Predicting Invisible Keypoints for Vehicle Location Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing keypoint extraction methods in vehicle location calculation systems can only predict visible keypoints, limiting the recognition of a target vehicle's shape and location.
Innovation Solution
A vehicle location calculation apparatus and method that predicts invisible keypoints using a model learning part to output invisible keypoint sets based on visible keypoint sets, with a dataset calculation part generating datasets including both visible and invisible keypoints, and a spatial coordinate calculation part determining the target vehicle's location using 2D and 3D camera coordinate values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only visible keypoints are predicted using existing keypoint extraction methods, then the system complexity remains low, but the recognition accuracy of vehicle shape and location is limited
Solution Approach 1:
The patent transitions from 2D visible keypoint detection to 3D invisible keypoint prediction by introducing depth information. The model learns to predict keypoints that are not directly visible in the 2D image but can be inferred through 3D spatial reasoning, thereby enhancing vehicle shape and location recognition accuracy.
Solution Approach 2:
The patent introduces a model learning part as an intermediary component that bridges visible keypoint detection and invisible keypoint prediction. This intermediary learns the relationship between visible and invisible keypoints through training data, enabling accurate prediction of hidden vehicle features without directly observing them.
2Measurement precision
If invisible keypoints are predicted using a model learning part, then the recognition of vehicle shape and location is enhanced, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary training of the model learning part offline using large datasets of visible and invisible keypoints. Once trained, the model can quickly predict invisible keypoints during real-time operation without requiring complex computations, thus reducing processing time while maintaining high accuracy.
Solution Approach 2:
The patent creates a learned mapping model that copies the relationship between visible and invisible keypoints from training data. This copied knowledge allows the system to predict invisible keypoints efficiently during inference without repeating the complex training process, reducing real-time computational burden.
3Measurement precision
If a dataset including both visible and invisible keypoints is generated, then the model learning accuracy is improved, but the data processing complexity increases
Solution Approach 1:
The patent segments the keypoint data into visible and invisible categories, allowing the model to learn their relationships separately. The dataset is structured to clearly distinguish between keypoints that can be directly observed and those that must be inferred, simplifying the learning process while improving accuracy.
Solution Approach 2:
The patent introduces a dataset calculation part as an intermediary that automatically generates and manages the complex dataset of visible and invisible keypoints. This intermediary component handles data pairing, validation, and organization, reducing the manual processing complexity while enabling high-quality model training.
Data Source
AI summary
A vehicle location calculation apparatus includes: a model learning part configured to perform learning to output an invisible keypoint set in a model image in which each vehicle is modeled, based on a visible keypoint set in the model image; and a dataset calculation part configured to generate a dataset including a visible keypoint and an invisible keypoint of a target vehicle, by inputting the visible keypoint of the target vehicle to the model learning part so that the invisible keypoint of the target vehicle is output, the visible keypoint of the target vehicle being detected in an image of the target vehicle while driving.