Radar Signal Processing for Object Categorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar signal processing devices face challenges in accurately determining whether an object is a braking subject or not, due to reliance on feature quantities derived from signal intensity, high data requirements for dictionary-based systems, and instability in radar cross-section variations, which can lead to delayed responses in dynamic situations.
Innovation Solution
A radar signal processing device that extracts first and second feature quantities from radar measurements, including relative distance, speed, direction, and reflection intensity, and calculates an attribution degree to determine the category of objects, using a database to associate these features with predefined categories, enabling accurate identification even with limited data and dynamic changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dictionary-based height estimation methods are used, then object categorization accuracy is improved, but data storage requirements and processing complexity increase
Solution Approach 1:
The patent extracts only the essential feature quantities (relative distance, relative speed, direction, reflection intensity) needed for object categorization, rather than storing and processing complete dictionary-based height estimation data. This extraction approach maintains categorization accuracy while significantly reducing data storage requirements and processing complexity.
Solution Approach 2:
The patent performs preliminary calculation of feature quantities from radar measurement results before object categorization. By pre-processing the radar data to extract key features, the system avoids the need for complex dictionary-based height estimation during real-time operation, thereby reducing processing complexity while maintaining accuracy.
2Measurement precision
If radar cross-section variation analysis is used for object identification, then object categorization is performed, but response time increases due to data accumulation requirements
Solution Approach 1:
The patent uses a predetermined number of measurement points (a partial amount of data) to calculate feature quantities and perform object categorization, rather than accumulating extensive radar cross-section variation data over time. This partial action approach achieves sufficient identification accuracy while significantly reducing response time for dynamic situations.
3Measurement precision
If multiple feature quantities are calculated and stored in time series, then object categorization accuracy is improved, but processing load increases
Solution Approach 1:
The patent extracts and stores only the essential feature quantities (relative distance, relative speed, direction, reflection intensity) in time series format, rather than processing and storing all possible radar measurement data. This selective extraction maintains object categorization accuracy while significantly reducing the processing load and computational resources required.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The device achieves highly accurate object identification and determination of braking subjects with a short time and small number of measurement points, maintaining accuracy even with missing data, and improves responsiveness in dynamic situations.
Implementation Method 1
a transmission/reception means which transmits radar waves ahead of an automobile, and which receives reflected waves from a target of the radar waves
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A highly accurate object identification is performed. A first feature quantity related to a relative distance and a relative speed to an object, the direction and the reflection intensity of the object, which are extracted by a first feature quantity extraction block (11), is made identical in time series in a data storage processing block (12); a second feature quantity is extracted in a second feature quantity extraction block (13) ; and a category of the object is determined by an object determination block (16) on the basis of an attribution degree to the distribution of the second feature quantity related to a predetermined category calculated by an attribution degree calculation block (14).