Subsea Video Object Detection Parallel Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic object detection systems in video processing face challenges in accuracy, robustness, efficiency, and cost-effectiveness, particularly in harsh environments where resource constraints are significant, and require calibration transforms that can lead to positional inaccuracy.
Innovation Solution
A computing device with parallel processing architecture, utilizing two data processors for object detection and 3D reconstruction, where the object detector assigns labels and 2D locations, and the 3D reconstruction generates models, allowing for efficient use of processing power and eliminating the need for calibration transforms by using the same image data for both tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single processor is used for both object detection and 3D reconstruction, then device complexity is reduced, but processing speed and efficiency deteriorate
Solution Approach 1:
The system divides the processing tasks into two separate processors: a first processor dedicated to object detection and labeling, and a second processor dedicated to 3D reconstruction. This segmentation allows each processor to specialize in its specific task, improving overall processing efficiency and speed while maintaining manageable device complexity through functional separation.
2Measurement precision
If calibration transforms are used to correlate multiple sensors, then measurement precision improves, but reliability deteriorates due to transform changes over time
Solution Approach 1:
The system uses the same image data for both object detection and 3D reconstruction tasks, eliminating the need for calibration transforms between separate sensors. By merging the data source and processing both tasks from a common image, the system achieves accurate object location without the reliability issues associated with time-varying calibration transforms.
3Measurement precision
If high bandwidth image data with large spatial resolution and high framerate is processed, then object detection accuracy improves, but use of energy and processing power increases
Solution Approach 1:
The system segments the processing workload across two processors, with the first processor handling object detection on high-bandwidth image data to maintain accuracy, while the second processor handles 3D reconstruction. This segmentation allows the system to process high-quality data without overloading a single processor, managing energy consumption through distributed computation.
4Productivity
If parallel processing is implemented with multiple processors, then productivity improves, but device complexity increases
Solution Approach 1:
The system implements parallel processing by segmenting tasks into two processors with clearly defined responsibilities: object detection/labeling and 3D reconstruction. This functional segmentation simplifies the parallel architecture compared to more complex multi-processor systems, as each processor has a specialized role and the data flow between them is well-defined.
Data Source
AI summary
A computing device for video object detection. Images from a camera are transferred in parallel to a first processor running object detection and a second processor running a 3D reconstruction. The object detection identifies a semantic object of interest and assigns a label to it and outputs the label information to an object mapper. The object mapper assigns the label to a component in the 3D model representing the object. The computing device can form part of a subsea or other harsh environment imaging system.