Network Camera Object Detection with Selective Masking Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In monitoring systems, existing techniques for applying change processing such as masking and encryption to images from network cameras often fail to allow observation of objects relevant to detected events, especially when these objects have moved out of the screen or when there is a delay in event detection, leading to incomplete or missed observations.
Innovation Solution
An image processing apparatus that includes a detection unit for identifying unidentifiable objects, an analysis unit for analyzing image data, and a processing unit that applies change processing to make objects unidentifiable, while a determination unit ensures that objects relevant to monitored events are made identifiable by canceling change processing, allowing for accurate observation and privacy protection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If change processing is regularly applied to all detected objects for privacy protection, then privacy protection is improved, but the ability to observe objects relevant to detected events deteriorates
Solution Approach 1:
The patent applies different processing qualities to different regions: objects identified as event-relevant receive no change processing (maintaining full visibility), while other objects receive change processing (masking, encryption, or blurring). This spatial differentiation of processing quality resolves the contradiction by making privacy protection local rather than universal, allowing event-relevant objects to remain observable while protecting privacy of irrelevant objects.
Solution Approach 2:
The system performs preliminary identification of event-relevant objects before applying change processing. By analyzing image data, detecting events, and identifying objects associated with detected events in advance, the system determines which objects should be exempt from change processing. This preliminary action ensures that when change processing is applied, event-relevant objects are already marked for exemption, preventing information loss.
2Productivity
If change processing is canceled only for objects currently in the screen when an event is detected, then processing efficiency is improved, but the ability to observe objects that have moved out of the screen deteriorates
Solution Approach 1:
The system performs preliminary tracking and identification of objects before events occur. By continuously tracking object positions and maintaining object information in a database, the system is prepared to quickly identify event-relevant objects when events are detected, even if those objects have moved out of the current screen. This preliminary preparation enables efficient processing without losing observation capability.
Solution Approach 2:
The patent extends the observation dimension from spatial (current screen only) to temporal (historical and future positions). By tracking object trajectories and using prediction algorithms, the system can identify objects that have moved out of the screen based on their movement patterns. This dimensional extension allows the system to maintain observation capability beyond the immediate screen boundaries while keeping processing efficient.
Data Source
AI summary
An object detection unit of a network camera detects an object including a human figure from image data. An event detection unit analyzes the image data based on a result of the object detection by the object detection unit. Based on a result of the image data analysis, an event detection unit determines to make identifiable to a client a human figure that should be monitored by the client and to apply masking to a human figure that should not be monitored by the client to make it unidentifiable to the client.