Common-Virtual-Axis 3D Camera Real-Time Ranging via YOLOv3-tiny and CSS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for real-time ranging with cameras, particularly in applications like unmanned driving and industrial measurement, face challenges in accurately and efficiently measuring distances using deep neural networks like YOLOv3-tiny in conjunction with common-virtual-axis 3D cameras.
Innovation Solution
A real-time ranging method utilizing a common-virtual-axis 3D camera that processes image data from both near and far images using the YOLOv3-tiny algorithm, involving steps like target recognition, curvature scale space corner recognition, and optical relation calculations to determine distance, with acceleration through field programmable gate arrays (FPGAs) and specific image processing operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If YOLOv3-tiny algorithm is used for target recognition and positioning, then operation speed is improved, but measurement precision may deteriorate
Solution Approach 1:
The patent segments the image processing into distinct modules: YOLOv3-tiny for target detection and positioning, CSS corner recognition for feature extraction, and optical relation calculation for distance computation. This segmentation allows each module to be optimized independently, maintaining fast operation while ensuring precise measurement through specialized processing at each stage.
Solution Approach 2:
The patent introduces CSS corner recognition as an intermediary between YOLOv3-tiny target detection and final distance measurement. This intermediary step extracts precise corner coordinates from the detected targets, serving as a bridge that transforms the fast but less precise YOLO detections into accurate measurement data through geometric relationship analysis.
2Productivity
If real-time ranging is implemented, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent makes the common-virtual-axis 3D camera system multi-functional by integrating YOLOv3-tiny target recognition, CSS corner recognition, and optical relation calculation into a single unified system. This allows one system to perform both fast target detection and precise distance measurement simultaneously, improving productivity without proportionally increasing complexity.
Solution Approach 2:
The patent utilizes parameter changes in the optical relation formula, specifically adjusting for different object distances (L1 and L2) and lens configurations. By dynamically adapting these optical parameters based on the detected target position and image characteristics, the system achieves real-time ranging while managing computational complexity through efficient parameter utilization.
3Measurement precision
If curvature scale space corner recognition is performed, then measurement precision is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary action by first detecting targets and extracting corners using YOLOv3-tiny and CSS algorithms before performing the more computationally intensive optical relation calculations. This preliminary processing identifies and pre-processes the essential features, reducing the time required for subsequent distance computation while maintaining high measurement precision through the CSS corner recognition method.
Data Source
AI summary
The present disclosure discloses a real-time ranging method of a common-virtual-axis three-dimensional (3D) camera based on YOLOv3-tiny, including: acquiring, by the common-virtual-axis 3D camera, a far image and a near image on a same optical axis; processing image data of the near image and image data of the far image by using a YOLOv3-tiny target-recognition neural-network acceleration algorithm; determining, according to the processed image data, two target recognition frames corresponding to a preset recognized object on the far image and the near image; performing curvature scale space (CSS) corner recognition on image data in the target recognition frames; obtaining an average value of distances from far and near corners to a center point of the image data in each of the target recognition frames according to recognized corner coordinates; and substituting the obtained average value into an optical relation of the common-virtual-axis 3D camera, to obtain distance information.