Camera Region Detection Using Local Gradient Pattern Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera systems face difficulties in detecting regions with specific shapes, such as human faces or license plates, under varying illumination conditions and when there are exceptional gradation differences between adjacent pixels, leading to reduced detection accuracy.
Innovation Solution
The method involves calculating gradation differences between central and peripheral pixels in local regions, comparing these differences to obtain local gradient pattern (LGP) values, and using these values to detect regions with specific shapes, thereby improving detection robustness across different illumination settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Local Binary Pattern (LBP) is used for detecting regions with specific shapes, then the detection is robust against illumination variations, but detection accuracy deteriorates when gradation differences between adjacent pixels are exceptionally great
Solution Approach 1:
The patent transforms the LBP parameter calculation by introducing a thresholding mechanism. Instead of directly using the raw gradation difference (in - ic), the patent applies a threshold T and uses the function s(|in - ic| - T) to determine binary values. This parameter transformation allows the system to ignore exceptional gradation differences that exceed the threshold while still responding to normal illumination variations, thereby resolving the contradiction between robustness and accuracy.
2Productivity
If conventional LBP method is used, then the processing is simple and fast, but detection accuracy reduces under challenging illumination conditions
Solution Approach 1:
The patent modifies the LBP calculation by introducing a threshold parameter T into the binary pattern generation process. The modified calculation s(|in - ic| - T) maintains the computational efficiency of LBP while improving accuracy under challenging illumination. The threshold acts as a filter that prevents exceptional gradation differences from corrupting the binary pattern, thus enhancing detection accuracy without significantly increasing processing complexity.
3Reliability
If the camera processes all image data to detect regions with specific shapes, then detection coverage is complete, but calculation amount and storage space increase
Solution Approach 1:
The patent divides the image processing into local regions and applies LBP analysis to each local region independently. By segmenting the image into 3×3 pixel blocks and calculating LBP values for each block, the system can identify regions with specific shapes more efficiently. This segmentation approach reduces the overall calculation amount by processing smaller units and enables early termination when a region of interest is detected, thus reducing storage requirements while maintaining complete detection coverage.
Data Source
AI summary
A method of detecting a region having a specific shape in a camera is provided. The method includes processing input image data in a camera and detecting the region having the specific shape. The method includes calculating gradation differences between a central pixel and respective peripheral pixels in each of local regions of an image frame, comparing an average gradation difference with each of the gradation differences and obtaining local gradient pattern (LGP) values based on a comparison result in each of the local regions, and detecting the region having the specific shape from the image frame using the LGP values obtained from the respective local regions.