Video Object Detection via Region Overlap Discrimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting moving objects in videos often fail to accurately identify objects due to mismatched detection regions between moving object detection and object detection, leading to incorrect or incomplete detection results.
Innovation Solution
A video processing apparatus that combines moving object detection and object detection by using a first detection unit to identify moving objects, a second detection unit to detect objects with specific shapes, and a discrimination unit to verify the accuracy of detected objects based on the overlap ratio between the moving object region and the extracted region.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If only moving object detection is used, then detection speed is improved, but detection accuracy deteriorates because non-target objects may be detected
Solution Approach 1:
The patent combines moving object detection results with object detection results by merging the detection regions and performing pattern matching on the combined region. This allows the system to maintain the speed advantage of moving object detection while adding the accuracy of object detection through pattern matching, thereby resolving the contradiction between detection speed and accuracy.
2Measurement precision
If only object detection is used, then detection accuracy is improved, but detection capability deteriorates because moving objects cannot be detected
Solution Approach 1:
The patent merges moving object detection and object detection by combining their detection regions and performing pattern matching on the merged region. This approach maintains the ability to detect moving objects while improving accuracy through pattern matching, thereby resolving the contradiction between detection accuracy and detection capability.
3Measurement precision
If both moving object detection and object detection are used together, then detection accuracy is improved, but device complexity increases due to region mismatch problems
Solution Approach 1:
The patent segments the detection process into distinct stages: moving object detection, region merging, and pattern matching. By dividing the complex task of combined detection into manageable segments, the system reduces overall complexity while maintaining the accuracy benefits of using both detection methods together.
Solution Approach 2:
The patent performs preliminary merging of detection regions from moving object detection and object detection before performing pattern matching. This preliminary action simplifies the subsequent pattern matching process by providing a pre-processed, combined detection region, thereby reducing the complexity of the overall system.
4Adaptability or versatility
If detection regions from moving object detection and object detection do not match, then detection flexibility is improved, but detection reliability deteriorates because objects cannot be accurately detected
Solution Approach 1:
The patent performs preliminary merging of detection regions from moving object detection and object detection before performing pattern matching. This preliminary merging action ensures that even when individual detection regions do not match perfectly, the combined region captures the complete object, thereby maintaining detection reliability while preserving the flexibility of using multiple detection methods.
Data Source
AI summary
A video processing apparatus includes: a first detection unit configured to detect a moving object from a movie; a second detection unit configured to detect an object having a predetermined shape from the movie; an extraction unit configured to extract a partial region of a region in which the second detection unit has detected the object having the predetermined shape in the movie; and a discrimination unit configured to discriminate whether the object detected by the second detection unit is a certain object depending on a ratio of a size of an overlapping region to a size of an extracted region extracted by the extraction unit, the overlapping region being a region where a region in which the first detection unit has detected the moving object in the movie and the extracted region overlap with each other.