Object Detection via Masked Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for security surveillance generates overwhelming amounts of video data, leading to high costs and human error due to limited attention from security practitioners, necessitating an efficient automatic solution for analyzing images captured by cameras.
Innovation Solution
A monitor method and system that captures images, calculates initial probabilities of object presence, applies a mask to cover portions of the image, and adjusts the mask's location to detect the object's presence and location using a convolutional neural network model, allowing for flexible resolution adjustment and resource optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual monitoring of surveillance images is used, then detection accuracy can be maintained, but labor costs and time consumption increase significantly
Solution Approach 1:
The patent divides the image analysis task into multiple stages: initial probability calculation for the entire image, followed by mask-based segmentation of high-probability regions, and finally detailed object detection only in those segmented regions. This segmentation approach allows the system to maintain high detection accuracy while significantly reducing the time and computational resources needed compared to manual monitoring of entire images.
Solution Approach 2:
The patent performs preliminary probability calculation across the entire image before detailed detection. By pre-calculating which regions have high probability of containing objects, the system prepares in advance for the detailed detection phase, avoiding the need to manually examine every pixel and reducing overall detection time while maintaining accuracy.
2Measurement precision
If full image analysis is performed to ensure accurate object detection, then detection precision is improved, but computational resources and processing time increase
Solution Approach 1:
The patent segments the image into high-probability regions (using masks) and low-probability regions, then applies detailed detection algorithms only to the high-probability segments. This reduces computational resource consumption significantly while maintaining detection precision in the regions that matter most.
Solution Approach 2:
The patent applies different levels of analysis quality to different regions of the image: high-quality detailed detection in high-probability regions and lower-quality or no detection in low-probability regions. This local quality approach optimizes the balance between detection precision and computational resource usage.
3Loss of energy
If mask-based region covering is applied to reduce computation, then resource consumption decreases, but detection accuracy may be compromised
Solution Approach 1:
The patent performs preliminary probability calculation across the entire image before applying masks. This preliminary action ensures that all regions containing objects are identified with high probability before the mask is applied, so no objects are missed in the subsequent reduced-computation phase, maintaining detection accuracy while reducing resource consumption.
Solution Approach 2:
The system uses the initial probability calculation results as feedback to determine where to apply masks and how to proceed with detailed detection. This feedback mechanism ensures that the mask application strategy is optimized to preserve detection accuracy in critical regions while reducing computation in less critical areas.
Data Source
AI summary
A monitor method for detecting an object includes capturing an image; calculating an initial probability of the object existing in the image; applying a mask to cover a first portion of the image if the initial probability is higher than a threshold; calculating a first probability of the object existing in the image excluding the first portion; and using at least the first probability to detect the location the object in the image.