Hierarchical Patch Analysis for Autonomous Vehicle Path Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing autonomous driving systems face challenges in efficiently processing complex road conditions and navigating around objects, requiring significant computational power and often bulky equipment to identify a clear path for vehicle operation.

Innovation Solution

A method involving patch-based analysis of images from cameras, where the image is divided into patches to determine clear path confidence likelihood, allowing for efficient identification of clear paths without the need for individual object classification, using a combination of camera and radar imaging systems to assess road conditions and obstacles.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If comprehensive object classification and separation is performed to identify clear paths, then navigation accuracy is improved, but computational processing time increases significantly

Engineering Contradiction:

Improveclear path identification accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The image is divided into multiple patches that are processed independently and in parallel. Each patch is analyzed separately to determine clear path confidence, allowing simultaneous processing of multiple regions without requiring sequential object classification throughout the entire image.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The method extracts only the essential clear path confidence information from each patch without performing complete object classification. By taking out only the necessary navigation-relevant data (clear path likelihood) rather than identifying every object type, processing is accelerated while maintaining navigation accuracy.

Inventive Principle:

Principle #2Taking out (Extraction)

2Reliability

If detailed object recognition and classification is implemented, then navigation reliability is improved, but device complexity and processing power requirements increase

Engineering Contradiction:

Improvenavigation reliabilityVSAvoidprocessing equipment complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Different processing approaches are applied to different regions of the image based on local clear path confidence. Patches with high clear path confidence require minimal processing, while only patches with uncertain clear path status undergo more detailed analysis, optimizing resource allocation and reducing overall system complexity.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system performs partial object classification only where necessary for navigation decisions. Instead of classifying all objects throughout the entire scene, the method applies classification only to regions that could potentially affect the clear path, reducing computational burden while maintaining navigation reliability.

Inventive Principle:

Principle #16Partial or excessive action

3Measurement precision

If complete scene analysis with full object identification is performed, then clear path detection accuracy is improved, but energy consumption increases

Engineering Contradiction:

Improveclear path detection accuracyVSAvoidcomputational energy consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The clear path detection is performed periodically at key decision points rather than continuously analyzing every pixel and object. The patch-based method samples the scene at strategically chosen intervals and locations, reducing computational energy while maintaining sufficient detection accuracy for safe navigation.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The system performs partial analysis of the scene by focusing computational energy only on patches that contain potential navigation concerns. By applying analysis selectively rather than uniformly across the entire scene, energy consumption is reduced while still achieving accurate clear path detection where it matters most.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS8421859B2Clear path detection using a hierachical approach

Publication Date: 2013.04.16 GM GLOBAL TECHNOLOGY OPERATIONS LLC

AI summary

A method for detecting a clear path of travel for a vehicle using an image generated by a camera includes monitoring the image, and utilizing a first patch-based analysis of the image including dividing the image into a plurality of patches, and analyzing each patch to determine a first clear path confidence likelihood for each of the patches of the first patch-based analysis. Each of the patches with the first clear path confidence likelihood greater than a first threshold confidence likelihood is defined as a clear path patch. A second analysis is utilized upon a portion of the patches with the first clear path confidence not greater than the first threshold confidence likelihood to determine a second clear path confidence likelihood for each of the patches of the second analysis. Each of the patches of the second analysis with the second clear path confidence likelihood greater than a second threshold confidence likelihood is defined as a clear path patch. The clear path of travel for the vehicle is identified based upon the clear path patches.