Mathematical Morphology Research Articles

Auto-segmentation of head and neck tumors in positron emission tomography images using non-local means and morphological frameworks.

Accurately segmenting head and neck cancer (HNC) tumors in medical images is crucial for effective treatment planning. However, current methods for HNC segmentation are limited in their accuracy and efficiency. The present study aimed to design a model for segmenting HNC tumors in three-dimensional (3D) positron emission tomography (PET) images using Non-Local Means (NLM) and morphological operations. The proposed model was tested using data from the HECKTOR challenge public dataset, which included 408 patient images with HNC tumors. NLM was utilized for image noise reduction and preservation of critical image information. Following pre-processing, morphological operations were used to assess the similarity of intensity and edge information within the images. The Dice score, Intersection Over Union (IoU), and accuracy were used to evaluate the manual and predicted segmentation results. The proposed model achieved an average Dice score of 81.47 ± 3.15, IoU of 80 ± 4.5, and accuracy of 94.03 ± 4.44, demonstrating its effectiveness in segmenting HNC tumors in PET images. The proposed algorithm provides the capability to produce patient-specific tumor segmentation without manual interaction, addressing the limitations of current methods for HNC segmentation. The model has the potential to improve treatment planning and aid in the development of personalized medicine. Additionally, this model can be extended to effectively segment other organs from limited annotated medical images.

Research on vehicle speed identification method based on time interpolation method and feature point recognition

Based on camera calibration in detection area, an improved method of time interpolation for vehicle speed identification is proposed. Firstly, calibration of camera parameters during photogrammetry is carried out through preset rectangular detection area. Secondly, the key frames are extracted from the video images and the target vehicle feature points and their image coordinates are identified by the image morphological processing. Then the actual coordinates of the characteristic points are determined through the coordinate conversion. Finally, the improved time interpolation method considering the acceleration between frames is used to identify the vehicle speed, so as to improve the accuracy of the vehicle speed measurement. The results proved that the vehicle speed calculated by improved time interpolation method is more reliable, and the error is reduced by 0.7%.

Moving target detection based on background modeling and frame difference

The integration of Unmanned Aerial Vehicle (UAV) technology with moving target detection has diverse applications in military reconnaissance, space remote sensing, and smart cities. Traditional motion-based target detection algorithms offer fast processing speeds but lack accuracy. Deep learning-based algorithms, while accurate for specific targets only, are complex and not suitable for resource-limited UAV platforms and lack real-time performance. Therefore, this study proposes a real-time moving target detection algorithm for UAV platforms based on traditional frame difference algorithm. The purpose of this algorithm is to improve detection accuracy, which has been hindered by the limitations of traditional algorithms caused by camera shake, background changes, and fast-moving targets. The algorithm involves rough background modeling, background updating during subsequent video image sequences, image morphology processing, and background compensation. Experimental results from multiple sets of UAV-borne video data demonstrate the algorithm's high target detection rate, low false alarm rate, and ability to detect moving targets stably in complex environments. The proposed algorithm achieves a speed of 25 FPS and a detection accuracy of 91.8%, meeting real-time and accurate detection requirements.

Alterations in Cortical Thickness in Children with Congenital Strabismus and Dissociated Vertical Deviation

Introduction: Little is known about the morphometric changes that occur in the cerebral cortex of children with congenital strabismus and Dissociated Vertical Deviation (DVD). Mathematical morphology provides the opportunity to discover alterations in cortical thickness. Objective: Identify and measure differences in cortical thickness in children with congenital strabismus and DVD. Materials and Methods: Twenty-two MRI brain studies were conducted, each comprising 120 sagittal slices of 1 mm. A cohort of 11 healthy children formed the control group, and a cohort of 11 children with congenital strabismus and Dissociated Vertical Deviation, both in Dissociated Esotropia and Dissociated Exotropia formed the case group. Differences in cortical thickness between both groups were compared and identified using the FreeSurfer suite. Subsequently, areas of interest were analyzed using the ImageJ program to determine density, area, size, and pixel quantity in these regions. Study results are presented through tables and figures. Results: Twelve inflated brain projections were obtained from 2640 MRI images. Thickness differences between both groups exhibited three areas with significant morphometric changes: the Left Angular Gyrus, Left Supramarginal Gyrus, and Right Pars Opercularis. Discussion: Children with congenital strabismus and DVD showed changes in cortical thickness that could be identified and measured through Voxel-Based Morphometry (VBM). It is plausible to consider that these changes are related to suppression, aiming to avoid diplopia.

