Object Detection System Using Saliency Masks and Parallel Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems suffer from errors and false alarms, leading to desensitization and devaluation of security systems, as they struggle to accurately and efficiently detect objects of interest in a scene.
Innovation Solution
A detection system utilizing multiple processors and memory with executable instructions to extract features through parallel processing, generate pyramid histograms of visual words, and apply a sliding window protocol with support vector machine classification to identify objects of interest, while employing a saliency mask to focus on attention zones and ignore irrelevant areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection methods are used, then the system can detect objects in a scene, but the detection accuracy is low leading to false alarms
Solution Approach 1:
The patent divides the image processing into multiple independent threads, each handling different scales of object detection. The detector is segmented into thread-safe components that can operate independently and concurrently, improving both accuracy and reliability by analyzing multiple scales simultaneously without interference.
Solution Approach 2:
The patent introduces a saliency map as an intermediary component that identifies regions of interest in the image before detailed object detection occurs. This intermediary step filters out irrelevant areas and focuses computational resources on promising regions, reducing false alarms while maintaining high detection accuracy.
2Reliability
If comprehensive feature analysis is performed to improve detection accuracy, then the reliability increases, but the processing time increases
Solution Approach 1:
The patent performs preliminary saliency detection and region identification before conducting comprehensive feature analysis. By pre-identifying regions of interest using the saliency map, the system prepares the data structure and focuses subsequent detailed analysis only on relevant areas, maintaining high reliability while reducing overall processing time.
Solution Approach 2:
The detection process is segmented into parallel threads that independently analyze different scales. This segmentation allows comprehensive multi-scale feature analysis to be performed simultaneously rather than sequentially, improving detection reliability through thorough analysis while reducing total processing time through parallel execution.
3Productivity
If multi-threaded parallel processing is implemented to reduce processing time, then the productivity increases, but the system complexity increases
Solution Approach 1:
The patent designs thread-safe detector components that are universal and can be safely shared across multiple threads without requiring complex synchronization mechanisms. These multi-functional components can operate independently in different threads while maintaining consistency, achieving parallel processing speedup while minimizing the complexity overhead of thread management.
4Reliability
If the entire image is analyzed to ensure no objects are missed, then the detection completeness improves, but the processing time increases
Solution Approach 1:
The system performs preliminary saliency detection across the entire image to identify regions of interest before conducting detailed object detection. This preliminary action ensures that no potential objects are missed by marking all relevant regions, while simultaneously reducing the area requiring comprehensive analysis to only these marked regions, thus maintaining detection completeness while reducing processing time.
Solution Approach 2:
The image analysis is segmented into two phases: a quick saliency detection phase that scans the entire image to identify regions of interest, and a detailed detection phase that performs comprehensive analysis only on these identified regions. This segmentation ensures detection completeness by covering the entire image initially, while reducing analysis time by focusing detailed processing only on relevant segments.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a system for detecting an object of interest in a scene. The system operates by receiving an image frame of a scene and extracting features from the image frame, the features being descriptors. Tire descriptors are quantized to generate PHOW features. A sliding window protocol is implemented to slide a window over the image and analyze the PHOW features that fall inside the window. Finally, the system determines if the PHOW features represent the object of interest and, if so, then designates the window as a location in the image with a detected object of interest.