Related Experiment Video
Updated: Sep 30, 2025
08:27
Author Spotlight: Efficient Image Recognition Using Directional Gradient Histogram Technique and Support Vector Machines
Published on: January 5, 2024
1.3K
Detection of Lower Body for AGV Based on SSD Algorithm with ResNet.
Xinbiao Gao1,2,3,4, Junhua Xu1,2, Chuan Luo1,2,5
1School of Mechanical Engineering, Shandong University, Jinan 250061, China.
Sensors (Basel, Switzerland)
|March 10, 2022
Summary
This study introduces R-SSD, an improved human lower body detection algorithm for autonomous vehicles. R-SSD enhances detection accuracy by 7% mAP, paving the way for reliable vehicle tracking.
Area of Science:
- Computer Vision
- Machine Learning
- Robotics
Background:
- Automatic vehicle tracking and relocation require accurate detection of human lower bodies.
- Existing detection models like SSD and ResNet have limitations in feature extraction and detection range.
Purpose of the Study:
- To propose an improved detection algorithm, R-SSD, for enhanced human lower body detection.
- To increase the detection range and accuracy for autonomous vehicle applications.
Main Methods:
- Utilized ResNet50 for improved feature extraction, replacing VGG16 in the SSD framework.
- Increased model input resolution to 448 × 448 to expand the detection range.
- Employed six feature maps for detection, with aspect ratio clustering of the dataset across these maps.
Main Results:
- The R-SSD model achieved a detection accuracy of 85.1% mean Average Precision (mAP).
- Demonstrated a 7% mAP improvement compared to the original SSD algorithm.
- Attained over 99% detection confidence in practical applications.
Conclusions:
- The R-SSD algorithm significantly improves human lower body detection accuracy.
- Provides a robust foundation for the tracking and relocation of autonomous vehicles.
- Highlights the effectiveness of integrating ResNet50 and increased resolution for enhanced detection performance.
