Superimposed Image Generation for Vehicle Detection Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting the location of a moving vehicle using external cameras require a large amount of data for artificial intelligence training, leading to a significant burden in preparing learning data sets.
Innovation Solution
A method that involves acquiring a moving route of a moving object, generating a superimposed image by combining images or three-dimensional data of the moving object and its environment, acquiring label information about the object's location and position, and creating a learning data set with the superimposed image and label information to reduce the burden of data preparation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If artificial intelligence is used in detecting a moving object from captured images, then detection accuracy is improved, but a huge quantity of learning data is required which increases data preparation burden
Solution Approach 1:
The system performs preliminary actions by automatically generating synthetic training data through virtual environment construction and image synthesis before actual AI training begins. This pre-generation of diverse training samples (including various lighting conditions, weather scenarios, and object positions) eliminates the need for manual data collection and preparation, directly resolving the contradiction between achieving high detection accuracy and reducing data preparation burden
Solution Approach 2:
The system creates virtual copies of real-world environments, objects, and scenarios through three-dimensional modeling and rendering. By synthesizing training images from virtual models rather than collecting real physical data, the system generates unlimited diverse training samples without the logistical burden of physical data collection, thereby improving detection accuracy while minimizing data preparation effort
2Measurement precision
If diverse training data is generated to improve AI performance, then model accuracy is improved, but data generation time and computational resources increase
Solution Approach 1:
The system pre-generates comprehensive training datasets covering diverse scenarios (different times of day, weather conditions, object positions) before training begins. This preliminary data preparation stage creates a ready-to-use training corpus that accelerates the actual model training process, achieving high model accuracy without extending total project timeline
Solution Approach 2:
The system implements dynamic parameter adjustment in the virtual environment, allowing automated variation of lighting conditions, weather scenarios, and object positions through programmable parameters rather than manual scene reconstruction. This dynamic generation approach efficiently creates diverse training data while minimizing computational time and resource requirements
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method comprises: acquiring a moving route of a moving object; generating a superimposed image by superimposing an image or three-dimensional data representing the moving object and an image or three-dimensional data representing an environment related to moving including the moving route on each other; acquiring a label information about the location and position of the moving object; and generating a learning data set including the superimposed image and the label information.