Object Detection via Frame Difference Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile electronic devices, due to their limited memory and processing power, are not ideal for accurate object detection, which is typically a complex process requiring significant resources.
Innovation Solution
A simplified object detection system that uses a camera to capture two images, calculates the difference between them, and employs a classifier to declare object detection with lower memory and processing requirements, utilizing a difference calculator to generate a difference image and a classifier to identify objects based on predefined features or patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional object detection processes are used, then detection accuracy is improved, but memory and processing power requirements increase
Solution Approach 1:
The patent segments the object detection process into two distinct stages: (1) a simplified difference calculation stage that generates a difference image by comparing current and previous frames, and (2) a classification stage that applies trained classifiers to the difference image. This segmentation allows the computationally intensive classification to be performed only on relevant changes rather than full images, reducing memory and processing requirements while maintaining accuracy
Solution Approach 2:
The patent applies preliminary action by pre-training classification models offline using large datasets, then deploying compact, pre-trained classifiers to mobile devices. The difference calculator also performs preliminary processing by generating simplified difference images before classification, reducing the computational burden on mobile device processors during runtime
2Device complexity
If simplified object detection processes are used, then memory and processing power requirements are reduced, but detection accuracy deteriorates
Solution Approach 1:
The patent introduces a difference image as an intermediary representation between the raw image frames and the classification process. This intermediary contains only the relevant change information, allowing classifiers to focus computational resources on detecting objects within the difference regions rather than processing entire images, thereby maintaining accuracy with reduced complexity
Solution Approach 2:
The patent changes the parameter representation from full-color image frames to binary or grayscale difference images. By transforming the data format to highlight only changes between frames and suppressing static background information, the system reduces the amount of data requiring processing while preserving the essential information needed for accurate object detection
3Productivity
If frame difference calculation is performed, then data processing efficiency is improved, but signal noise increases
Solution Approach 1:
The patent extracts only the relevant change information from the difference calculation by applying thresholding and binary masking operations. This extraction process removes noise components that do not represent actual object changes, keeping only the significant differences that correspond to moving objects while discarding minor variations and noise
Data Source
AI summary
Object detection using a difference between image frames may include receiving a first image of a field of view, receiving a second image of the field of view, determining a difference between portions of the first image and corresponding portions of the second image, and declaring based on the difference between the portions of the first image and the corresponding portions of the second image that a specific object has been detected in the field of view.