Image Processing Device Human Body Detection Region Exclusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies require significant computational resources for human body detection in images due to the need to scan images at various scales, leading to inefficient processing times.
Innovation Solution
An image processing device that includes a detection unit for identifying objects and an excluding unit to skip processing areas where objects have already been detected, reducing the computational load by generating reduced images and setting non-detection areas based on detected object positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern matching processing is performed to detect human bodies of various sizes by scanning the entire captured image at various scales, then detection accuracy is improved, but the calculation amount increases and processing time lengthens
Solution Approach 1:
The patent divides the image processing into multiple scales by generating reduced images at different resolution levels. Detection is performed hierarchically from large-scale reduced images to small-scale original images, segmenting the overall detection task into manageable stages that reduce total calculation amount while maintaining comprehensive detection accuracy
Solution Approach 2:
The patent performs preliminary detection on reduced images at lower resolution before conducting detailed detection on the original high-resolution image. This preliminary action identifies potential detection regions in advance, allowing the system to focus computational resources on areas of interest and avoid exhaustive scanning of the entire original image
2Measurement precision
If the entire captured image is scanned at various scales for human body detection, then comprehensive detection coverage is achieved, but the calculation amount increases significantly
Solution Approach 1:
The patent segments the detection process into multiple resolution levels using reduced images. By performing detection on progressively refined image versions rather than the complete original image at all scales, the system achieves comprehensive coverage while significantly reducing the total computational energy required
Solution Approach 2:
The patent creates reduced copies of the original image at different resolution levels. These copied reduced images serve as proxies for preliminary detection, allowing the system to evaluate large areas with minimal computational cost before investing resources in detailed analysis of specific regions
3Measurement precision
If detection processing is performed on the entire captured image without exclusion, then no detection areas are missed, but processing time increases due to redundant calculations
Solution Approach 1:
The patent performs preliminary detection on reduced images to identify regions containing human bodies before processing the original image. This preliminary action allows the system to exclude areas where no human bodies are present from detailed processing, eliminating redundant calculations while ensuring all actual human bodies are detected
Solution Approach 2:
The patent extracts and excludes regions determined to be free of human bodies based on reduced image analysis. By taking out these non-relevant areas from the processing region of the original image, the system eliminates wasted computational time while maintaining complete detection coverage of actual human bodies
Data Source
AI summary
The present invention is directed to reducing a calculation amount for human body detection to achieve high speed processing. Detection processing is executed to detect a predetermined object in an image captured by an image capturing unit. When the predetermined object is detected using a pattern having a predetermined size, a partial area of the image is excluded from a processing region in which the detection processing of detecting the predetermined object is executed using the pattern having the predetermined size, based on a position at which the predetermined object is detected.