Rail Segmentation Obstacle Detection for Switch-Aware Train Vision

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing electro-optical systems face challenges in real-time imaging and processing of railway scenes to accurately detect rails and obstacles under diverse weather conditions and extended ranges, particularly in scenarios involving switches and varying track paths, which are crucial for safe train operation.

Innovation Solution

A method and system utilizing a rails and switches states detection neural network (RSSD NN) for dynamic segmentation of image frames, combined with an objects and obstacles detection and tracking neural network (OODT NN), to determine the current and impending railway path, detect obstacles, and generate alerts, employing visual and thermal infrared sensors for enhanced accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If electro-optical sensors are used to survey and monitor railway scenes in real time, then obstacle detection capability is improved, but system complexity increases

Engineering Contradiction:
Improveobstacle detection capabilityVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the railway scene analysis into multiple specialized neural networks: RSSD NN for rail and switch detection, and OODT NN for obstacle detection and tracking. Each network focuses on specific detection tasks, improving overall measurement precision while managing system complexity through functional decomposition

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from traditional 2D image analysis to 4D spatiotemporal analysis by incorporating historical segmentation masks and temporal sequences. This dimensional expansion enables real-time obstacle detection with improved accuracy by analyzing motion patterns across multiple time frames

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Reliability

If image processing is applied to detect rails and obstacles in diverse weather conditions, then detection reliability is improved, but processing time increases

Engineering Contradiction:
Improvedetection reliabilityVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The RSSD NN performs preliminary action by detecting and segmenting rails and switches before obstacle detection begins. This pre-processing establishes the railway path and switch states in advance, enabling faster and more reliable obstacle detection in subsequent frames across diverse weather conditions

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system maintains continuous detection reliability by processing every image frame through the neural networks without interruption. The continuous analysis of rail, switch, and obstacle elements ensures consistent detection performance across varying weather conditions while maintaining real-time processing throughput

Inventive Principle:
Principle #20Continuity of useful action

3Measurement precision

If neural networks are used for real-time segmentation and detection, then measurement precision is improved, but computational requirements increase

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidcomputational requirements
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The computational workload is segmented across two specialized neural networks: RSSD NN handles rail and switch segmentation, while OODT NN handles obstacle detection. This segmentation of computational tasks improves segmentation accuracy for each specific function while distributing computational requirements across multiple optimized processes

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The RSSD NN performs preliminary segmentation of rails and switches before the OODT NN begins obstacle detection. This preliminary action reduces the computational burden on the obstacle detection network by pre-establishing the railway context, thereby improving overall segmentation accuracy while managing total computational requirements

Inventive Principle:
Principle #10Preliminary action

4Reliability

If the system detects and classifies potential obstacles in real time, then safety response capability is improved, but device complexity increases

Engineering Contradiction:
Improvesafety response capabilityVSAvoiddevice complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The safety response system is segmented into specialized functional modules: RSSD NN for rail/switch detection, OODT NN for obstacle detection and tracking, and a classification system for obstacle categorization. This segmentation improves safety response capability by dedicating specific computational resources to each safety-critical function while managing overall device complexity

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The neural network system performs multiple safety functions simultaneously: detecting rails, identifying switches, determining switch states, tracking obstacles, and classifying obstacle types. This multi-functionality improves comprehensive safety response capability while consolidating detection tasks into a unified system architecture

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3820760B1Method and system for railway obstacle detection based on rail segmentation
Publication Date: 2025.11.26 RAIL VISION LTD
  • EP3820760B1 patent drawingFigure 1A
  • EP3820760B1 patent drawingFigure 1B
  • EP3820760B1 patent drawingFigure 2

AI summary

Systems and methods for rails and obstacles detection based on forward-looking electrooptical imaging, novel system architecture and novel scene analysis and image processing are disclosed. The processing solution utilizes a deep learning semantic scene segmentation approach based on a rails and switches states detection neural network that determines the railway path of the train in the forward- looking imagery, and an objects and obstacles detection and tracking neural network that analyzes the vicinity of the determined railway path and detects impending obstacles.