Region-Based CNN Object Detection in Transformed Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection technologies in images are limited by relying on pre-defined features, failing to recognize objects when they appear transformed or in combination with other objects, and are computationally intensive, making real-time detection challenging, especially in transformed images like x-rays or infrared.
Innovation Solution
The use of a region-based convolutional neural network (RCNN) trained on transformed images without relying on pre-defined features, combined with a modified fuzzy C-means algorithm and Generative Adversarial Networks to generate and identify transformed objects, allowing for efficient detection in various image types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If pre-defined features are used for object detection, then detection speed is improved, but detection accuracy deteriorates when objects appear transformed or in combination with other objects
Solution Approach 1:
The patent transforms the input image into a different domain (e.g., from visible light to infrared or thermal imaging) to change the parameters of the image data. This transformation allows the CNN to detect objects based on their thermal or infrared signatures rather than visual appearance, improving detection accuracy for transformed objects while maintaining detection speed through efficient transformation algorithms
2Measurement precision
If the entire image is processed by CNN, then detection accuracy is improved, but computational cost increases
Solution Approach 1:
The patent extracts and processes only the transformed image data that contains relevant object information, rather than processing the entire original image. By focusing computational resources on the transformed representation and using region-based CNN approaches, the system achieves high detection accuracy while reducing overall computational cost
3Ease of manufacture
If traditional object detection methods are used, then implementation simplicity is maintained, but adaptability to transformed images deteriorates
Solution Approach 1:
The patent implements a universal detection system using CNN that can handle multiple types of images (visible light, infrared, thermal, and transformed combinations) through a single trained model. The CNN architecture is designed to be multi-functional, accepting various input transformations and detecting objects across different domains, thereby improving adaptability while maintaining implementation simplicity through a unified approach
Data Source
AI summary
A transformed image is received. The transformed image includes an other-than-visible light image that has been captured using a transformation device. A region of the transformed image is isolated, the region being less than an entirety of the transformed image. By applying to the region a convolutional Neural Network (CNN) which executes using a processor and a memory, and by processing only the region of the transformed image, an object of interest is detected in the region. Upon detecting, an indication is produced to indicate the presence of the object of interest in the region.