Object Detection Post-Processing for Position Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Object detection techniques using AI, such as deep learning convolutional neural networks, face challenges in maintaining accurate object position information due to fluctuations when the background image changes, leading to unstable object detection regions.
Innovation Solution
An object detection post-processing device that accumulatively stores position information from multiple frames, calculates average or median positions, and determines final position information to stabilize object detection, even when the object detection region changes, by comparing current and previous frame data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If object detection is performed using AI techniques such as deep learning convolutional neural networks, then object detection capability is achieved, but position information becomes unstable when the background image changes
Solution Approach 1:
The patent applies preliminary action by storing position information from multiple previous frames before determining the final position. The processing device accumulates position data from N previous frames and uses this historical data to stabilize the current frame's object position, preventing instability caused by background changes.
Solution Approach 2:
The patent implements feedback by comparing the current frame's object position with accumulated position information from previous frames. The system uses this feedback to determine whether to update the stored position information, ensuring that only stable and accurate position data is retained, thereby maintaining position information stability.
2Measurement precision
If position information from multiple frames is accumulated and averaged, then position accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent applies partial action by accumulating position information from a specific number N of previous frames rather than all available frames. This selective accumulation achieves sufficient position accuracy while limiting the processing burden, balancing precision improvement with time efficiency.
Solution Approach 2:
The patent uses parameter changes by adjusting the number of frames N used for accumulation based on processing requirements. The system can dynamically change the parameter N to balance between position accuracy and processing time, allowing flexible optimization for different application scenarios.
Data Source
AI summary
An object detection post-processing device of a display device, includes: an input to receive position information of an object that is detected from input image data for the display device in a current frame; a position information storage device to accumulatively store the position information of a plurality of previous frames; an accumulated position calculator to generate accumulated position information according to the position information of the current frame and the position information of the plurality of previous frames; and a position information determiner to determine final position information according to the accumulated position information, and to output the final position information.