Railcar Wheelroom Image Masking Using Catenary Gantry Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video recording systems for railway vehicles face challenges in anonymizing captured images due to ethical and regulatory concerns regarding privacy and personal data protection, with existing solutions failing to conform to railway right-of-way geometries and often obscuring valuable data.
Innovation Solution
An anonymization method that detects catenary pole gantries in the railway right-of-way to define bounding boxes, traces a masking polygon, and applies masking to the image, ensuring all or almost all railway right-of-way data is retained while anonymizing private areas and individuals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If opaque frames are placed on either side of the image to retain only the central portion, then privacy protection is improved, but railway right-of-way data is lost
Solution Approach 1:
The patent applies different treatments to different regions of the image based on their content. Instead of uniformly obscuring fixed side portions, the system dynamically identifies and selectively masks only the specific areas containing private information (people, vehicles, buildings) while preserving railway infrastructure data. This localized approach ensures privacy protection where needed without sacrificing valuable railway data.
Solution Approach 2:
The anonymization masks are dynamic rather than static. The system continuously processes video frames, detects private information in real-time, and adjusts the masking regions accordingly. This allows the anonymization zones to move and change shape based on the position of detected objects, ensuring that private areas are consistently obscured while railway infrastructure remains visible throughout the video sequence.
2Device complexity
If fixed opaque frames are used for anonymization, then processing simplicity is improved, but adaptability to different railway configurations is worsened
Solution Approach 1:
The system employs dynamic masking regions that automatically adapt to different railway configurations. By detecting catenary poles and gantries in each video frame and calculating their positions, the system generates anonymization masks that conform to the actual geometry of the railway right-of-way in each scene. This enables the system to handle various track layouts, pole positions, and gantry configurations without requiring manual reconfiguration.
Solution Approach 2:
The system performs self-adjustment by automatically detecting railway infrastructure elements and generating appropriate anonymization masks without external intervention. The algorithm independently identifies catenary poles, determines the railway right-of-way boundaries, and creates masking polygons that adapt to each scene's specific geometry, making the system universally applicable to different railway configurations.
3Productivity
If real-time anonymization is implemented, then operational efficiency is improved, but processing resource requirements are worsened
Solution Approach 1:
The system segments the image processing task by focusing computational resources only on detecting and masking specific elements (catenary poles, gantries, and associated private information) rather than processing the entire image uniformly. This targeted approach reduces the overall computational burden while maintaining real-time performance.
Solution Approach 2:
The system extracts only the essential elements needed for anonymization (catenary pole positions and gantry geometries) from each video frame, processes this extracted information to generate masks, and applies them. By working with extracted key features rather than the complete high-resolution image data throughout the pipeline, the system reduces processing resource requirements while achieving real-time anonymization.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention relates to a method for anonymizing images of a video stream captured from a railway vehicle, the method comprising a step of receiving at least one image of the video stream, each image representing a scene of a railway right-of-way crossed by the railway vehicle, a step of detecting, on each received image, at least one catenary pole gantry arranged in the railway right-of-way, a step of creating for each received image and from each catenary pole gantry detected on the received image, a bounding box defining a polygonal area globally surrounded by the catenary pole gantry, a step of tracing a masking polygon (70) for each received image, connecting at least one point of each bounding box and defining an anonymization zone on the image, a step of applying the masking to the received image from the masking polygon (70) associated with said image.