Skip to content
DSML.KZ
Publications

Article · Вестник КазАТК · December 3, 2025

RESEARCH OF A MOBILE AGRICULTURAL ROBOT ON THE BASIS OF LOCAL ARTIFICIAL INTELLIGENCE FOR REAL-TIME DETECTION OF CLUBERRY DISEASES IN THE FIELD

Raikhan Amanova, Baurzhan Belgibaev, Yersaiyn Mailybayev, Gulshat A. Amirkhanova, Gulnur Tyulepberdinova

Baurzhan Belgibaev
Open via DOIPDF

Abstract

Strawberry diseases can severely reduce both yield and fruit quality, and conventional visual scouting by agronomists is difficult to scale in large or densely planted production systems. This work introduces a method for automatic detection of strawberry leaf diseases based on the lightweight YOLOv10n object detector, tailored for deployment on edge devices and mobile agro-robots. A dedicated dataset was assembled for this purpose, consisting of 3,863 images and 9,386 annotated instances from seven disease classes, collected in open-field and greenhouse environments under natural, highly variable lighting with frequent leaf occlusions. The YOLOv10n model was trained using moderate data augmentation to improve robustness under class imbalance, while a lightweight transformer-based MobileViT-S model and a classical «Random Forest» classifier were employed as baseline approaches for comparison in the classification setting. Experimental evaluation demonstrates that YOLOv10n attains an mAP@0.5 of 0.96 and an F1-score of 0.93, with inference speeds suitable for near real-time operation on embedded hardware without cloud support. These results indicate that YOLOv10n-based detectors can be effectively integrated into precision agriculture workflows for targeted spraying, localized irrigation, and early warning of strawberry disease development.

Smart Agriculture and AIAdvanced Neural Network ApplicationsIndustrial Vision Systems and Defect DetectionYOLOv10nlightweight object detectionedge deployment

More by Baurzhan Belgibaev