Recognition for Frontal Emergency Stops Dangerous Activity Using Nano IoT Sensor and Transfer Learning

Currently, it is difficult to extract the depth feature of the frontal emergency stops dangerous activity signal, which leads to a decline in the accuracy and efficiency of the frontal emergency stops the dangerous activity. Therefore, a recognition for frontal emergency stops dangerous activity algorithm based on Nano Internet of Things Sensor (NIoTS) and transfer learning is proposed. First, the NIoTS is installed in the athlete’s leg muscles to collect activity signals. Second, the noise component in the activity signal is removed using the de-noising method based on mathematical morphology. Finally, the depth feature of the activity signal is extracted through the deep transfer learning model, and the Euclidean distance between the extracted feature and the depth feature of the frontal emergency stops dangerous activity signal is compared. If the European distance is small, it can be judged as the frontal emergency stops dangerous activity, and the frontal emergency stops dangerous activity recognition is realized. The results show that the average time delay of activity signal acquisition of the algorithm is low, the signal-to-noise ratio of the action signal is high, and the activity signal mean square error is low. The variance of the frontal emergency stops dangerous activity recognition does not exceed 0.5. The difference between the appearance time of the dangerous activity and the recognition time of the algorithm is 0.15 s, it can accurately and quickly recognize the frontal emergency stops the dangerous activity.

Experimental and Numerical Research on Water Transport during Adsorption and Desorption in Cement-Based Materials

The durability of cement-based materials is related to water transport and storage in their pore network under different humidity conditions. To understand the mechanism and characteristics of water adsorption and desorption processes from the microscopic scale, this study introduces different points of view for the pore space model generation and numerical simulation of water transport by considering the “ink-bottle” effect. On the basis of the pore structure parameters (i.e., pore size distribution and porosity) of cement paste and mortar with water-binder ratios of 0.3, 0.4 and 0.5 obtained via mercury intrusion porosimetry, randomly formed 3D pore space models are generated using two-phase transformation on Gaussian random fields and verified via image analysis method of mathematical morphology. Considering the Kelvin-Laplace equation and the influence of “ink-bottle” pores, two numerical calculation scenarios based on mathematical morphology are proposed and applied to the generated model to simulate the adsorption-desorption process. The simulated adsorption and desorption curves are close to those of the experiment, verifying the effectiveness of the developed model and methods. The obtained results characterize water transport in cement-based materials during the variation of relative humidity and further explain the hysteresis effect due to “ink-bottle” pores from the microscopic scale.

Automated Detection of Malaria and Sickle Cells in Blood Smears Using Edge Detection and Morphological Operations

Automated Detection of Malaria and Sickle Cells in Blood Smears Using Edge Detection and Morphological Operations

Multi-scale local feature enhanced transformer network for pavement crack detection

