Video Object Removal Detection via Segmented Reference Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video image processing systems for security and surveillance struggle with accurately detecting removed objects due to environmental changes and occlusion, often resulting in false alarms.
Innovation Solution
A method and system that generates 'reference with hole' and 'object reference' images using segmentation and hole-filling algorithms, allowing for robust detection of object removal and occlusion by comparing these reference images with input images, and updating them periodically or upon environmental changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If change detection approaches are used to detect removed objects, then the detection capability is improved, but false alarms increase due to occlusion and environmental changes
Solution Approach 1:
The patent segments the reference image into two distinct components: the object reference image (containing only the target object) and the background reference image (containing only the background). This segmentation allows independent comparison of object presence and background changes, enabling the system to distinguish between actual object removal and occlusion events where the object is merely obscured by foreground elements.
Solution Approach 2:
The patent introduces an intermediary component - the object reference image - that serves as a mediator between the input image and the detection decision. By comparing the input image against this intermediary reference that contains only the object without background context, the system can detect object removal while being insensitive to background environmental changes, thus reducing false alarms.
2Measurement precision
If conventional image processing techniques are used, then the system complexity is low, but the ability to distinguish object removal from occlusion is poor
Solution Approach 1:
The patent applies segmentation by dividing the reference image into two separate reference images: one containing only the object and another containing only the background. This is achieved by detecting the object in the reference image and creating a mask, then using this mask to generate the object reference image and the background reference image separately.
Solution Approach 2:
The patent applies local quality by using different reference images for different comparison purposes: the object reference image is used specifically for detecting object removal by comparing object regions, while the background reference image is used for detecting environmental changes by comparing background regions. Each reference image has optimized quality for its specific detection task.
3Productivity
If change detection is applied to monitor object presence, then the detection speed is fast, but the reliability decreases due to sensitivity to environmental changes
Solution Approach 1:
The patent segments the monitoring task into two parallel processes: object removal detection using the object reference image and environmental change detection using the background reference image. This allows simultaneous processing without interference, maintaining fast detection speed while improving reliability by preventing false alarms from environmental changes.
Solution Approach 2:
The patent optimizes each reference image for its specific purpose: the object reference image preserves object characteristics for accurate removal detection, while the background reference image captures background characteristics for environmental change monitoring. This local optimization ensures each comparison is highly reliable for its intended function.
Data Source
AI summary
A method, system and program product for the detection of removed objects in video images is disclosed. An object (or objects) is within a fixed camera's FOV (Field of View), is designated to be monitored. If an object is not designated, an algorithm can be used to segment and identify all the objects in the image that is acquired by a fixed camera. In such case, all these objects can be monitored. A reference frame around the monitored object that also includes the surrounding background can be determined. The object can be extracted out of the reference frame, thereby creating a “reference with hole” image and an object reference image. The hole in the “reference with hole” image can be filled by either an in-painting algorithm or a texture synthesis algorithm depending on the complexity of the surrounding background.