Image Processing Change Detection for Moving Camera
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional change detection algorithms are not applicable to video data taken with a moving camera, as the changing background and perspective of the camera complicate the detection of changes in a scene.
Innovation Solution
A change detection algorithm implemented on an unmanned aircraft using a camera and processor, which captures high-resolution images and employs a Self Adaptive Discriminant (SAD) Filter and RGB histogram analysis to track regions of interest and detect significant changes, even with a shifting background.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional change detection algorithms are used with fixed cameras, then change detection is accurate and reliable, but the system cannot handle video data from moving cameras
Solution Approach 1:
The patent applies dynamics by making the background model adaptive and time-varying to accommodate the moving camera. Instead of using a static background image, the system continuously updates the background model to track the changing scene, allowing reliable change detection despite camera motion and perspective changes.
Solution Approach 2:
The system changes parameters by using multiple background models with different time constants and adapting the background subtraction thresholds dynamically. This allows the detection algorithm to maintain accuracy across varying camera positions and lighting conditions by adjusting detection parameters rather than relying on fixed thresholds.
2Reliability
If all video frames are stored for analysis, then complete change detection is possible, but bandwidth and storage requirements increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for change detection by comparing each frame against a learned background model and transmitting only the difference data or alerts about detected changes. This extraction approach maintains detection completeness while dramatically reducing the quantity of data that needs to be stored and transmitted.
Solution Approach 2:
The system performs partial action by applying change detection selectively - it processes all frames for background learning but only flags and transmits frames where actual changes are detected. This partial processing approach ensures no changes are missed while minimizing the data volume that requires storage and bandwidth.
3Measurement precision
If high-resolution images are captured to improve detection precision, then change detection accuracy improves, but data processing complexity and resource requirements increase
Solution Approach 1:
The patent segments the image processing task by first applying background subtraction to identify regions of interest, then applying more complex analysis only to those specific regions rather than the entire high-resolution image. This segmentation maintains detection precision in critical areas while reducing overall processing complexity and resource requirements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Apparatus and method for processing a sequence of images of a scene, the method comprising: tracking a region of interest (14) in the sequence of images (e.g. using a Self Adaptive Discriminant filter); selecting a particular image (12) in the sequence; selecting a set of images from the sequence,the set comprising one or more images that precede the particular image (12)in the sequence of images; for each pixel in the region of interest (14) in the particular image (12), determining a value for a parameter; for each pixel in the region of interest (14) of each image in the set of images, determining a value for the parameters; and comparing a function of the determined values for the region of interest (14) in the particular image to a further function of the determined values for the regions of interest (14) in the images in the set of images.