This paper proposes a detection algorithm DDC-YOLO based on improved YOLOv10 to address the issues of occlusion interference and insufficient lighting in tree detection. Firstly, a dynamic convolutional mix block (DCMB) was designed to enhance the multi-scale feature fusion capability through adaptive dynamic convolution, solving the problem of singularity in traditional convolution kernels; Secondly, a dual backbone dynamic feature fusion network was proposed, combining the backbone structures of RT-DETR and YOLOv10, and utilizing the dynamic alignment fusion (DAF) module to adjust feature weights and enhance the model's adaptability to different features; Further introduced pyramid context feature extraction and spatial feature reconstruction techniques to optimize the neck network and achieve deep fusion of multi-level semantic information. The experiment was validated based on the self built dataset TreeImages (containing 7475 images), and the results showed that the mAP50 of DDC-YOLO reached 46.7%, which was 5.0 percentage points higher than the original YOLOv10 model. The parameter size decreased from 2.27 M to 2.26 M (a decrease of 0.44%), and the detection speed (FPS) increased from 202 to 254 (an increase of 25.4%). The improved model exhibits higher robustness and real-time performance in complex scenarios, providing an efficient solution for forest resource surveys.
精确率指正样本中被预测为正样本占比;召回率指在预测为正的样本中确是正样本的占比。AP指标反映目标检测对单个类别的准确性,mAP表示求平均AP值,mAP50指标代表了IoU(intersection over union)阈值为50%时的mAP值,mAP50-95则为IoU从50%开始,以步长为0.05上升到95%时的平均mAP值,判断predicted bbox和ground truth bbox之间的IoU大于50才被认定为正确的检测。检测速度每秒帧数(FPS)用于评估模型在批量大小为1的推理速度。
AKDOĞANC, ÖZERT, OĞUZY.PP-YOLO:Deep learning based detection model to detect apple and cherry trees in orchard based on histogram and wavelet preprocessing techniques[J].Computers and Electronics in Agriculture,2025,232:110052.
DOUA L, LUH, DUY J,et al.Urban tree detection based on enhanced YOLO v5[J].Journal of Nanjing Institute of Technology(Natural Science Edition),2023,21(3):1-7.
ZHUJ, YUW H, TANM X,et al.Object detection optimization method based on YOLO v10[J].Computer Engineering,2024,50(2):123-132.
[16]
YUW H, ZHOUP, YANS C,et al.InceptionNeXt:When inception meets ConvNeXt[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).June 16-22,2024,Seattle,WA,USA.IEEE,2024:5672-5683.
[17]
TANM X, PANGR M, LEQ V.EfficientDet:Scalable and efficient object detection[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).June 13-19,2020,Seattle,WA,USA.IEEE,2020:10778-10787.
[18]
LIUX Y, PENGH W, ZHENGN X,et al.EfficientViT:Memory efficient vision transformer with cascaded group attention[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).June 17-24,2023,Vancouver,BC,Canada.IEEE,2023:14420-14430.
[19]
CHENJ R, KAOS H, HEH,et al.Run,don’t walk:Chasing higher FLOPS for faster neural networks[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).June 17-24,2023,Vancouver,BC,Canada.IEEE,2023:12021-12031.
[20]
WOO S, DEBNATHS, HUR H,et al.ConvNeXt V2:Co-designing and scaling ConvNets with masked autoencoders[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).June 17-24,2023,Vancouver,BC,Canada.IEEE,2023:16133-16142.
[21]
LIH L, LIJ, WEIH B,et al.Slim-neck by GSConv:A lightweight-design for real-time detector architectures[J].Journal of Real-Time Image Processing,2024,21(3):62.
YANGZ Q, GUANQ, ZHAOK,et al.Multi-Branch auxiliary fusion YOLO with re-parameterization heterogeneous convolutional for accurate object detection[EB/OL].arXiv preprint,2024.arXiv:
[24]
SHID.TransNeXt:Robust foveal visual perception for vision transformers[EB/OL].arXiv preprint,2024.arXiv:
[25]
LIB X, FANGJ D, ZHAOY.RTDETR-Refa:A real-time detection method for multi-breed classification of cattle[J].Journal of Real-Time Image Processing,2025,22(1):38.
[26]
ZHAOY A, LYUW Y, XUS L,et al.DETRs beat YOLOs on real-time object detection[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).June 16-22,2024,Seattle,WA,USA.IEEE,2024:16965-16974.
[27]
FINDERS E, AMOYALR, TREISTERE,et al.Wavelet convolutions for large receptive fields[EB/OL].arXiv preprint,2024.arXiv:
[28]
LIH, WUA Y, JIANGZ Y,et al.Improving object detection in YOLOv8n with the C2f-f module and multi-scale fusion reconstruction[C]//2024 IEEE 6th Advanced Information Management,Communicates,Electronic and Automation Control Conference (IMCEC).May 24-26,2024,Chongqing,China.IEEE,2024:374-379.
[29]
ZHUJ, HUT, ZHENGL H,et al.YOLOv8-C2f-faster-EMA:An improved underwater trash detection model based on YOLOv8[J].Sensors,2024,24(8):2483.
[30]
ZHAOZ L, DONGM.Channel-spatial dynamic convolution:An exquisite omni-dimensional dynamic convolution[C]//2023 8th International Conference on Intelligent Computing and Signal Processing(ICSP).April 21-23,2023,Xi’an,China.IEEE,2023:1707-1711.
[31]
WIGHTMANR, TOUVRONH, JÉGOUH.ResNet strikes back:An improved training procedure in timm[EB/OL].arXiv preprint,2021.arXiv:
[32]
TANM X, PANGR M, LEQ V.EfficientDet:Scalable and efficient object detection[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).June 13-19,2020,Seattle,WA,USA.IEEE,2020:10778-10787.
[33]
ZHAOX D, CHENJ L, LIUM M,et al.Multi-scale attention-based feature pyramid networks for object detection[M]//Image and Graphics.Cham:Springer International Publishing,2021:405-417.
[34]
KHANAMR, HUSSAINM.What is YOLOv5:A deep look into the internal features of the popular object detector[EB/OL].arXiv preprint,2024.arXiv:
[35]
VARGHESER, SAMBATHM.YOLOv8:A novel object detection algorithm with enhanced performance and robustness[C]//2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems(ADICS).April 18-19,2024,Chennai,India.IEEE,2024:1-6.