Ground Mark Extraction via Deep Learning Network

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for extracting ground marks in high-precision maps rely heavily on manual editing, which is inefficient, prone to errors, and lacks accuracy due to the sparse distribution of ground marks on road segments.

Innovation Solution

A ground mark determining method utilizing a mark-extraction network model, trained on sample grayscale images of road segment maps, performs convolution and pooling to extract and recognize ground marks, improving efficiency and accuracy by automatically identifying and filtering ground marks based on confidence levels and attribute information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If manual editing method is used to extract ground marks, then the extraction process can be performed with simple tools, but the workload is heavy and efficiency is low

Engineering Contradiction:
Improvesimplicity of extraction processVSAvoidextraction efficiency
Core Design Contradiction:
Ease of manufactureVSProductivity

Solution Approach 1:

The patent replaces the manual mechanical editing process with an automated deep learning-based image processing system. The mark-extraction network model automatically identifies and extracts ground marks from road segment maps, substituting human manual operations with intelligent algorithms that perform convolution, pooling, and feature extraction to detect ground mark patterns efficiently.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If manual traversal of all road segments is performed to find ground marks, then all road segments can be checked, but the process is time-consuming and accuracy is low

Engineering Contradiction:
Improvecompleteness of ground mark detectionVSAvoidextraction time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent uses a trained mark-extraction network model that has learned ground mark patterns from training data. This model acts as a intelligent copy of human detection capability, automatically applying learned patterns to new road segment maps to identify ground marks without requiring manual traversal of each segment, thereby maintaining detection completeness while dramatically reducing time consumption.

Inventive Principle:
Principle #26Copying

3Ease of operation

If manual editing is used to draw boundary boxes and edit attributes, then the extraction method can be straightforward, but errors are prone and accuracy is low

Engineering Contradiction:
Improvesimplicity of editing operationVSAvoidground mark extraction accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent implements self-service automation where the mark-extraction network model autonomously performs ground mark detection, boundary box drawing, and attribute extraction without human intervention. The system automatically identifies ground mark locations, determines their shapes and positions, and extracts relevant attributes, eliminating manual editing errors while maintaining operational simplicity through automated processing.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3792901B1Ground mark extraction method, model training method, device and storage medium
Publication Date: 2023.05.03 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3792901B1 patent drawingFigure 1~2A
  • EP3792901B1 patent drawingFigure 2B~2D
  • EP3792901B1 patent drawingFigure 3~4

AI summary

Disclosed in the present application is a ground mark determination method, comprising: acquiring a point cloud grayscale map, the point cloud grayscale map comprising a road segment map; running a mark extracting network model, acquiring ground mark information from the road segment map, the mark extracting network model being used to extract ground marks included in the road segment map, the ground mark information comprising information concerning each of the ground marks extracted by the mark extracting network model, the ground mark being travel indication information marked on the ground of the road segment; according to the ground mark information, determining a target ground mark from each of the ground marks. In the embodiments of the present application, a mark extracting network model is used to acquire information concerning ground marks in a road segment map, and the ground marks in the road segment map is determined according to the ground mark information, thereby improving the efficiency and accuracy in determining the ground mark in the road segment map.