Deep Learning Security Check Image Sample Generation via Real-Shot Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating training samples for deep learning in security checks are inefficient, relying heavily on manual data collection and identification, which is labor-intensive, costly, and prone to errors, especially for difficult case samples.
Innovation Solution
An image sample generating method and system that uses a simple algorithm to fuse real-shot security check images with labeled target images, normalizing and processing pixel gray values to create new training samples that can adapt to different scenarios, reducing the need for extensive manual data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If manual data collection and identification methods are used to generate training samples, then sample data can be obtained, but the process is labor-intensive, costly, and time-consuming
Solution Approach 1:
The patent uses real-shot security check images as templates to generate synthetic training samples through image processing and fusion techniques. Instead of manually collecting and identifying real contraband images, the system creates copies and variations of real images by fusing them with simulated images, thereby generating large quantities of training data automatically without manual intervention
Solution Approach 2:
The system performs automatic image fusion and sample generation without requiring manual data collection and identification. The algorithm automatically processes real-shot images, fuses them with simulated images, and generates training samples independently, eliminating the need for human labor in data preparation
2Measurement precision
If more real-shot images are collected manually to improve model training, then detection accuracy can be improved, but labor cost and identifying efficiency problems persist
Solution Approach 1:
The patent introduces an image fusion algorithm as an intermediary between real-shot images and training samples. Instead of directly using manually collected real images, the system uses the fusion algorithm to combine real images with simulated images, creating enhanced training samples that maintain the benefits of real data while eliminating the need for extensive manual collection
Solution Approach 2:
The system changes the parameters of training samples by fusing real images with simulated images at different ratios and applying various image processing techniques. This creates diverse training samples with varied characteristics, improving model generalization without requiring proportional increases in manual data collection
3Quantity of substance
If existing sample generation methods are used, then some training data can be produced, but the algorithms are complex and application flexibility is limited
Solution Approach 1:
The patent divides the sample generation process into distinct segments: obtaining real-shot images, obtaining simulated images, normalizing images, fusing images at different ratios, and generating final training samples. This modular approach simplifies the overall algorithm by breaking it into manageable steps that can be independently optimized and applied flexibly to different scenarios
4Reliability
If deep learning models are trained with more data, then detection performance improves, but the cost of collecting and identifying data increases
Solution Approach 1:
The system creates multiple copies and variations of training samples through image fusion, generating large datasets without proportional increases in collection costs. By fusing real images with simulated images, the system automatically generates diverse training data that maintains realism while reducing manual intervention requirements
Solution Approach 2:
The image fusion algorithm serves multiple functions: it generates training samples, enhances existing samples, creates difficult case samples, and adapts to different detection scenarios. This multi-functional approach eliminates the need for separate data collection processes for different purposes, reducing overall costs while improving detection performance
Data Source
AI summary
Provided are a target detection method and an image sample generating method and system for deep learning. The image sample generating method includes performing a scenario composition analysis on an item to be detected in a security check place; obtaining a real-shot security check image of a target scenario having a corresponding composition ratio according to the scenario composition analysis; obtaining a target security check image having a label, where the target security check image is captured by a security check device; processing a pixel gray value of an i-th feature layer in the real-shot security check image and a pixel gray value of an i-th feature layer in the target security check image separately; determining images to be fused; normalizing sizes of the images to be fused; fusing the size-normalized images to be fused to form a new sample; and performing the determining the images to be fused.