Neural Network Road Projection Recognition for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional information projection techniques on road surfaces, used to assist driving, are ineffective in autonomous driving systems as they rely on driver recognition and cannot be utilized when a vehicle operates without a driver, thereby failing to enhance driving safety.
Innovation Solution
An artificial neural network-based system that includes an object detecting device using a Convolution Neural Network (CNN) to identify objects, a projection information classifying device to distinguish projection information from the road surface, and a controller to recognize and respond to projection information in a Region Of Interest (ROI) corresponding to the vehicle's driving direction, allowing for safe lane changes and speed adjustments based on projected information from neighboring vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If driver recognition of projected marks is used, then driving safety is improved for manual driving, but the system becomes unusable for autonomous driving mode
Solution Approach 1:
The patent replaces the mechanical system of driver visual recognition with an automated image processing and neural network recognition system. The camera captures projected marks, the preprocessing unit enhances image quality, and the neural network automatically classifies the marks, eliminating the need for driver participation while maintaining safety information delivery.
Solution Approach 2:
The system enables the autonomous vehicle to independently recognize and interpret projection information without driver intervention. The automated recognition system processes images, identifies projected marks, and provides safety information to the autonomous driving controller, allowing the vehicle to serve itself in information acquisition.
2Adaptability or versatility
If automated recognition system is implemented, then applicability to autonomous driving is improved, but system complexity increases
Solution Approach 1:
The recognition system is divided into distinct functional modules: image capture unit, preprocessing unit, neural network recognition unit, and output unit. Each module performs a specific task, making the overall complex system manageable through functional segmentation and independent optimization of each component.
Solution Approach 2:
The neural network recognition system is designed to identify multiple types of projection information (turn signals, lane changes, braking warnings) using a unified architecture. This multi-functional approach avoids the need for separate specialized systems for each type of projection mark, reducing overall system complexity.
Data Source
AI summary
An artificial neural network-based projection information recognition apparatus for a vehicle is capable of learning information (projection information) projected on a road surface by a neighboring vehicle based on an artificial neural network and also recognizing information projected on a region of interest (ROI) determined based on a driving direction of the vehicle. The apparatus includes: an object detecting device to detect an object in an image based on a first Convolution Neural Network (CNN), a projection information classifying device to classify projection information located on a road surface among objects detected by the object detecting device, and a controller that recognizes the projection information located in a Region Of Interest (ROI).