Object Detection Device Using Low-Resolution Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection systems face challenges with excessive processing load and insufficient accuracy in certain scenarios.
Innovation Solution
The proposed object detection device and method involve generating a low-resolution image from a captured image, defining partial area sets with varying pixel numbers, aggregating feature amount differences between partial areas and their peripherals, and extracting points of interest to reduce processing load while maintaining detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution processing is performed on the entire captured image to ensure detection accuracy, then object detection accuracy is improved, but processing load increases excessively
Solution Approach 1:
The captured image is divided into multiple blocks, and each block is further divided into sub-blocks. This segmentation allows the system to process only specific regions of interest at high resolution rather than the entire image, thereby reducing overall processing load while maintaining detection accuracy for relevant areas.
Solution Approach 2:
The patent applies different processing qualities to different regions of the image. High-resolution processing is applied selectively to blocks containing objects of interest, while other regions are processed at lower resolution or skipped entirely. This local quality approach ensures detection accuracy is maintained where needed while reducing processing load in less critical areas.
2Measurement precision
If the entire captured image is processed at high resolution to improve detection accuracy, then detection performance is improved, but processing time increases
Solution Approach 1:
The image processing is segmented into multiple stages: initial low-resolution scanning of the entire image to identify potential objects, followed by high-resolution processing only of identified regions of interest. This multi-stage segmentation significantly reduces total processing time while maintaining detection accuracy.
Solution Approach 2:
The patent performs preliminary low-resolution processing of the entire image to identify candidate regions before applying high-resolution processing. This preliminary action filters out areas that do not contain objects of interest, preventing unnecessary high-resolution processing and thereby reducing overall processing time.
3Device complexity
If uniform block sizes are used for dividing the image, then processing is simpler, but detection robustness against object size variations deteriorates
Solution Approach 1:
The patent employs asymmetric block division where blocks are divided into different numbers of sub-blocks based on their content and position. Rather than uniform division, each block is adaptively segmented into varying numbers of sub-blocks (e.g., 2×2, 2×3, 3×3 configurations), which improves detection robustness against varying object sizes and positions while managing processing complexity through systematic variation.
Data Source
AI summary
An object detection device includes a storage device which has stored a program, and a hardware processor, in which the hardware processor executes the program stored in the storage device, thereby acquiring a captured image in which a road is captured, generating a low-resolution image in which an image quality of the captured image is lowered, defining one or more partial area sets with each of the one or more partial area sets defined by cutting out a plurality of partial areas from the low-resolution image, deriving an aggregated value obtained by aggregating differences in feature amount between the partial areas included in each of the one or more partial area set and the peripheral partial areas, and extracting a point of interest based on the aggregated value.