Identification of leaves of wild Ussurian Pear (Pyrus ussuriensis) based on YOLOv10n-MCS

Niman Li^1,2Xingguang Dong^2*Yongqing Wu^1*Luming Tian²Ying Zhang²Hongliang Huo²Dan Qi²Jiayu Xu²Chao Liu²Zhiyan Chen¹Yulu Mou²

• ¹School of Software, Liaoning Technical University, Huludao, China
• ²Institute of Pomology, Chinese Academy of Agricultural Sciences, Xingcheng, Liaoning Province, China

Wild Ussurian Pear germplasm resource has rich genetic diversity, which is the basis for genetic improvement of pear varieties. Accurately and efficiently identifying wild Ussurian Pear accession is a prerequisite for germplasm conservation and utilization. In order to achieve accurate identification of the accession of wild pear, an improved YOLOv10n-MCS network model based on YOLOv10n was proposed. Specifically, the Mixed Local Channel Attention (MLCA) module was introduced at the neck to enhance the model's feature extraction capability and improve recognition accuracy. Then, the Simplified Spatial Pyramid Pooling -Fast (SimSPPF) module is used instead of the original network pyramid pooling layer to improve the detection efficiency of the model. Furthermore, the C2f SCConv module is designed to replace C2f in backbone, reducing computational redundancy and improving detection performance. The proposed model is validated using a self-made dataset of wild pear leaves images. Experiment results demonstrate that the precision, recall, mAP50, parameters, FLOPs, and model size of YOLOv10n-MCS reached 97.7(95% CI: 97.18 to 98.16)%, 93.5(95% CI: 92.57 to 94.36)%, 98.8(95% CI: 98.57 to 99.03)%, 2.52M, 8.2G, and 5.4MB, respectively. The precision, recall, and mAP50 are significant improved of 2.9%, 2.3%, and 1.5% respectively over the YOLOv10n model (p <0.05). By comparing with other mainstream target detection algorithms, it is verified that the proposed model has advantages in detection precision, model complexity, model size, and other aspects. The YOLOv10n-MCS proposed in this study can quickly and accurately identify wild pear leaves in natural backgrounds. It effectively meets the requirements of accuracy and real-time performance, which helps to achieve automated identification of wild Ussurian Pear accession and provides technical support and reference for the protection, utilization, classification research of wild pear germplasm resource, as well as the identification of other crop varieties.