Lightweight YOLOv8-PSS Model for Obstacle Detection in Agricultural Vehicles

Sample size: 7344 publication 10 minutes Evidence: high

Author Information

Author(s): Chen Zhijian, Fang Yijun, Yin Jianjun, Lv Shiyu, Sheikh Muhammad Farhan, Liu Lu

Primary Institution: School of Agricultural Engineering, Jiangsu University, Zhenjiang, China

Can a lightweight YOLOv8-PSS model improve obstacle detection for unmanned agricultural vehicles?

The YOLOv8-PSS model achieves high accuracy and efficiency in detecting obstacles for unmanned agricultural vehicles.

The YOLOv8-PSS model achieved a precision of 85.3%, recall of 88.4%, and average accuracy of 90.6%.
It reduced the number of parameters by 55.8% and computational cost by 51.2% compared to the original model.
The model performed well in real-world conditions, demonstrating effective obstacle detection.
Positioning accuracy tests showed average and maximum errors of 2.73% and 4.44% respectively.
The model outperformed other algorithms like Faster R-CNN and YOLOv5 in terms of accuracy and efficiency.

This study created a smart model that helps farming machines see and avoid obstacles, making them safer to use.

The study developed a YOLOv8-PSS model that integrates a depth camera and uses advanced convolution techniques for real-time obstacle detection.

The model struggles with detecting static obstacles and may misidentify objects in complex lighting conditions.

Access the complete publication on the publisher's website