Video Object Detection With Dynamic Region Frequencies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection technologies struggle to achieve both object tracking performance and object detection performance when processing ultra-high definition videos, often leading to inaccurate object tracking during sudden changes due to increased processing load and reduced partial surfaces.
Innovation Solution
An object detection device and method that divides images into partial surfaces based on differences between consecutive frames, allocates detection frequencies, and combines results from both partial and entire surfaces to optimize performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the total number of partial surfaces is reduced to increase object detection frequency, then object tracking performance is improved, but image reduction on each partial surface leads to deterioration of object detection performance
Solution Approach 1:
The patent dynamically adjusts the number of partial surfaces and detection frequencies based on object characteristics and movement patterns. The system determines optimal detection parameters in real-time, changing the number of partial surfaces from a fixed value to a dynamic parameter that adapts to scene complexity, object speed, and detection requirements, thereby resolving the contradiction between tracking smoothness and detection accuracy
Solution Approach 2:
The system changes multiple parameters simultaneously including the number of partial surfaces, detection frequency, and partial surface size based on object attributes such as movement speed, size, and importance. By adjusting these parameters dynamically, the system optimizes both tracking performance and detection accuracy for different scenarios without being constrained by fixed parameter settings
2Productivity
If the number of partial surfaces on which detection is executable is small, then processing amount is reduced, but the number of times of thinning increases and object tracking accuracy deteriorates during sudden changes
Solution Approach 1:
The patent applies different detection frequencies to different partial surfaces based on local characteristics such as object presence, movement intensity, and scene complexity. High-priority regions with objects or sudden changes receive higher detection frequencies while low-priority regions use lower frequencies, thereby maintaining tracking accuracy during sudden changes while reducing overall processing load
Solution Approach 2:
The system performs preliminary analysis of image differences between consecutive frames to predict regions where objects may appear or move suddenly. By pre-identifying these high-risk regions, the system can allocate higher detection frequencies to specific partial surfaces before objects actually appear, ensuring tracking accuracy is maintained during sudden changes without uniformly increasing processing across all surfaces
Data Source
AI summary
An object detection device that detects an object from an image included in a moving image includes: an acquisition unit configured to acquire the image from the moving image; a number-of-faces setting unit configured to set a number of faces for dividing the image into a plurality of partial faces using a difference between consecutive images; an allocation control unit configured to allocate a frequency of detecting the object for each of the divided partial faces; a division processing unit configured to divide the image into a plurality of partial faces depending on the set number of faces and to detect an object from the partial faces in accordance with the allocated frequency; an overall processing unit configured to reduce the image to an entire face indicating the entire image and to detect an object from the entire face; and a combination processing unit configured to combine respective detection results detected from the partial faces and the entire face to detect an object from the image.