-target likeliness and uses this parameter to filter and rank detected images. By maintaining only images above a certain evaluation threshold and erasing others when the quantity exceeds the preset number, the system improves processing efficiency while retaining the most significant detection results, resolving the contradiction between productivity and quantity of detected images.
3Reliability
If all detected images are stored in the storage device, then information retention is improved, but storage requirements and processing load increase causing system overload
Solution Approach 1:
The patent employs an evaluation value parameter to differentiate between high-quality and low-quality detections. By systematically erasing images with lower evaluation values when the total number exceeds the preset limit, the system maintains stability and reduces processing complexity while preserving the most valuable detection data, thus resolving the contradiction between system stability and processing complexity.
Data Source
AI summary
If a target-image portion to be detected is detected anew, an evaluation value of this target-image portion is calculated. If the target-image portion is detected anew, it is determined whether the number of target-image portions has exceeded n owing to such detection. If, in a case where n has been exceeded, information concerning a target-image portion having an evaluation value smaller than that of the newly detected target-image portion has been stored, then the information having this evaluation value is deleted from a list and the information concerning the newly detected target-image portion is stored in the list. Since the number of items of information stored in the list will not exceed n, it is possible to avoid a malfunction that can occur if processing is executed following storage of a large number of items of information.