Industrial Plant Video Mapping Using 3D Models and Camera Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing industrial plant monitoring systems rely heavily on manual annotation and lack automated methods to assign two-dimensional image data components to three-dimensional plant models, making it difficult to determine camera properties such as position, orientation, and zoom factor, especially with multiple video streams from varying cameras.
Innovation Solution
A method that involves receiving a video stream from an industrial plant, assigning two-dimensional image components to plant objects using a three-dimensional model, determining recording properties such as camera position, orientation, and zoom factor, and outputting the video stream along with these properties, all done in an automated manner.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used to assign 2D image components to 3D plant models, then the assignment can be performed, but the process is time-consuming and not scalable for multiple video streams
Solution Approach 1:
The system performs self-annotation by automatically determining camera properties and assigning 2D image components to 3D plant model objects without requiring manual user annotation. The automated process uses the video stream data and 3D model to compute camera position, orientation, and zoom factor, thereby eliminating time-consuming manual work while maintaining assignment accuracy.
2Area of stationary object
If multiple video streams from different cameras are used to monitor the plant, then comprehensive coverage is achieved, but the complexity of managing and evaluating all streams increases significantly
Solution Approach 1:
The system processes multiple video streams from different cameras using a unified automated approach that determines camera properties and assigns image components to 3D model objects for all streams simultaneously. This universal processing method maintains comprehensive monitoring coverage while reducing the operational complexity of managing multiple streams by applying the same automated evaluation logic across all cameras.
3Adaptability or versatility
If camera settings and positions are undocumented, then the system is more flexible in deployment, but the ability to analyze and interpret video data accurately is compromised
Solution Approach 1:
The system automatically determines and recovers camera properties (position, orientation, zoom factor) from the video stream data and 3D plant model, creating a feedback loop that reconstructs the camera configuration information that would otherwise be lost. This allows the system to maintain deployment flexibility with undocumented cameras while recovering the necessary information for accurate video data analysis through automated computation.
Data Source
AI summary
Disclosed herein is a method for monitoring an industrial plant including plant objects, where the following steps are carried out: receiving (S1) a video stream (VS) of the plant generated with the aid of a video camera device; assigning (S2) two-dimensional image components (PT) of the video stream (VS) to plant objects (A′, B′, C′, D′) with the aid of a three-dimensional (3D) model (CAD) of the plant; determining (S3) a recording property of the image data depending on the image components (PT) assigned to one or more plant objects (C′) and the respective object property (XYZ); and outputting (S4) the video stream (VS′) together with the determined object properties (XYZ) and/or recording properties. Further disclosed herein are a computer program product and an apparatus.