LiDAR Object Detection Fusion for Unknown Objects and Direction Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing 3D object detection techniques for LiDAR systems face challenges in accurately identifying object types and directions due to limitations in signal processing-based methods, while deep learning-based methods struggle with recognizing objects not in their training dataset and those with varied shapes and sizes.

Innovation Solution

An object detection apparatus and method that fuse results from signal processing-based and deep learning-based object detection techniques, using a processor to determine overlap ratios and associations between detected objects, thereby improving detection performance and addressing misrecognition and non-recognition issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If signal processing-based object detection is used, then ease of operation and reliability are improved, but measurement precision and ability to identify object type and direction deteriorate

Engineering Contradiction:

Improvedetection stabilityVSAvoidobject type and direction identification accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The patent combines signal processing-based object detection results with deep learning-based detection results through a fusion algorithm. The signal processing provides stable detection results, while deep learning supplements object type and direction information, resolving the contradiction between reliability and measurement precision.

Inventive Principle:

Principle #5Merging (Combining)

2Measurement precision

If deep learning-based object detection is used, then measurement precision for object size and posture is improved, but adaptability to objects not in training dataset deteriorates

Engineering Contradiction:

Improveobject size and posture prediction accuracyVSAvoidrecognition of objects not in training dataset

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent uses signal processing-based detection results as an intermediary to bridge the gap for objects not in the deep learning training dataset. The signal processing method can detect any object regardless of training data, and its results are fused with deep learning results to provide both precision and adaptability.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If deep learning-based object detection is used, then classification accuracy is improved, but device complexity increases

Engineering Contradiction:

Improveclassification accuracyVSAvoiddetection network complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies deep learning partially - only for objects where it provides significant classification accuracy improvement. The fusion algorithm selectively combines deep learning results with signal processing results, using deep learning's classification strength without fully relying on its complexity for all detection scenarios.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250191352A1Object detection apparatus and method thereof

Publication Date: 2025.06.12 HYUNDAI MOTOR CO LTD

AI summary

In an object detection apparatus and a method therefor, the object detection apparatus may include: a processor configured to detect an object based on deep learning using data obtained from LiDAR and to detect the object based on signal processing; and a storage operatively connected to the processor and configured to store algorithms and data driven by the processor, wherein the processor is configured to output a final object detection result by fusing the deep learning-based object detection result and the signal processing-based object detection result.