Vehicle Recognition Fusion for Area-Adaptive Target Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recognition technologies face challenges in securing accurate recognition of target movable bodies in vehicles due to varying recognition performance requirements based on the vehicle's traveling area, leading to potential inaccuracies in identifying pedestrians and other road users.
Innovation Solution
A processing system that acquires probability distributions of target movable bodies based on distance from the vehicle, adjusts recognition rates for multiple models using fusion rates correlated with recognition scores, and optimizes fusion rates to reflect changes in recognition performance based on the vehicle's traveling area, ensuring accurate recognition through data fusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple recognition models are used to improve recognition accuracy, then recognition reliability improves, but device complexity increases
Solution Approach 1:
The patent combines multiple recognition models (dictionary-based model and deep learning model) into a unified recognition system that processes sensor data together. The fusion processor integrates results from both models to produce a final recognition output, achieving improved reliability through model combination while managing complexity through structured integration.
Solution Approach 2:
The recognition system is designed to handle multiple types of target movable bodies (pedestrians, cyclists, animals, etc.) using a universal framework that accommodates different recognition models. The system can selectively apply appropriate models based on the detection scenario, making the system versatile without requiring separate dedicated systems for each target type.
2Measurement precision
If recognition models are optimized for specific traveling areas, then recognition precision improves, but adaptability decreases
Solution Approach 1:
The system dynamically adjusts the fusion rate between different recognition models based on the detected traveling area. The fusion rate determination unit changes the weighting of each model's output according to the current environment (e.g., urban, rural, highway), allowing the system to optimize precision for each area while maintaining adaptability across different scenarios through real-time parameter adjustment.
Solution Approach 2:
The patent changes the fusion rate parameter according to the traveling area type. By adjusting this key parameter, the system optimizes recognition precision for specific environments (e.g., higher weight on dictionary-based model for structured urban areas, higher weight on deep learning model for complex rural areas) while maintaining the ability to adapt to any environment by simply changing the parameter value.
Data Source
AI summary
A processing system including a processor is configured to perform recognition-related processing related to recognition of a target movable body for a host movable body. The processor is configured to execute acquiring a recognition rate of recognizing a target movable body for each of a plurality of recognition models, and fusing recognition data by each of the plurality of recognition models according to a fusion rate based on a recognition score.