Attention-Based Target Detection–You Only Look Once: A Detection Model for <i>Locusta migratoria</i> ssp. <i>manilensis</i> in Complex Environments

Locusts have always been among the important hazards affecting crop growth and the grassland ecological environment. Accurate and timely detection of locusts is crucial for effective control of insect development. Aiming at the problem of false detection and missed detection caused by locust occlusi...

Full description

Saved in:
Bibliographic Details
Main Authors: Peng Wang, Jiandong Fang, Xiuling Wang, Yudong Zhao
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Agronomy
Subjects:
Online Access:https://www.mdpi.com/2073-4395/15/6/1381
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Locusts have always been among the important hazards affecting crop growth and the grassland ecological environment. Accurate and timely detection of locusts is crucial for effective control of insect development. Aiming at the problem of false detection and missed detection caused by locust occlusion and background similarity in complex field environments, this paper proposes a lightweight Attention-based Target Detection (ATD) model while constructing the dataset Real-Locust with the theme of <i>Locusta migratoria</i> ssp. <i>manilensis</i>. By introducing to attention mechanism and lightweight design, the model achieves a mean average precision (mAP) of 90.9% on the Real-Locust dataset, and the precision and recall rate are increased by 0.6% and 4.3%, respectively. At the same time, the number of parameters and computational complexity are reduced by 27.4% and 22.9%, showing that this provides an efficient solution for real-time monitoring of locusts.
ISSN:2073-4395