Motion-Vector Object Detection for Video Frame Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection methods for moving images face inefficiencies in processing large numbers of frames, particularly when new objects appear, and methods using motion compensation fail to maintain accuracy in scenes with significant motion changes.
Innovation Solution
An object detection device that calculates motion vectors between frames, using a threshold to determine if full detection is needed, and applies motion correction for efficient detection without reducing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection by CNN is performed in all frames of a moving image, then detection accuracy is maintained, but processing time and computational load increase significantly
Solution Approach 1:
The patent implements periodic full-frame object detection at key frames (e.g., every N frames or when motion threshold is exceeded) rather than continuous detection in every frame. Between key frames, object tracking is performed using motion compensation. This periodic approach maintains detection accuracy at critical moments while significantly reducing overall processing time and computational load.
Solution Approach 2:
The patent segments the detection process into two distinct modes: full-frame object detection using CNN at key frames, and object tracking using motion compensation at non-key frames. This segmentation allows the system to apply computationally intensive methods only when necessary while using lighter methods for routine frames, resolving the contradiction between accuracy and processing time.
2Productivity
If object tracking is performed only based on previous detection results without full detection, then processing efficiency improves, but new objects appearing in the frame are not detected
Solution Approach 1:
The patent uses motion compensation to generate predicted object positions and compares these predictions with actual frame content. When discrepancies exceed a threshold (indicating potential new objects or significant changes), the system triggers a full-frame object detection. This feedback mechanism ensures new objects are detected while maintaining high processing efficiency during stable periods.
Solution Approach 2:
The patent dynamically switches between tracking-only mode and full detection mode based on motion thresholds and detection confidence. When motion is small and tracking confidence is high, the system uses efficient tracking. When motion exceeds thresholds or confidence drops (suggesting new objects), full detection is activated. This dynamic adaptation resolves the contradiction between efficiency and reliability.
3Productivity
If motion compensation is applied in frames with large motion changes, then processing efficiency is maintained, but detection accuracy deteriorates
Solution Approach 1:
The patent calculates motion vectors and motion magnitude for each frame, then uses this motion information as feedback to determine whether to apply motion compensation or perform full-frame detection. When motion exceeds a predefined threshold, the system switches to full detection, preventing accuracy deterioration. When motion is within acceptable ranges, motion compensation is applied to maintain efficiency. This feedback-based decision resolves the contradiction between efficiency and accuracy.
Solution Approach 2:
The patent changes the detection strategy parameter based on motion magnitude. For frames with small motion, motion compensation parameters are applied to maintain efficiency. For frames with large motion exceeding thresholds, the system switches to full-frame detection parameters. This parameter change based on motion conditions resolves the contradiction by adapting the detection approach to the actual scene dynamics.
Data Source
AI summary
A calculation unit (22) sets each frame, of a moving image including a plurality of frames, as a target frame, calculates a motion vector with reference to a reference frame over an entirety of the target frames, and calculates an index indicating a magnitude of change between a key frame and the target frame, which is represented by using the motion vector, a judgement unit (25) judges whether or not the calculated index is equal to or greater than a predetermined threshold value, a first detection unit (27) detects a region indicating an object from the target frame by using an object detection model in a case in which the index is equal to greater than the threshold value, a second detection unit (28) detects a region on the target frame, which is obtained by correcting a position of a region detected in the reference frame by using the motion vector in a case in which the index is less than the threshold value, and an output unit (30) outputs information of the region detected by the first detection unit (27) and information of the region detected by the second detection unit (28).