Surveillance Image Sampling Using HSV Background Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Surveillance systems face challenges in accurately recognizing objects due to background interference from high color saturation and brightness variation, leading to increased computational load and reduced recognition accuracy.
Innovation Solution
An object sampling method that transforms surveillance images to HSV color space, computes saturation and brightness variation, sets a bonding box, and computes a background score based on these parameters to select images with low interference backgrounds for analysis, thereby reducing computational load and enhancing recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the surveillance system retrieves surveillance images with the suspicious person appearing on the surveillance region of the photographed camera captured at other points of time, then the recognition accuracy can be improved, but the systematic computation increases significantly
Solution Approach 1:
The patent applies preliminary action by computing background scores for all surveillance images in advance, before the actual object recognition process. The operation processor pre-calculates background scores based on color saturation and brightness variation, stores these scores, and then uses them to quickly filter candidate images during retrieval operations. This pre-computation approach reduces the systematic computation required during actual recognition while maintaining high recognition accuracy.
2Adaptability or versatility
If the object recognition technology is applied to images with colorful background or obvious edges, then the coverage of detection can be improved, but the recognition accuracy decreases due to background interference
Solution Approach 1:
The patent applies local quality by differentiating between background regions and foreground objects in the surveillance images. It computes background scores specifically for background pixels based on color saturation and brightness variation, then uses these local background characteristics to filter images. This allows the system to maintain detection coverage across various scene types while improving recognition accuracy by identifying and excluding images where the background would interfere with object recognition.
3Measurement precision
If the surveillance system analyzes all captured images, then the recognition accuracy can be improved, but the recognition speed decreases due to increased computational load
Solution Approach 1:
The patent pre-computes and stores background scores for all surveillance images before the recognition process. When recognition is needed, the system quickly filters images using these pre-computed scores, selecting only those with low background interference for detailed analysis. This preliminary filtering action maintains recognition accuracy by ensuring only suitable images are analyzed, while significantly improving recognition speed by avoiding unnecessary computation on unsuitable images.
Data Source
AI summary
An object sampling method is applied to an image analysis apparatus with an image receiver and an operation processor. The image receiver acquires an image stream. The object sampling method includes transforming color space of at least one surveillance image in the image stream, computing saturation and gradient of specific pixels within the at least one surveillance image, setting a bonding box within the at least one surveillance image, and computing a background score relevant to the saturation and value of all pixels in the bonding box, and comparing the background score with a first sampling condition so as to determine whether the at least one surveillance image is applicable for the image analysis apparatus via a comparison result of comparing the foreground score with the first sampling condition.