Image Recognition Security with Adaptive Detection Modes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition-based security systems face inefficiencies in resource utilization and false acceptance rates due to the repetitive performance of face detection, recognition, and unauthorized object detection on all input frames, especially in environments where the security context remains unchanged.
Innovation Solution
A processor-implemented method and apparatus that determine the operation mode based on the state of a terminal (locked or unlocked) and the state maintenance time, selectively performing face detection, facial recognition, and unauthorized object detection only when necessary, such as when a background region change is detected or when the state maintenance time meets a preset threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If face detection, facial recognition, and unauthorized object detection are performed on all input frames, then security reliability is improved, but computing resource usage increases and processing efficiency decreases
Solution Approach 1:
The system dynamically adjusts the operation mode based on the terminal state (locked/unlocked) and state maintenance time. When the terminal is unlocked and state maintenance time is short, only face detection is performed. When state maintenance time meets the threshold, full detection including unauthorized object detection is performed. This dynamic adjustment resolves the contradiction by making processing intensity adaptive to security needs.
Solution Approach 2:
The system changes the parameter of detection intensity based on the operation mode. In the first operation mode, only face detection is performed. In the second operation mode, face detection, facial recognition, and unauthorized object detection are sequentially performed. This parameter change allows the system to balance security reliability with processing efficiency by adjusting detection comprehensiveness according to security context.
2Reliability
If full detection operations are performed on all frames, then false acceptance rates are reduced, but resource consumption increases
Solution Approach 1:
The system performs partial detection actions based on security context. When the terminal is in unlocked state with short state maintenance time, only essential face detection is performed (partial action). When state maintenance time meets the threshold, full detection including unauthorized object detection is performed. This partial action approach reduces computing resource usage while maintaining adequate security by performing only necessary detections.
Solution Approach 2:
The system performs comprehensive detection operations periodically based on the state maintenance time threshold. Instead of continuous full detection on every frame, the system switches between operation modes periodically, performing full detection (unauthorized object detection included) only when the state maintenance time meets the preset threshold, thereby reducing overall resource consumption while maintaining security.
3Reliability
If repetitive detection operations are performed on all frames, then security coverage is improved, but processing time increases
Solution Approach 1:
The system dynamically adjusts processing operations based on the current operation mode determined by terminal state and state maintenance time. This dynamic approach reduces unnecessary processing time by performing only essential face detection when security context indicates lower risk, while maintaining comprehensive security coverage by performing full detection including unauthorized object detection when state maintenance time meets the threshold.
Data Source
AI summary
An apparatus and a method with image recognition-based security are disclosed. The method includes, for an unlocked terminal, tracking a face detected in a previous frame, detecting a background region change between the previous frame and a current frame based on a region of the tracked face, when the background region change is not detected, determining whether a state maintenance time fails to meet a preset time, in response to the state maintenance time failing to meet the preset time, determining an operation mode to be a first operation mode for determining whether recognition succeeds for the current frame, performing the first operation mode, including performing face detection with respect to the current frame, and maintaining the unlocked state of the terminal for the current frame when the face is detected as a result of the performing of the face detection, representing that the recognition succeeded for the current frame.