Radar Object Classification Using Multi-Cycle Spectrum Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for classifying objects around a motor vehicle using radar sensors are not precise enough, leading to potential misidentification of objects, especially in cases where objects have multiple reflections or varying speeds, such as pedestrians moving their arms.

Innovation Solution

A method that utilizes a radar sensor to generate a power spectrum from received radar signals, selects potential objects based on intensity thresholds, and assigns them to specific classes by comparing features such as intensity range, speed, and acceleration, using a rectangular section of the power spectrum and a prediction algorithm like the Kalman filter to enhance accuracy and reduce noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If radar signals are used to detect objects in the surrounding area, then objects can be detected at a distance, but the classification precision is insufficient leading to misidentification of objects with multiple reflections or varying speeds

Engineering Contradiction:
Improveobject classification precisionVSAvoidobject identification reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the detection process into multiple cycles, where each cycle performs detection and classification independently. By dividing the continuous detection into discrete cycles and comparing results across cycles, the system achieves more reliable classification while maintaining distance detection capability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary detection in the first cycle to identify potential objects, then uses this information to guide subsequent detection cycles. The first cycle establishes baseline data that is used to compare against later cycles, improving overall classification precision before final identification.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If multiple spectrum areas are considered for object detection, then detection coverage is improved, but noise and false positives increase

Engineering Contradiction:
Improveobject detection accuracyVSAvoidnoise and false positives
Core Design Contradiction:
Measurement precisionVSObject-generated harmful factors

Solution Approach 1:

The patent merges detection results from multiple spectrum areas across different cycles by requiring consistent identification in at least two cycles. This combining approach maintains comprehensive detection coverage while filtering out noise through temporal consistency validation.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system uses feedback from cycle-to-cycle comparisons to validate detections. Detection results from one cycle inform and are compared against subsequent cycles, allowing the system to distinguish true objects from noise through consistent pattern recognition across multiple measurement cycles.

Inventive Principle:
Principle #23Feedback

3Reliability

If objects are tracked over multiple cycles, then classification reliability is improved, but detection time increases

Engineering Contradiction:
Improveclassification reliabilityVSAvoiddetection time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent requires only a minimum number of consistent detections (at least two cycles) rather than continuous tracking. This partial action approach achieves sufficient reliability for safe operation without the excessive time cost of prolonged tracking, balancing reliability needs with response time requirements.

Inventive Principle:
Principle #16Partial or excessive action

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

This approach allows for more precise classification of objects, reducing noise and improving the reliability of object recognition, enabling better tracking and assignment to specific classes like pedestrians or vehicles, which is crucial for autonomous driving scenarios.

Implementation Method 1

Sending radar signals into the surrounding area from at least one motor vehicle-side radar sensor and receiving radar signals reflected on the object

Methodology Applied
Scientific EffectRadar: Radar

Implementation Method 2

receiving radar signals reflected on the object

Methodology Applied
Scientific EffectReflection: Reflection

Implementation Method 3

The radar sensor calculates a Fourier transform from the received radar signal in the range and Doppler directions, whereby a two-dimensional Fourier transform is obtained

Methodology Applied
Scientific EffectFourier transform:

Implementation Method 4

The radar sensor calculates a Fourier transform from the received radar signal in the range and Doppler directions

Methodology Applied
Scientific EffectDoppler effect: Doppler Effect

Data Source

PatentEP3175256B1Method for classifying an object in an area surrounding a motor vehicle, driver assistance system and motor vehicle
Publication Date: 2020.08.12 VALEO SCHALTER & SENSOREN GMBH
  • EP3175256B1 patent drawingFigure 1
  • EP3175256B1 patent drawingFigure 2~3
  • EP3175256B1 patent drawingFigure 4~5

AI summary

The invention relates to a method for classifying an object (18) in an area (4) surrounding a motor vehicle (1), having the steps: a) emission of radar signals (5) into the surrounding area (4) by at least one radar sensor (3) on the motor vehicle side, and receiving radar signals (6) reflected at the object (18), b) making available the information obtained by means of the radar signals (5, 6) as radar data, c) generating a power spectrum (11) as a function of at least one distance value (13) and/or speed value (14) and/or intensity value (12) contained as information in the radar data, d) comparing at least one spectrum range (15), contained in the power spectrum (11) with an intensity threshold, and selecting the spectrum range (15) as a potential object (16) if the intensity (12) thereof is greater than the intensity threshold value, e) carrying out a further cycle with the steps a) to d), f) detecting the potential object (16) as an actual object (18) as a function of whether the potential object (16) was selected both in a first cycle according to steps a) to d) as well as at least in the second cycle according to step e), g) comparing the at least one feature (Vabs, Aabs, σ2V, σ2Α, y, RC) of the detected object (18) with reference features (Vabs, Aabs, σ2V, σ2Α, y, RC) of a reference object, h) assigning the object (18) to a specific class (23, 24, 25) of multiple different classes (23, 24, 25) as a function of the comparison according to step G)