Traffic Light ROI Mapping for Low-Latency Autonomous Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous vehicles face increased computational resource usage and detection latencies when using on-board computing for traffic light detection, especially in complex road topologies, due to the need for high-resolution image processing.

Innovation Solution

An edge computing system that processes a sequence of images from one autonomous vehicle to generate traffic light region of interest (ROI) parameters, including visual feature templates and ROI scaling ratios, which are then transmitted to another vehicle to enable real-time traffic light detection, reducing the computational load on individual vehicles.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If on-board computing resources are used to process high-resolution images for traffic light detection, then detection accuracy is improved, but computational resource usage increases and detection latency increases

Engineering Contradiction:

Improvetraffic light detection accuracyVSAvoidcomputational resource usage

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent introduces an edge computing system as an intermediary between the autonomous vehicle and the cloud. This edge server pre-processes images from multiple vehicles, extracts traffic light information, and generates candidate regions of interest. The vehicle's onboard system then only needs to verify these pre-processed results rather than processing full high-resolution images from scratch, significantly reducing computational load while maintaining detection accuracy

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The edge computing system performs preliminary processing of images before they reach the autonomous vehicle. By pre-detecting traffic lights, pre-defining regions of interest, and pre-extracting visual features, the system prepares data in advance so that the vehicle's onboard computer only needs to perform final verification and decision-making, reducing both computational resource usage and detection latency

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If on-board computing resources are used to process high-resolution images for traffic light detection, then detection accuracy is improved, but traffic light detection latency increases

Engineering Contradiction:

Improvetraffic light detection accuracyVSAvoiddetection latency

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent merges computational resources across multiple vehicles by using an edge computing system that aggregates images from several vehicles. This distributed approach allows the system to pool computing power and share processing results, reducing the time any single vehicle needs to spend on detection while improving overall system accuracy through multiple observations

Inventive Principle:

Principle #5Merging (Combining)

3Measurement precision

If complex road topologies are compensated for to enable accurate traffic light detection, then detection accuracy is improved, but computational resource strain increases

Engineering Contradiction:

Improvetraffic light detection accuracyVSAvoidcomputational resource strain

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The edge computing system acts as an intermediary that handles the complex computational tasks of compensating for road topologies. It receives images from multiple vehicles, performs sophisticated processing including perspective transformation and geometric correction to account for bumps, curves, and hills, and returns corrected results to individual vehicles. This distributes the computational burden away from each vehicle's onboard system

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system creates and shares visual feature templates and region of interest definitions across multiple vehicles through the edge computing system. Instead of each vehicle independently handling complex topology compensation, they share pre-processed templates and corrections that have already accounted for road geometry, reducing individual computational strain while maintaining accuracy

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12131555B2Systems and methods for providing traffic light region of interest parameters to an autonomous vehicle

Publication Date: 2024.10.29 GM GLOBAL TECHNOLOGY OPERATIONS LLC

AI summary

A sequence of images and a vehicle location associated with each of the images is received at a traffic light ROI management system. At least one traffic light is detected in each image. A ECS traffic light ROI is defined for each image. The ECS traffic light ROI encloses the detected traffic lights. A visual feature template is generated for each image. The visual feature template is based on the ECS traffic light ROI for the image. Each visual feature template is mapped to the vehicle location associated with the image to a HD map. The HD map is transmitted to an autonomous vehicle to enable the autonomous vehicle to identify a real-time traffic light ROI in a real-time image based on a match between a first visual feature template and real-time visual features of the real-time image at the vehicle location associated with the first visual feature template.