Automated Training Image Generation for Surveillance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating training images for machine learning tools in surveillance systems, such as automatic target recognition, rely on manual or semi-automatic processes, which are time-consuming and inefficient, requiring large volumes of labeled images and involving operators in the data processing loop.
Innovation Solution
A method for automatically generating training images by estimating object positions based on valid information at a specific recording time, separating the image into 'good' and 'bad' image tiles using predefined selection criteria, and providing these images without the need for an operator, allowing for efficient and effective training data creation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual or semi-automatic processes are used to generate training images, then the quality and accuracy of training data can be maintained, but the data processing time and time consumption increase significantly
Solution Approach 1:
The system automatically generates training images by having the machine learning model detect objects in original images and use those detections to create labeled training data, eliminating the need for manual operator involvement while maintaining training quality
Solution Approach 2:
The method creates training images by copying and processing existing original images, adding synthetic labels based on model detections rather than manual annotations, thus reproducing training data efficiently without manual intervention
2Measurement precision
If manual processes involving operators are used for training image generation, then accurate object labeling can be achieved, but the productivity and throughput of the system decrease
Solution Approach 1:
The machine learning model performs self-training by automatically generating its own training data through object detection in original images, eliminating the bottleneck of manual operator processing and significantly increasing system throughput while maintaining labeling accuracy
Solution Approach 2:
The system performs preliminary object detection on original images to generate labels before creating training images, preparing the training data in advance through automated processes rather than waiting for manual annotation, thus improving productivity
3Loss of time
If automated methods are used to generate training images, then the data processing time is reduced, but the complexity of the generation process increases
Solution Approach 1:
The machine learning model serves multiple functions: it detects objects in original images, generates labels for training images, and trains itself iteratively, simplifying the overall system architecture despite the automated complexity by using a single multi-functional component
Data Source
AI summary
A method for providing training images including receiving information about objects which is at least valid at a specific recording time. The method includes receiving an image of an area imaged at the specific recording time. The method includes estimating respective positions of the objects at the specific recording time based on the information and selecting estimated positions of respective objects from the estimated respective positions which fulfill a predefined selection criterion. The method includes generating training images by separating the imaged area into first and second pluralities of image tiles. Each of the first plurality of image tiles differs from each of the second plurality of image tiles. Each of the first plurality of image tiles images (depicts) a respective one or several of the selected positions. The method includes providing the training images. Further, a training image product and a device for providing training images are provided.