目的 道路裂缝是路面病害的早期征兆。定期监测路面状况、及时准确地发现路面裂缝对于交通养护机构降低成本、保证路面结构的可靠性和耐久性以及提高驾驶安全性、舒适性有重要意义。目前基于卷积神经网络的深度学习模型在长距离依赖建模方面存在不足，模型精度难以满足真实路面环境下的裂缝检测任务。一些模型通过引入空间/通道注意力机制进行长距离依赖建模，但是会导致计算量和计算复杂程度增加，无法实现实时检测。鉴于此，本文提出一种基于Transformer编码—解码结构的深度神经网络道路裂缝检测模型CTNet （crack transformernetwork）。方法 该模型主要由Transformer注意力模块、多尺度局部特征增强模块、上采样模块和跨越连接4部分构成。采用Transformer注意力机制能更有效提取全局和长距离依赖关系，克服传统卷积神经网络表征输入信息的短距离相关缺陷。同时，为适应裂缝尺寸变化多样性，将Transformer与多尺度局部特征增强模块相结合，从而有效整合不同尺度局部信息，克服Transformer局部特征建模不足。结果 通过与DeepCrack模型在不同裂缝检测数据集中的比较表明，本文提出的多尺度局部特征增强Transformer网络能快速、准确地分割路面裂缝，且效率更优。定量研究结果表明，CTNet在更有挑战性的CrackLS315数据集中的精度、召回率和F1值达到91.38%、80.38%和85.53%，明显优于对比方法。在CrackWH100数据集中，精度、召回率和F1值进一步提升，分别达到92.70%、90.52%和91.60%。此外，CTNet的训练速率提升至DeepCrack模型的6.78倍。结论 CTNet可以实现强噪声背景下的道路裂缝检测，检测效果优于目前最优方法，且参数量小，易于训练和部署。;Objective The pavement-relevant inspection is focused on the optimization for pavement cracks early-alarming detection and the preservation of pavement structure. However，conventional image processing-based techniques are laborintensive and time-consuming，such as edge detection，threshold segmentation，template matching，and morphology operations. It is challenged to the geometric and spectral complexities of pavement crack and its contexts（e. g. ，illumination variation，oil or water stains，and shadows caused by trees and vehicles）. The convolution neural network（CNN）based deep learning image processing techniques have been developing intensively. However，the CNN-based methods are less effective in long-range dependency modeling，which may cause insufficient detection results in complicated road surface scenarios. Some works are related to attention mechanisms like spatial or channel attention modules，and self-attention modules. However，these attention mechanism-based operations are still challenged for their sophistication and computational cost. Method To detect pavement cracks efficiently and effectively，we develop a novel Transformer-based encoderdecoder neural network，called CTNet，which consists of Transformer blocks，multi-scale local feature enhanced blocks， upsampling blocks，and skip connections. The CTNet can achieve more long-range dependency and global receptive field in terms of multi-head self-attention-based Transformer mechanism. Although Transformer is featured by high running efficiency and low computational overhead demand，it is infeasible to model local contextual information because Token generation can break the connections of neighboring regions. Thus，to capture more multi-scale local information，we design a multi-scale local feature-enhanced block in terms of a multiple dilation ratios-relevant dilation convolution block. Especially，the designed multi-scale local feature enhancement block is melted into each Transformer block for local information complement. Both of local and global low-level contextual features can be captured for feature enhancement. Afterwards，a novel decoder path is implemented to extract high-level features. The decoder consists of the Transformer blocks similar to the up-sampling blocks and the spatial details can be restored for end-to-end segmentation. Result To demonstrate the efficiency and effectiveness of our proposed CTNet，a series of comparative analyses and ablation studies are carried out on three datasets. First，the CTNet can optimize running efficiency，as well as comparable computation overhead and complexity compared to the current UNet，SegNet，DeepCrack，and SwinUnet. Second，CTNet is 6. 78 times faster than the second-best DeepCrack model in terms of training speed. On CrackLS315 dataset，quantitative analyses are also showed that the optimal CTNet is obtained a precision of 91. 38%，a recall of 80. 38%，and a F1 measure of 85. 53% of each；on CrackWH100 dataset，CTNet can obtain a precision of 92. 70%，a recall of 90. 52%，and a F1 measure 91. 60% of each as well. However，it is still challenged to lack of local information when pure-Transformer-based Swin-UNet performed not well compared to fully convolution networks. Furthermore，the CTNet is insufficient to converge when the local blocksenhanced are removed. In summary，the Transformer-based CTNet is beneficial to multi-scenario pavement cracks in terms of the global receptive field. The CTNet can get pavement crack detection results consistently. Conclusion The proposed CTNet has its potentials to deal with noisy pavement images for pavement crack detection.

Video Sensor-Based Automatic License Plate Recognition of Static and Moving Vehicles

In Bangladesh, present transportation system needs suitable control schemes to ensure safety and mobility to the road users. License Plate Recognition (LPR) is one of the control schemes which can be used to ensure road safety using video sensors. Moreover, from installation point of view, it is cheaper than Radiofrequency Identification (RFID) system. Automated LPR is adopted to reduce human involvement which causes loss in accuracy and reliability in reading license plate. The extracted information from LPR can be used in various schemes, such as electronic payment system, traffic surveillance and extracting vehicle trajectory. In this paper, we are presenting an offline dynamic LPR system, which uses geometric properties of Bengali alphabets stored in alphabet template database to recognize license plates from video. The system comprises of three main components: 1) License Plate Locating (LPL), 2) Character Segmentation (CS), and 3) Geometric Character Recognition (GCR). In this regard, the image is converted into a binary image considering morphological operations and color modification. A new ratio-based detection algorithm has been proposed for the segmentation of license plate. Finally, the alphabet template database and the character extracted from detected license plate are correlated to recognize alphanumeric character which deduces the plate number. Thus, the plate number can be utilized for further action.

