ASFL-YOLOX: an adaptive spatial feature fusion and lightweight detection method for insect pests of the Papilionidae family

Lijia Xu; Xiaoshi Shi; Zuoliang Tang; Yong He; Ning Yang; Wei Ma; Chengyu Zheng; Huabao Chen; Taigang Zhou; Peng Huang; Zhijun Wu; Yuchao Wang; Zhiyong Zou; Zhiliang Kang; Jianwu Dai; Yongpeng Zhao

doi:10.3389/fpls.2023.1176300

ASFL-YOLOX: an adaptive spatial feature fusion and lightweight detection method for insect pests of the Papilionidae family

Front Plant Sci. 2023 Jun 14:14:1176300. doi: 10.3389/fpls.2023.1176300. eCollection 2023.

Authors

Lijia Xu¹, Xiaoshi Shi^{1

2}, Zuoliang Tang^{1

2}, Yong He³, Ning Yang⁴, Wei Ma⁵, Chengyu Zheng⁶, Huabao Chen⁷, Taigang Zhou⁸, Peng Huang¹, Zhijun Wu¹, Yuchao Wang¹, Zhiyong Zou¹, Zhiliang Kang¹, Jianwu Dai¹, Yongpeng Zhao¹

Affiliations

¹ College of Mechanical and Electrical Engineering, Sichuan Agricultural University, Ya'an, China.
² College of Resources, Sichuan Agricultural University, Chengdu, China.
³ College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou, China.
⁴ College of Electrical and Information Engineering, Jiangsu University, Zhenjiang, China.
⁵ Institute of Urban Agriculture, Chinese Academy of Agricultural Sciences, Chengdu, China.
⁶ Regulation Department, China Telecom Corporation Limited Sichuan Branch, Chengdu, China.
⁷ College of Agronomy, Sichuan Agricultural University, Chengdu, China.
⁸ Changhong Digital Agriculture Research Institute, Sichuan Changhong Yunsu Information Technology Co., Ltd, Chengdu, China.

Abstract

Introduction: Insect pests from the family Papilionidae (IPPs) are a seasonal threat to citrus orchards, causing damage to young leaves, affecting canopy formation and fruiting. Existing pest detection models used by orchard plant protection equipment lack a balance between inference speed and accuracy.

Methods: To address this issue, we propose an adaptive spatial feature fusion and lightweight detection model for IPPs, called ASFL-YOLOX. Our model includes several optimizations, such as the use of the Tanh-Softplus activation function, integration of the efficient channel attention mechanism, adoption of the adaptive spatial feature fusion module, and implementation of the soft Dlou non-maximum suppression algorithm. We also propose a structured pruning curation technique to eliminate unnecessary connections and network parameters.

Results: Experimental results demonstrate that ASFL-YOLOX outperforms previous models in terms of inference speed and accuracy. Our model shows an increase in inference speed by 29 FPS compared to YOLOv7-x, a higher mAP of approximately 10% than YOLOv7-tiny, and a faster inference frame rate on embedded platforms compared to SSD300 and Faster R-CNN. We compressed the model parameters of ASFL-YOLOX by 88.97%, reducing the number of floating point operations per second from 141.90G to 30.87G while achieving an mAP higher than 95%.

Discussion: Our model can accurately and quickly detect fruit tree pest stress in unstructured orchards and is suitable for transplantation to embedded systems. This can provide technical support for pest identification and localization systems for orchard plant protection equipment.

Keywords: ASFF; GhostNet-ECA; YOLOX; pest detection; plant protection; pruning strategy.

Grants and funding

This study was funded by the Science and Technology Innovation Cultivation Project of Department of Science and Technology of Sichuan province (Grant No. 2021JDRC0091), the Key R & D project of Department of Science and Technology of Sichuan province (Grant No. 2020YFN0025), Natural Science Youth Program of Department of Education of Sichuan Provincial (Grant No. 16ZA0032), the Key R & D project of Department of Science and Technology of Sichuan province (Grant No. 22ZDYF0095), Intelligent Agriculture Engineering Technology Center of Sichuan Provincial.