Radar Point-Cloud Denoising for Accurate Obstacle Distance Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing distance measurement methods using radar signals are prone to inaccuracies due to interference from surrounding obstacles, leading to low measurement accuracy.
Innovation Solution
A method involving the reception of multiple echo signals, creation of a point cloud dataset with distance, rate, and signal-to-noise ratio values, denoising based on these values, and clustering to identify and measure the distance to a target obstacle, while filtering out noise and interference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If radar signals are used for distance measurement in existing frequency bands, then distance measurement capability is achieved, but measurement accuracy deteriorates due to interference from surrounding obstacles
Solution Approach 1:
The patent segments the detected targets into multiple categories based on their motion characteristics. By dividing the point cloud data into different clusters representing different target types (e.g., stationary obstacles, moving vehicles, pedestrians), the system can selectively process and measure distances to specific target categories, thereby improving measurement accuracy by excluding interference from irrelevant obstacles
Solution Approach 2:
The patent applies dynamic filtering based on motion characteristics of targets. By analyzing the velocity and acceleration patterns of detected objects, the system dynamically adjusts which targets are considered valid measurement objects versus interference. This dynamic approach allows the measurement system to adapt to changing environmental conditions and maintain high accuracy
2Measurement precision
If denoising and clustering operations are performed on point cloud data, then measurement accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent extracts only the essential features from the raw point cloud data that are necessary for accurate distance measurement. By identifying and extracting key parameters such as range, velocity, and acceleration, while discarding redundant information, the system achieves high measurement accuracy without requiring overly complex processing algorithms
Solution Approach 2:
The patent creates simplified representations (copies) of the complex point cloud data through clustering and classification. Instead of processing every individual point, the system generates cluster centroids and representative target models that capture the essential measurement information, significantly reducing computational complexity while maintaining measurement precision
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
Improves measurement accuracy by reducing interference from obstacles and environmental noise, enhancing the precision of distance measurements.
Implementation Method 1
A frequency modulated continuous wave (frequency modulated continuous wave, FMCW) modulation mode may be used to achieve centimeter-level distance measurement precision
Implementation Method 2
receiving a plurality of first echo signals generated within a detection range by a plurality of first radar signals transmitted
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A distance measurement method and apparatus (100) are provided. The method includes: receiving a plurality of first echo signals generated within a detection range by a plurality of first radar signals transmitted in a first time segment (S210); determining a first point cloud dataset based on the plurality of first echo signals (S220); performing denoising on the first point cloud dataset based on a signal-to-noise ratio value and a rate value that are in each piece of first point cloud data included in the first point cloud dataset, to obtain a target dataset (S230), where noise and rates at obstacle points in different motion statuses are different; clustering, based on a distance value in each piece of first point cloud data included in the target dataset, first point cloud data included in the target dataset, to obtain at least one classification, where the at least one classification corresponds to at least one obstacle (S240); and determining a distance between an obstacle corresponding to each classification and a transmitting origin based on a distance value in each piece of first point cloud data included in each of the at least one classification (S250). The distance measurement method and apparatus (100) can improve accuracy of distance measurement.