Neural Network Flame Detection via Synthetic Image Augmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sensors for fire detection, including those using image processing, face challenges in accurately and efficiently detecting fires across wider areas due to environmental influences and the difficulty in securing high-quality learning data for neural network models, which affects their performance in determining fire occurrences.
Innovation Solution
A learning method for neural network models involves generating learning images by combining real and fake fire images with background images, using multiple neural networks to update feature extraction layers and improve data quality, and adapting to various data sizes and resolutions, enabling efficient and accurate fire detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional sensors (heat, smoke, flame) are used for fire detection, then local area detection is achieved, but detection accuracy deteriorates due to environmental influences and limited coverage area
Solution Approach 1:
The patent transitions from conventional point-based sensor detection to wide-area image-based detection by capturing and analyzing flame images across large spatial areas, enabling simultaneous coverage of extensive regions while maintaining detection accuracy through image processing techniques
Solution Approach 2:
The patent uses image copying and processing techniques to create multiple representations of flame patterns from captured images, allowing analysis of flame characteristics across different regions and improving both coverage area and detection accuracy through comparative analysis
2Area of stationary object
If image processing technology is used for wide-area fire detection, then detection coverage area is improved, but processing time and cost increase significantly
Solution Approach 1:
The patent performs preliminary actions by pre-processing and organizing flame images during the learning phase, creating optimized neural network models that can rapidly detect fires in real-time without requiring extensive processing of raw images during actual detection operations
Solution Approach 2:
The patent replaces complex mechanical image processing systems with a neural network-based intelligent detection system that automatically learns flame patterns and performs rapid detection, significantly reducing processing time while maintaining wide-area coverage capability
3Productivity
If machine learning is applied to fire detection, then detection speed is improved, but data quality and quantity requirements increase, making data acquisition more difficult and costly
Solution Approach 1:
The patent uses image copying and synthesis techniques to generate additional training data by creating variations of existing flame images through transformations such as rotation, scaling, and color adjustments, thereby increasing data quantity without requiring proportional increases in real-world data collection
Solution Approach 2:
The patent performs preliminary data preparation and augmentation during the learning phase, organizing and preprocessing flame images in advance to create comprehensive training datasets that enable fast and accurate detection without requiring extensive data collection during deployment
Data Source
AI summary
Disclosed herein is a learning method of a neural network model for flame determination. The learning method of a neural network includes generating a learning image including a fake image generated by combining a real fire image and an arbitrary flame image with a background image; inputting the learning image to a first neural network model and outputting a determination result for whether a flame is present; and updating a weight in a layer extracting features of the learning image from the first neural network model using the determination result. According to the present invention, data of various fire situations may be secured, a performance of the neural network model that determines an occurrence of the fire through the secured data may be increased, and a quality of data for learning may be increased to allow the neural network model itself to predict various situations of fires.