Object Detection Using Optical Flow for Moving Cameras
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection devices are ineffective in accurately detecting objects from images captured by moving cameras, such as those on robots or vehicles, due to the need for extensive pattern recognition processing and inability to adapt to changing camera positions and sizes.
Innovation Solution
An object detection device that incorporates an image capturer, an optical flow calculator to calculate motion between frames, and an object detector using the calculated optical flow to identify object areas, allowing for accurate detection even in moving images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern recognition processing is continuously executed while changing window size and position to detect objects, then object detection accuracy is improved, but operation amount increases enormously
Solution Approach 1:
The patent applies preliminary action by performing luminance change detection and optical flow calculation before the main pattern recognition processing. This pre-processing step identifies candidate regions that are more likely to contain objects, thereby reducing the number of windows that need to undergo computationally intensive pattern recognition processing while maintaining detection accuracy.
Solution Approach 2:
The patent segments the image processing task into multiple stages: first performing luminance change detection to identify regions with significant changes, then calculating optical flow in those regions, and finally applying pattern recognition only to candidate areas. This segmentation reduces the overall operation amount by avoiding exhaustive pattern recognition across the entire image.
2Device complexity
If luminance change detection is used to reduce the number of discrimination processing times, then operation amount is reduced, but detection accuracy fails for moving cameras
Solution Approach 1:
The patent introduces optical flow calculation as an intermediary step between luminance change detection and pattern recognition processing. The optical flow information serves as a mediator that compensates for camera movement by providing motion compensation data, allowing the system to maintain detection accuracy even when the camera is moving while still reducing the overall operation amount compared to exhaustive pattern recognition.
3Reliability
If exhaustive pattern recognition is performed on all possible windows, then detection coverage is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary luminance change detection and optical flow calculation to identify candidate regions before conducting pattern recognition processing. This preliminary action ensures that pattern recognition is applied only to regions that are likely to contain objects, maintaining detection coverage while significantly reducing processing time by avoiding exhaustive processing of all possible windows.
Solution Approach 2:
The patent applies partial action by performing exhaustive pattern recognition only on candidate regions identified through luminance change detection and optical flow analysis, rather than on all possible windows in the image. This selective approach maintains sufficient detection coverage for objects while reducing overall processing time.
Data Source
AI summary
An object detection device includes: an optical flow calculator to calculate an optical flow between images captured by the image capturer at different times; an evaluation value calculator to divide the image captured by the image capturer into areas, and calculate, for each divided area, an evaluation value by using the optical flows of pixels belonging to the divided area, the evaluation value indicating a measure of a possibility that the divided area is an object area representing part or whole of the object to be detected; and an area determinator to determine an area in an image, in which the object to be detected exists, by comparing the evaluation value of each divided area calculated by the evaluation value calculator with a threshold value.