Human Video Detection Using Frame Difference Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing surveillance cameras, such as doorbell cameras, face challenges in accurately detecting humans due to interference from other moving objects like cars, animals, and tree branches, leading to false alarms and reduced detection accuracy.
Innovation Solution
The integration of a four-channel image processing method, where a frame difference image is converted to grayscale and added as a fourth channel to the original RGB image, allows the camera to differentiate human motion from background motion, improving detection accuracy by training deep learning models to correlate human presence with motion information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion detectors are used to detect human presence, then detection coverage is improved, but false alarms increase due to other moving objects like cars, animals, and tree branches
Solution Approach 1:
The patent segments the detection task by separating motion detection from human classification. The motion detector identifies moving regions, and then video analytics with deep learning algorithms specifically analyze those regions to determine if humans are present, rather than treating all motion as human presence
Solution Approach 2:
The patent introduces video analytics and deep learning algorithms as an intermediary between motion detection and human detection. This intermediary layer processes the motion data and distinguishes human motion from other moving objects, reducing false alarms while maintaining detection coverage
2Measurement precision
If deep learning algorithms are used to differentiate human motion from background motion, then human detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies preliminary action by using motion detectors to pre-identify regions of interest before applying complex deep learning algorithms. This preprocessing step reduces the amount of data that needs to be processed by the algorithms, maintaining high accuracy while reducing overall system complexity
Solution Approach 2:
The patent uses partial action by applying deep learning algorithms only to regions identified as containing motion, rather than processing the entire video feed. This selective processing maintains detection accuracy while significantly reducing computational complexity
3Measurement precision
If four-channel image processing is used to fuse motion information with RGB images, then human detection accuracy is improved, but data processing requirements increase
Solution Approach 1:
The patent merges motion information from frame difference images with RGB image data by converting the difference image to grayscale and adding it as a fourth channel. This combination allows deep learning models to simultaneously process visual appearance and motion information, improving detection accuracy
Data Source
AI summary
Methods, systems, and apparatus for motion-based human video detection are disclosed. A method includes generating a representation of a difference between two frames of a video; providing, to an object detector, a particular frame of the two frames and the representation of the difference between two frames of the video; receiving an indication that the object detector detected an object in the particular frame; determining that detection of the object in the particular frame was a false positive detection; determining an amount of motion energy where the object was detected in the particular frame; and training the object detector based on penalization of the false positive detection in accordance with the amount of motion energy where the object was detected in the particular frame.