A MODEL-INVARIANT APPROACH FOR 3D ASSESSMENT OF STRUCTURAL CHANGES IN PRECLINICAL ANIMAL MODELS OF OA

Quantitative morphometric analysis (QMA) of joint tissues has been shown to be sensitive to OA in preclinical rabbit and rat models [1,2]. Traditionally performed using workflows based on morphological operations such as dilation and erosion, its sensitivity is dependent on animal scale and voxel size, requiring parametric studies to be performed in every new implementation. We hypothesize that using probabilistic approaches, it is possible to capture OA structural changes while remaining invariant to the animal model and voxel size. In this work, a pilot study developing an empirical and model-invariant approach is conducted to automatically assess osteophyte presence, a hallmark feature of OA. Develop an empirical probabilistic approach providing an automatic and model-invariant solution to statistically capture morphological changes due to OA by detecting osteophyte activity. Micro-computed tomography scans (SCANCO Medical AG, Switzerland) from previous rabbit and rat studies of OA were used [1,2]. The rabbit dataset (18 μm voxel size) consists of scans from rabbits that underwent ACL transection in the right knee (rabbit OP, n = 8), while the left knee was kept as contralateral control (rabbit NO, n = 8). The rat dataset (10 μm voxel size) consists of scans from rats that underwent ACL transection and medial meniscectomy in the right knee (rat OP, n = 11), while the other was kept as contralateral control (rat NO, n = 11). For each image of the joint, the cortical bone of the femur and tibia were segmented, and its 3D thickness was calculated using a sphere-fitting distance transform [3]. A one-voxel-thick outer layer was segmented from the resulting thickness map, and its thickness histogram was empirically estimated by a series of statistical distributions. Parameters describing the best-fit distribution, determined with negative log-likelihood, were analyzed to determine their sensitivity to OA and the presence of osteophytes using paired two-sided t-tests. Visual inspections revealed clustering of surface voxels with small thickness values around areas with osteophytes (Figure 1a), as compared to those without (Figure 1b). Surface thickness values were optimally fitted to a Gamma distribution using negative log-likelihood for most femur and tibia across both animals. The shape parameter of the fitted Gamma distribution was found to be vary significantly with OA (p < 0.01) and in the presence of osteophytes (p < 0.05). By combining traditional image processing with empirical distribution fitting, an automated and model-invariant method for structural assessment that is sensitive to OA and the presence of osteophyte can be achieved. Future work will focus on characterizing the underlying biological processes of abnormal bone remodeling to gain insight into the emergence of the fitted Gamma distribution observed in this study. Discovery Projects scheme from the Australian Research Council (DP180101838). None CORRESPONDENCE ADDRESS: [email protected]

Noncontact High-Precision Aperture Detection Based on Distortion Compensation and Adaptive Mean Blur Algorithm

High-precision apertures are of great significance for the manufacture and assembly of precision instruments. There are currently two common types of methods for aperture measurement: manual measurement methods and computer vision-based methods. The computer vision-based methods work by using industrial cameras and morphological operators and image interpolation algorithms, and can dramatically reduce human errors. However, these methods still suffer from two problems. First, common morphological operations would damage the edge information of the image and fail to capture the accurate edge of the hole. Second, although the magnification algorithm based on image interpolation has an obvious edge positioning effect, it tends to make the edge jagged and blurred. To reduce the influence of image distortion on the error of subsequent image processing, we perform distortion compensation on the image, and use the grid intersection and pixel information in the vicinity for fitting and analysis, and calculate the lens distortion coefficient to correct the image distortion. Then, to obtain more accurate edge information, we improve the Canny operator by using the mean blur with an adaptive threshold that preserves the edge in place of the Gaussian blur and adding a new convolution kernel to smooth the image to provide a more accurate profile for subsequent aperture measurements. With these efforts, we propose a non-contact high-precision aperture measurement method based on a distortion compensation and adaptive mean blur algorithm. The proposed method has a measurement error no greater than ±0.03 mm, and achieves a 100% pass rate. The method solves the problem that the edge contour of the hole cannot be accurately extracted due to edge blurring, and allows for high-precision detection of apertures in industrial fields. The method is expected to be used with a number of precision instruments for aperture size measurement.

