CNN Kernels Matched to Reflection-Ghost Offset for Radar Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current radar reflection detection techniques for convolutional neural networks face challenges in optimizing object detection in complex scenes, particularly due to the reflection-ghost offset which affects the accuracy of radar signal processing.

Innovation Solution

The method involves obtaining reflective radar signals from a monitored scene using an array of multiple antennas, determining a reflective-intensity (RI) spectrum, and applying a trained convolutional neural network (CNN) to filter the RI spectrum using kernels that incorporate the reflection-ghost offset, thereby enhancing object detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional radar signal processing is used, then the processing is simpler, but the object detection accuracy deteriorates due to reflection-ghost offset artifacts

Engineering Contradiction:

Improveobject detection accuracyVSAvoidsignal processing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The CNN filter kernels are pre-trained offline to incorporate the reflection-ghost offset characteristics of the specific antenna array configuration. This preliminary action embeds the compensation logic into the filter weights before runtime, allowing the processing system to achieve high accuracy without complex real-time calculations

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional signal processing methods (which rely on mathematical models and manual parameter tuning) with a data-driven CNN approach. The mechanical/mathematical processing is substituted with neural network-based pattern recognition that automatically learns to distinguish main lobes from reflection ghosts

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If traditional filtering methods are applied, then the processing is faster, but the discrimination between main lobes and side lobes deteriorates

Engineering Contradiction:

Improvelobe discrimination accuracyVSAvoidprocessing speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The CNN filters are pre-trained offline to learn the specific patterns of main lobes and reflection ghosts for the given antenna configuration. This preliminary training phase allows the system to achieve high discrimination accuracy during runtime without complex real-time computations

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates a virtual model of the reflection-ghost offset characteristics through the CNN filter kernels. This virtual model captures the essential patterns of main lobes and side lobes, allowing the system to distinguish between them through pattern matching rather than complex physical analysis

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12339396B2Radar reflection detection with convolutional neural network kernels matched to reflection-ghost offset

Publication Date: 2025.06.24 GM GLOBAL TECHNOLOGY OPERATIONS LLC

AI summary

A method that includes obtaining reflective radar signals regarding a scene monitored by a radar sensor system having an antenna array that is characterized by effecting a reflection-ghost offset in one or more domains, determining a reflective-intensity (RI) spectrum in three domains based on the reflective radar signals, producing a filtered RI spectrum by applying a trained convolutional neural network (CNN) to the RI spectrum by, at least in part, filtering the RI spectrum using one or more CNN kernels that incorporate the reflection-ghost offset; and detecting objects in the monitored scene based, at least in part, on the filtered RI spectrum.