Wood Species Recognition with Small Data: A Deep Learning Approach

Sun, Ye; Lin, Qizhao; He, Xin; Zhao, Youjie; Dai, Fei; Qiu, Jian; Cao, Yong

doi:10.2991/ijcis.d.210423.001

Cited by 19 publications

(14 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The current (practical) experiences show that a combination of the described approaches offer the best solution for the development of machine learning systems (Sun et al 2021). The advantages of "classification" and "structural model" are based on a direct calculation of why an image is assigned to a certain class.…”

Section: Studies Using Wood Anatomy Methodsmentioning

confidence: 99%

Tracing the world’s timber: the status of scientific verification technologies for species and origin identification

Low

Schmitz

Boeschoten

et al. 2022

IAWA J.

View full text Add to dashboard Cite

Summary Illegal logging and illegal timber trade is a global problem. Anatomical, genetic, and chemical techniques support illegal logging legislation by verifying the species and geographic origin of timber. In principle, these methods can be used to identify timber species and the origin of harvest, however, the availability of specific tests for important timber species is unclear. We review the status of these methods for the top 322 global priority timber taxa. Our results show that for species identification, reference data exist for 100% of taxa using wood anatomy, 86% using genetics, 41% for using DART TOFMS, and 6% using NIRS. For origin identification, data exist for 24% of taxa, with most studies applying genetic approaches (23%). No studies have developed forensic-ready tests for the global priority timber taxa. The review highlights that the current potential for identifying species is greater than for geographic origin and more research focused on determining the geographical origin of timber is required. Based on the current rate, it will take approx. 27 years to generate geographic data for all 322 priority taxa. Finally, we identify research opportunities to improve global timber tracing efforts. Our findings indicate more research is needed, and quickly so that scientific verification can support regulators to combat illegal logging.

show abstract

Section: Studies Using Wood Anatomy Methodsmentioning

confidence: 99%

Tracing the world’s timber: the status of scientific verification technologies for species and origin identification

Low

Schmitz

Boeschoten

et al. 2022

IAWA J.

View full text Add to dashboard Cite

show abstract

“…It has been widely used in fields including text classification, the prediction of short-term water demand, and annual average rainfall forecasts [15][16][17][18]. However, the KNN method needs to continuously store the known training data during the learning process and requires a large amount of RAM [19,20]. The VGG-16 model includes a convolution layer, a full connection layer, and a pooling layer.…”

Section: Introductionmentioning

confidence: 99%

LF-CNN: Deep Learning-Guided Small Sample Target Detection for Remote Sensing Classification

Li¹,

Liu²,

Zhao³

et al. 2022

Computer Modeling in Engineering &Amp; Sciences

View full text Add to dashboard Cite

Target detection of small samples with a complex background is always difficult in the classification of remote sensing images. We propose a new small sample target detection method combining local features and a convolutional neural network (LF-CNN) with the aim of detecting small numbers of unevenly distributed ground object targets in remote sensing images. The k-nearest neighbor method is used to construct the local neighborhood of each point and the local neighborhoods of the features are extracted one by one from the convolution layer. All the local features are aggregated by maximum pooling to obtain global feature representation. The classification probability of each category is then calculated and classified using the scaled expected linear units function and the full connection layer. The experimental results show that the proposed LF-CNN method has a high accuracy of target detection and classification for hyperspectral imager remote sensing data under the condition of small samples. Despite drawbacks in both time and complexity, the proposed LF-CNN method can more effectively integrate the local features of ground object samples and improve the accuracy of target identification and detection in small samples of remote sensing images than traditional target detection methods.

show abstract

“…They also applied data augmentation to reduce overfitting and improve performance. Sun et al [11] selected 25 wood species for training and testing. The woodblocks of these species collected in two kinds of ways (wood factories in Yunnan Province, China, the others were obtained from the Wood Herbarium of Southwest Forestry University).…”

Section: Introductionmentioning

confidence: 99%

“…Figueroa-Mata [10] et al proposed a deep CNN for wood samples of tree species recognition, which used pre-trained weights to finetune the Resnet50 model to achieve the highest accuracy of 98.03%. Sun et al [11] proposed a deep-learning-based wood species recognition method, which used 20X amplifying glass to acquire wood images, they extracted the image features with ResNet50 neural network, refined the features with linear discriminant analysis (LDA), and recognized the wood species with a k-Nearest Neighbor algorithm (KNN classifier). Lopes et al [12] presented the feasibility of the InceptionV4_ResNetV2 convolutional neural network to classify 10 North American hardwood species with 92.60% of accuracy and precision-recall rate of 0.98.…”

Section: Introductionmentioning

confidence: 99%

Multi-Fusion Approach for Wood Microscopic Images Identification Based on Deep Transfer Learning

et al. 2021

View full text Add to dashboard Cite

With the wide increase in global forestry resources trade, the demand for wood is increasing day by day, especially rare wood. Finding a computer-based method that can identify wood species has strong practical value and very important significance for regulating the wood trade market and protecting the interests of all parties, which is one of the important problems to be solved by the wood industry. This article firstly studies the establishment of wood microscopic images dataset through a combination of traditional image amplification technology and Mix-up technology expansion strategy. Then with the traditional Faster Region-based Convolutional Neural Networks (Faster RCNN) model, the receptive field enhancement Spatial Pyramid Pooling (SPP) module and the multi-scale feature fusion of Feature Pyramid Networks (FPN) module are introduced to construct a microscopic image identification model based on the migration learning fusion model and analyzes the three factors (Mix-up, Enhanced SPP and FPN modules) affecting the wood microscopic image detection model. The experimental results show that the proposed approach can identify 10 kinds of wood microscopic images, and the accuracy rate has increased from 77.8% to 83.8%, which provides convenient conditions for further in-depth study of the microscopic characteristics of wood cells and is of great significance to the field of wood science.

show abstract

Wood Species Recognition with Small Data: A Deep Learning Approach

Cited by 19 publications

References 29 publications

Tracing the world’s timber: the status of scientific verification technologies for species and origin identification

Tracing the world’s timber: the status of scientific verification technologies for species and origin identification

LF-CNN: Deep Learning-Guided Small Sample Target Detection for Remote Sensing Classification

Multi-Fusion Approach for Wood Microscopic Images Identification Based on Deep Transfer Learning

Contact Info

Product

Resources

About