The Interaction of Nominalisation and Compounding in Polish: On the Analysis of Listonosz ‘Mail Carrier’ and Korkociąg ‘Corkscrew’ in Construction Morphology

This paper compares two analyses of interfixal-paradigmatic formations in Polish, such as listonosz ‘mail carrier’ and korkociąg ‘corkscrew’. They are often interpreted as exocentric compounds consisting of a noun stem and a verb stem connected by a linking vowel, i.e. as [N+LV+V]N formations (Kurzowa; Booij, “Morphology”). The proposal defended here is the treatment of such lexemes as [N+LV+[V]N]N compounds, in which the right-hand verb stems are assumed to undergo conversion into nouns. Semantic arguments are presented in support of such an assumption. Moreover, differences are emphasised between properties of the interfixal-paradigmatic formations in question and features of exocentric compounds in which the verb stem stands in the initial position, e.g. moczymorda ‘drunkard’. The analysis adopted here in the framework of Construction Morphology involves the simultaneous application of two morphological operations, namely compounding and nominalisation (i.e. verb-to-noun conversion), in the formation of interfixal-paradigmatic lexemes. Such an assumption allows for the explanation of the phenomenon of “embedded productivity” (Booij, Construction Morphology).

Image Manipulation Localization Using Multi-Scale Feature Fusion and Adaptive Edge Supervision

Image manipulation localization is a technique that can efficiently segment the tampered regions from a suspicious image. Existing work usually trains a detection model by fusing the features from diverse data streams, e.g., noise inconsistency, recompression inconsistency, and local inconsistency. They, however, ignore a fact that not all tampered images contain these data streams. As a result, high feature redundancy may cause a large number of false detection for tampered region. To address this problem, this paper designs an end-to-end high-confidence localization network architecture. First, deep convolutional neural networks are utilized to extract multi-scale feature sets from the RGB streams. We then design a semantic refined bi-directional feature integration module to fully fuse multi-scale adjacent features and significantly enhance feature representation. Subsequently, morphological operations are introduced to extract multi-scale edge information, which can efficiently reduce feature redundancy by generating wider high-resolution edges during image reconstructing. Finally, a deep semantic residual decoder is sequentially re-constructed by spreading deep semantic information into each decoding stage. The proposed method can not only improve the manipulation localization accuracy, but also guarantee the model robustness. Extensive experiments demonstrate that our method can obtain an effective performance in locating forged regions over different large-scale image sets, and outperforms most of state-of-the-art methods with higher localization accuracy and stronger robustness.

Segmentation of Moiré Fringes of Scoliotic Spines Using Filtering and Morphological Operations

Segmentation of Moiré Fringes of Scoliotic Spines Using Filtering and Morphological Operations

Fuzzy text/non-text classification of document images based on morphological operator, wavelet transform, and strong feature vector

Fuzzy text/non-text classification of document images based on morphological operator, wavelet transform, and strong feature vector

Efficient Crack Severity Level Classification Using Bilayer Detection for Building Structures

Detection of cracks at the early stage is considered as very constructive since precautionary steps need to be taken to avoid the damage to the civil structures. Moreover, identifying and classifying the severity level of cracks is inevitable in order to find the stability of buildings. Hence, this paper proposes an efficient strategy to classify the cracks into fine, medium, and thick using a novel bilayer crack detection algorithm. The bilayer crack detection algorithm helps in extracting the requisite features from the crack for efficient classification. The proposed algorithm works well in the dark background and connects the discontinued cracks too. The first layer is used to detect cracks under texture variations and manufacturing defects, through segmented adaptive thresholding and morphological operations. The residual noise present in the output of the first layer is removed in the second layer of crack detection. The second layer includes the double scan and the noise reduction algorithms and is used to join the missed crack parts. As a result, a segmented crack is formed. Further classification is done using an ensemble classifier with bagging, and decision tree techniques by extracting the geometrical features and the weaker crack criterion from the segmented part. The results of the proposed technique are compared with the existing techniques for different datasets and have obtained a rise in True Positive Rate (TPR), accuracy and precision value. The proposed technique is also implemented in Raspberry Pi for further real-time evaluation.

