Recognition Object Detection Using Multi-Threshold Binarization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional recognition object detection systems face challenges in accurately detecting vehicles at varying distances due to luminance issues, leading to increased processing load and cost, especially when using multiple imaging sensitivity settings.
Innovation Solution
A recognition object detecting apparatus with an imaging unit that generates image data using an imager with a high dynamic range characteristic, where the relation between luminance and output pixel values varies, and a detection unit that binarizes pixel values using multiple threshold values to generate binary images for accurate object detection, reducing the need for multiple sensitivity settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If imaging sensitivity is increased to detect low luminance objects at long distance, then detection accuracy for distant vehicles improves, but high luminance objects at short distance become saturated
Solution Approach 1:
The patent divides the luminance range into multiple segments (first luminance range and second luminance range) and applies different processing methods to each segment. Low luminance regions are processed with higher sensitivity while high luminance regions are processed with lower sensitivity, preventing saturation while maintaining detection accuracy for distant vehicles.
Solution Approach 2:
The patent applies different quality characteristics to different regions of the image based on local luminance properties. Regions with low luminance receive enhanced processing to improve detection, while regions with high luminance receive different processing to avoid saturation, creating locally optimized image quality throughout the frame.
2Measurement precision
If multiple images are taken with different sensitivity settings to detect both low and high luminance objects, then detection accuracy improves, but processing load increases
Solution Approach 1:
Instead of taking multiple separate images with different sensitivities, the patent segments a single image into different luminance ranges and processes each segment independently. This achieves the same detection accuracy as multiple images while processing only one image, significantly reducing computational load.
Solution Approach 2:
The patent applies processing only to the extent necessary for each luminance range. Rather than processing the entire image with multiple sensitivity settings, it applies appropriate processing only to the relevant luminance segments, reducing overall processing requirements while maintaining detection accuracy.
Data Source
AI summary
A recognition object detecting apparatus is provided which includes an imaging unit which generates image data representing a taken image, and a detection unit which detects a recognition object from the image represented by the image data. The imaging unit has a characteristic in which a relation between luminance and output pixel values varies depending on a luminance range. The detection unit binarizes the output pixel values of the image represented by the image data by using a plurality of threshold values to generate a plurality of binary images, and detects the recognition object based on the plurality of binary images.