Object Detection via Region-Specific Candidate Bounding Boxes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In vehicle-to-everything (V2X) applications, existing object detection models face challenges in achieving high accuracy and real-time performance due to the need for detecting small and densely distributed objects, which requires a high detection precision but increases computational load and reduces processing speed.
Innovation Solution
The method involves determining non-overlapping basic candidate bounding boxes and expanded candidate bounding boxes with different dimensions, focusing more on regions of interest such as small object dense areas, and randomly reserving candidate bounding boxes for training to improve detection precision and stability while reducing computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection is performed on the entire driving scene image, then detection coverage is improved, but computational load increases and processing speed decreases
Solution Approach 1:
The driving scene image is divided into multiple regions including a first region (far end with small dense objects) and a second region (near end with large sparse objects). Different detection strategies are applied to each region: dense sampling in the first region and sparse sampling in the second region, thereby reducing overall computational load while maintaining detection coverage.
Solution Approach 2:
Different detection densities are applied to different regions based on their specific characteristics. The first region (far end) uses dense sampling with more candidate bounding boxes per unit area, while the second region (near end) uses sparse sampling with fewer candidate bounding boxes per unit area, optimizing computational resources according to local detection needs.
2Measurement precision
If detection precision for small dense objects is improved, then detection accuracy is improved, but computational load increases
Solution Approach 1:
The image is segmented into regions with different object density characteristics. The first region containing small dense objects is separately handled with increased sampling density, while other regions use standard or reduced sampling, thereby concentrating computational resources where they are most needed for small object detection.
Solution Approach 2:
The sampling density parameter is dynamically adjusted based on region characteristics. In the first region where small dense objects are located, the sampling density is increased to improve detection accuracy, while in other regions the sampling density is reduced to lower computational load, achieving an optimal balance between accuracy and efficiency.
Data Source
AI summary
A method of object detection for driving assistance, including: acquiring a driving scene video, and acquiring a driving scene image of each moment of the driving scene video; determining a plurality of non-overlapping basic candidate bounding boxes on the driving scene image; determining a region of interest of the driving scene image; determining a plurality of expanded candidate bounding boxes with different dimensions near each of the basic candidate bounding boxes in the region of interest; and inputting the basic candidate bounding boxes and the expanded candidate bounding box into an object detection model to detect an object affecting a driving behavior, and obtaining an object detection result of the driving scene image.