Moving Image Tracking Scene Detection Using Object Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately distinguish between scenes where an imaging apparatus is tracking an object and where it is shifting the line of sight to another object, especially when the object's action changes during tracking.
Innovation Solution
A processing device and method that extracts sections where the camera's angle of view continuously moves, derives object information, and identifies whether the section is tracking an object or not based on object features, using processors and memories to output appropriate section information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If motion vector-based correlation image detection is used to distinguish tracking scenes, then the processing method is simple, but the distinction accuracy deteriorates when object actions change during tracking
Solution Approach 1:
The patent changes the detection parameters from motion vector-based correlation to optical flow field characteristics. By analyzing the continuity and smoothness of optical flow vectors across multiple frames, the system can distinguish tracking scenes more accurately even when object actions change, while maintaining computational feasibility through parameter optimization.
Solution Approach 2:
The patent introduces optical flow field analysis as an intermediary between raw video frames and scene distinction results. The optical flow field serves as a mediator that captures continuous motion patterns, allowing the system to differentiate between tracking and non-tracking scenes more reliably than direct motion vector comparison.
2Measurement precision
If optical flow field analysis is used to improve tracking scene distinction accuracy, then the distinction accuracy improves, but the processing complexity increases
Solution Approach 1:
The patent segments the optical flow field analysis into distinct processing stages: optical flow calculation, correlation image generation, and scene distinction determination. This segmentation allows each stage to be optimized independently, reducing overall system complexity while maintaining high distinction accuracy through specialized processing at each stage.
3Productivity
If motion vector correlation detection is used, then the processing speed is fast, but the ability to correctly identify sequential actions of the same object deteriorates
Solution Approach 1:
The patent maintains continuity of useful action by analyzing optical flow fields across multiple consecutive frames rather than comparing isolated motion vectors. This continuous analysis preserves the temporal relationships in object motion, enabling reliable identification of sequential actions while maintaining processing efficiency through optimized frame sampling and optical flow computation.
Data Source
AI summary
A processing device includes one or more memories, and one or more processors in communication with the one or more memories, wherein the one or more processors and the one or more memories are configured to extract, from a captured moving image, a section where an angle of view of an imaging apparatus continuously moves, derive object information indicating a description according to a feature of an object in the extracted section, derive, based on the derived object information, first section information for identification of a state where the section included in the moving image is a section where an object is tracked, or second section information for identification of a state where the section included in the moving image is a section where no object is tracked and output the second section information.