Multioperator Morphological Undecimated Wavelet for Wheelset Bearing Compound Fault Detection

Wheelset bearing compound faults are observed as impulses in the vibration measurements but immersed in noise. The morphological undecimated wavelet (MUW) is an effective tool for recovering fault-related impulses from vibration mixtures. To date, the reported MUWs adopt an identical morphological operator in each level of decomposition without exception. However, effectively capturing signal signatures by repeatedly using a noise elimination operator, an impulse extraction operator, or even the product of two operators is unattainable. Spurred by this deficiency, in this paper, we propose a multi-operator MUW (MOMUW). A three-level structure: noise reduction, impulse extraction, and further denoising and feature enhancement, is developed in the MOMUW, and different morphological operators are designed for each decomposition level to more purposefully denoise and extract impulse features. The developed MOMUW is applied to measured wheelset bearing vibration data, with the results demonstrating that it can accurately detect wheelset bearing compound faults. Compared with the reported MUWs, the proposed approach presents superior performance. Furthermore, MUWs with varying operators and levels are analyzed and compared. Their success and failure in bearing fault diagnosis are interpreted and discussed, laying a theoretical foundation for constructing new MUWs.

Auxiliary Classifier of Generative Adversarial Network for Lung Cancer Diagnosis

The classification of lung nodules is a challenging problem as the visual analysis of the nodules and non-nodules revealed homogenous textural patterns. In this work, an Auxiliary Classifier (AC)-Generative Adversarial Network (GAN) based Lung Cancer Classification (LCC) system is developed. The proposed AC-GAN-LCC system consists of three modules; preprocessing, Lungs Region Detection (LRD), and AC-GAN classification. A Wiener filter is employed in the preprocessing module to remove the Gaussian noise. In the LRD module, only the lung regions (left and right lungs) are detected using iterative thresholding and morphological operations. In order to extract the lung region only, flood filling and background subtraction. The detected lung regions are fed to the AC-GAN classifier to detect the nodules. It classifies the nodules into one of the two classes, i.e., binary classification (such as nodules or non-nodules). The AC-GAN is the extended version of the conditional GAN that predicts the label of a given image. Three different optimization techniques, adaptive gradient optimization, root mean square propagation optimization, and Adam optimization are employed for optimizing the AC-GAN architecture. The proposed AC-GAN-LCC system is evaluated on the Lung Image Database Consortium (LIDC) database Computed Tomography (CT) scan images. The proposed AC-GAN-LCC system classifies ∼15000 CT slices (7310 non-nodules and 7685 nodules). It provides an overall accuracy of 98.8% on the LIDC database using Adam optimization by a 10-fold cross-validation approach.

Automatic Optic Disc Detection in Retinal Images Using FKMT‒MOPDF

In recent days, detecting Optic Disc (OD) in retinal images has been challenging and very important to the early diagnosis of eye diseases. The process of detecting the OD is challenging due to the diversity of color, intensity, brightness and shape of the OD. Moreover, the color similarities of the neighboring organs of the OD create difficulties during OD detection. In the proposed Fuzzy K‒Means Threshold (FKMT) and Morphological Operation with Pixel Density Feature (MOPDF), the input retinal images are coarsely segmented by fuzzy K‒means clustering and thresholding, in which the OD is classified from its neighboring organs with intensity similarities. Then, the segmented images are given as the input to morphological operation with pixel density feature calculations, which reduce the false detection in the small pixel of the OD. Finally, the OD area is detected by applying the Sobel edge detection method, which accurately detects the OD from the retinal images. After detection optimization, the proposed method achieved Sensitivity (SEN), Specificity (SPEC) and Accuracy (ACC), with 96.74%, 96.78% and 96.92% in DiaretDB0 (Standard Diabetic Retinopathy Database Calibration level 0), 97.12%, 97.10% and 97.75% in DiaretDB1 (Standard Diabetic Retinopathy Database Calibration level 1) and 97.19%, 97.47% and 97.43% in STARE (Structured Analysis of the Retina) dataset respectively. The experimental results demonstrated the proposed method’s efficiency for segmenting and detecting OD areas.

Morphological Transformation and Spatial-Logical Aggregation for Tree Species Classification Using Hyperspectral Imagery

Hyperspectral image (HSI) consists of abundant spectral and spatial characteristics, which contribute to a more accurate identification of materials and land covers. However, most existing methods of hyperspectral image analysis primarily focus on spectral knowledge or coarse-grained spatial information while neglecting the fine-grained morphological structures. In the classification task of complex objects, spatial morphological differences can help to search for the boundary of fine-grained classes, e.g., forestry tree species. Focusing on subtle traits extraction, a spatial-logical aggregation network (SLA-NET) is proposed with morphological transformation for tree species classification. The morphological operators are effectively embedded with the trainable structuring elements, which contributes to distinctive morphological representations. We evaluate the classification performance of the proposed method on two tree species datasets, and the results demonstrate that the proposed SLA-NET significantly outperforms the other state-of-the-art classifiers.