基于改进语义分割网络的落叶松电阻率层析成像的图像心材区域识别

打开文本图片集

关键词：落叶松；心材；ERT；DeepLabv3+；混合注意力中图分类号：S791.222；TP391.4 文献标识码：A DOI：10.7525/j.issn.1006-8023.2025.06.018

Heartwood Region Recognition in Larch ERT Images Based on Improved Semantic Segmentation Network

LINGHao，XU Huadong*，GUO Xuhui （College of Mechanical and Electrical Engineering，NortheastForestryUniversity，Harbin15O040，China）

Abstract：Aiming atthedificultyofaccuratelyrecognizing the heartwoodoflarch standing trees，this study proposes an intellgent segmentation method based on electrical resistance tomography（ERT）.Bycomparing andanalyzing ERTimagesandphysicalsectionsoflarchsamples，itisfound thattheresistancechangerateofthe junctionregionof the heart sapwood reached 90 % -94%.Based on this threshold range，the segmentation annotation criterion for ERT image segmentationof larch standing tree heartwood was established.Due tothe dificulty inacquiring ERTimages，twodatasets， Mini-200 small samples and Mids-32OO large samples，were constructed to quickly adapt to the segmentation task throughthesmalltraining set，andcombined withthelarge trainingsettoimprovethe modelrobustness andreduceoverfitting.The improved semantic segmentation network（DeepLabv3 + ）model was proposed to optimize its feature extraction capability byintroducing ResNet101，convolutional block atention mechanism（CBAM）and data communication module（DCM）.The ablation experiments showed that the five evaluation indexes of accuracy（A），precision （ P ），intersection over union（IoU），mean intersection over union（mIoU）and Dice lossfunction of the improved model were improved by 0.14%-0.44% compared with the base model on the Mini-2OO dataset；on the Mids-32OO dataset，the improvedDeepLabv3 †+ model had the optimal segmentation performance，and compared with the original model，the pixel accuracy（PA）and IoU of theheartwood were improved by 0.32% and 2. 45% ，respectively，and the classpixel accuracy（CPA），mIoU and Dice coefficient were improved by 0.47% ，2. 13% ， 0.25% ，respectively，and the IoU reached 98. 80% ，compared with the original model. It proves that the improved model works wellfor the segmentation of the heartwood of the ERT image of larch.

Keywords：Larch;heartwood;ERT;DeepLabv3 + ;hybridattention

0 引言

在林木培育领域，心材的精准识别直接影响材料价值评估与资源高效利用。（剩余11966字）