Compressed Domain Motion Detection for Smart Cameras
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion detection systems in smart camera devices are computationally expensive and often generate false positives, particularly on under-powered devices with low-end camera hardware, due to their inability to effectively differentiate between foreground and background motion.
Innovation Solution
The system employs a method to detect motion using compressed video data by extracting macroblocks and motion vectors, calculating frame-level and similarity features, and using a trained model to predict motion, thereby reducing false positives and processing complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing motion detection algorithms are used, then motion detection capability is provided, but computational expense increases and false positives increase
Solution Approach 1:
The video processing is segmented into multiple stages: extracting only relevant features (macroblocks and motion vectors) from compressed video data, calculating frame-level features from these extracted elements, then calculating similarity features between consecutive frames. This segmentation allows motion detection to operate on a subset of video data rather than full frames, reducing computational complexity while maintaining detection accuracy.
Solution Approach 2:
The patent extracts only the essential motion-related components (macroblocks and motion vectors) from the compressed video data, discarding redundant information. By working exclusively with these extracted features rather than full video frames, the system achieves accurate motion detection with significantly reduced computational requirements.
2Reliability
If existing motion detection algorithms are used, then motion detection is performed, but false positives increase due to inability to differentiate foreground and background motion
Solution Approach 1:
The system calculates similarity features by comparing frame-level features across consecutive frames, creating a feedback mechanism that tracks motion patterns over time. This temporal feedback allows the system to distinguish between transient background motion (like swaying trees) and sustained foreground motion (like intruders), reducing false positives while maintaining detection sensitivity.
3Measurement precision
If full video decoding and processing is performed, then accurate motion detection is achieved, but processing speed decreases on under-powered devices
Solution Approach 1:
The patent segments the video data processing by operating directly on compressed video bitstreams and extracting only motion-relevant features (macroblocks and motion vectors) rather than fully decoding entire video frames. This segmentation enables accurate motion detection while maintaining processing speed suitable for under-powered smart camera devices.
4Reliability
If comprehensive video analysis is performed, then accurate event detection is provided, but cloud storage and delivery costs increase due to unnecessary video recording
Solution Approach 1:
The system performs preliminary motion analysis on compressed video data before deciding whether to record full video clips. By pre-processing the compressed stream to identify frames with genuine motion (distinguishing from background motion), the system can selectively record only relevant events, improving detection accuracy while reducing unnecessary video storage and delivery costs.
Data Source
AI summary
Methods and systems for detecting motion using video data from a smart camera device. In some embodiments, and by non-limiting example, this disclosure is directed to detecting motion in the compressed video domain. One aspect is a method comprising receiving compressed video data, extracting macroblocks and motion vectors for a plurality of frames in the compressed video data, identifying frame-level features for each of the plurality of frames based on the macroblocks and the motion vectors, calculating similarity features for each of the identified frame-level features based on the frame-level features identified in consecutive frames, and predicting motion for each of the plurality of frames by providing the frame-level features and the similarity features into a model trained to detect motion.