Image Processing Device Static Dynamic Region Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face challenges in accurately estimating the position and attitude of objects when dynamic and static regions are mixed in an image, leading to difficulties in capturing the change over time of position and attitude, especially when camera work involves both types of regions.
Innovation Solution
An image processing device that combines image data and depth data to enhance estimation accuracy by calculating position and attitude changes using a statistic of the calculated changes, employing a configuration with an acquisition unit, calculating unit, and determining unit to differentiate between static and dynamic regions, thereby suppressing the influence of dynamic regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional background region extraction using difference between continuous frames is employed, then the processing is simple, but the position and attitude estimation accuracy deteriorates when dynamic and static regions are mixed in the image
Solution Approach 1:
The image is segmented into static regions and dynamic regions based on depth information and motion detection. The calculating unit separately calculates position and attitude changes for each region type, preventing dynamic region motion from contaminating the static region estimation. This segmentation approach resolves the contradiction by maintaining high estimation accuracy through region-specific processing while managing complexity through systematic classification.
Solution Approach 2:
Depth data serves as an intermediary to distinguish between static and dynamic regions. By utilizing depth information from the depth image alongside the captured image, the system can identify which regions are likely to be static (e.g., background) versus dynamic (e.g., moving objects). This intermediary depth information enables accurate position and attitude estimation without requiring complex processing of the entire image.
2Measurement precision
If depth data is used together with image feature alignment, then the position and attitude estimation accuracy is improved, but the processing complexity increases
Solution Approach 1:
Instead of processing the entire image with both depth data and feature alignment, the system applies these methods selectively to identified static regions. The calculating unit focuses computational resources on regions that contribute most accurately to position and attitude estimation, performing partial processing that achieves high precision without the full complexity applied uniformly across the entire image.
3Measurement precision
If motion vector-based estimation is used, then the processing is straightforward, but the estimation accuracy deteriorates when dynamic regions are present in the image
Solution Approach 1:
The image is segmented into static and dynamic regions using depth information and motion analysis. By isolating static regions from dynamic regions, the system can apply motion vector-based estimation only to the static portions of the image. This segmentation maintains processing simplicity while eliminating the accuracy deterioration caused by dynamic region contamination.
Solution Approach 2:
Dynamic regions are extracted and excluded from the position and attitude estimation calculation. The determining unit identifies and removes the influence of dynamic regions from the overall estimation process, allowing the straightforward motion vector-based method to operate on purified static region data, thereby maintaining both simplicity and accuracy.
Data Source
AI summary
An image processing device includes an acquisition unit configured to acquire an image data and a depth data correspond to the image data; a calculating unit configured to calculate a position and attitude change for each depth from the image data and the depth data; and a determining unit configured to determine a position and attitude change of a whole image by a position and attitude change data calculated by the calculating unit based on a statistic of the position and attitude change calculated by the calculating unit.