Multispectral Research Articles

Color router-based long-wave infrared multispectral imaging.

In the field of long-wave infrared multispectral imaging, traditional snapshot techniques often deploy broadband filters in front of the sensor to encode spectral information about the scene. However, this approach causes a significant loss of precious optical energy, especially for the limited radiation energy of the long-wave infrared region. To address this issue, we first propose an imaging strategy that replaces conventional filters with specially designed diffractive elements, which are optimized by a gradient descent algorithm. The diffractive elements enable effective steering of diverse wavelengths to their designated pixels, significantly minimizing the reflection losses throughout light transmission and thereby augmenting the system's optical energy efficiency. Secondly, we use the MST neural network to reconstruct the spectral information and realize the snapshot computational multispectral imaging. In the experiments, we concentrate the wavelength band within 8-12 μm, simulating and optimizing the design of the diffractive elements. We also discuss how this innovative design can adapt to the field change of image plane that may be encountered in the actual imaging system. Emulation experiments show that our proposed method ensures excellent spectral separation and high imaging quality under different field conditions. This study provides new ideas and practical guidance for the lightweight and efficient development of long-wave infrared multispectral imaging technology.

Enhancing Turbidity Predictions in Coastal Environments by Removing Obstructions from Unmanned Aerial Vehicle Multispectral Imagery Using Inpainting Techniques

High-resolution remote sensing of turbidity in the coastal environment with unmanned aerial vehicles (UAVs) can be adversely affected by the presence of obstructions of vessels and marine objects in images, which can introduce significant errors in turbidity modeling and predictions. This study evaluates the use of two deep-learning-based inpainting methods, namely, Decoupled Spatial–Temporal Transformer (DSTT) and Deep Image Prior (DIP), to recover the obstructed information. Aerial images of turbidity plumes in the coastal environment were first acquired using a UAV system with a multispectral sensor that included obstructions on the water surface at various obstruction percentages. The performance of the two inpainting models was then assessed through both qualitative and quantitative analyses of the inpainted data, focusing on the accuracy of turbidity retrieval. The results show that the DIP model performs well across a wide range of obstruction percentages from 10 to 70%. In comparison, the DSTT model produces good accuracy only with low percentages of less than 20% and performs poorly when the obstruction percentage increases.

Identification of the best method for detecting surface water in Sentinel-2 multispectral satellite imagery

Surface water maps are useful in a variety of disciplines from climate change analysis to water resource management. Multispectral satellite imagery can be used to derive such surface water maps using a variety of image processing methods. The medium resolution Sentinel-2 multispectral satellite imagery catalogue is currently used extensively for surface water mapping. The quality and accuracy of these maps produced from Sentinel-2 imagery can vary greatly depending on the method applied to classify the image pixels into land or water. Thus far, there has not been a consensus on which method produces the highest accuracy surface water maps, warranting a direct comparison to assess these methods in a wide range of geographic settings. Here we show that among some of the most commonly applied surface water mapping methods (NDWI, MNDWI, AWEI_SH, AWEI_NSH, AWEI_BOTH, SVM, RT, MLC, and KNN) that no single method produced the most accurate maps across the four locations studied, but AWEI_NSH performed the best overall across the four locations, and SVM was the best performing machine learning technique. Rather, each method's performance was shown to depend on the objects present in the image (e.g., built-up, shadows, vegetated shorelines, narrow waterbodies, etc.) and how successfully the method was able to classify those objects properly. This is in-line with current understanding of spectral index methods' performance, and we provide recommendations to aid remote sensing data users in choosing a suitable method based on their image's characteristics. Using these recommendations, we hope that the quality of surface water maps derived from multispectral satellite imagery will be improved for all disciplines that utilize such data by allowing users to choose the method that is best fit to the application.

Multispectral and hyperspectral image fusion via multi-model deep prior regularization

ABSTRACT Due to the difficulty of collecting a sufficient number of photons within a narrow spectral window, hyperspectral images often have low spatial resolution. In order to solve this problem without increasing hardware costs, a large number of image super-resolution algorithms have been proposed and have shown promising results. However, the performance of these algorithms is inconsistent in different datasets. To fix this situation, we proposed a multi-model deep prior regularization method for multispectral and hyperspectral image fusion. This method leverages mechanistic knowledge to combine various super-resolution results into a single algorithm, significantly improving the quality of the hyperspectral images. Specifically, we first proposed a variable residual mixed attention network. This network takes a low-resolution hyperspectral image and a high-resolution multispectral image as inputs, producing deep prior results from three different modes by adjusting spatial and channel attention. Then, we reconstructed the prior results along the spectral dimension and used them as regularization terms in our mechanistic model. We transformed the problem of finding the optimal solution of the model into solving the Sylvester equation and applied a fast solver for this equation to create the high-resolution hyperspectral images. Additionally, we used a genetic algorithm to determine the best hyperparameters for different deep prior results. Through validation on three datasets, we have demonstrated that our method performs better than existing methods in both quantitative and qualitative comparisons.

Intelligent detection of coal gangue in mining Operations using multispectral imaging and enhanced RT-DETR algorithm for efficient sorting

The coal gangue generated during coal mining causes serious environmental pollution. To efficiently screen coal gangue, we developed an efficient approach combining multi-spectral imaging with an improved Real-Time Detection Transformer (RT-DETR) for coal gangue detection. Firstly, spectral data from 25 bands were optimized into 3 bands and combined into a pseudo RGB image. Then, the backbone network was enhanced through lighter Faster-Block and Efficient Multi-scale Attention (EMA) mechanisms, with learnable position coding introduced in the Attention based Intra scale Feature Interaction (AIFI) layer and the Re-param Convolutional (RepC3) structure of the Cross scale Feature Fusion Module (CCFM) layer upgraded to Dilated Re-param Block (DRB). Additionally, we propose a Stable Diffusion model based on Low-Rank Adaptation (LoRA) fine-tuning for image generation and enhance the model’s robustness by combining it with data augmentation to simulate specific environments. Our results showed that the method achieves detection accuracy, recall, and mean average precision of 92.18%, 86.78%, and 91.95%, respectively. The lightweight model and high detection accuracy provide an effective solution for mine sorting.

UAV-Based Phenotyping: A Non-Destructive Approach to Studying Wheat Growth Patterns for Crop Improvement and Breeding Programs

Rising food demands require new techniques to achieve higher genetic gains for crop production, especially in regions where climate can negatively affect agriculture. Wheat is a staple crop that often encounters this challenge, and ideotype breeding with optimized canopy traits for grain yield, such as determinate tillering, synchronized flowering, and stay-green (SG), can potentially improve yield under terminal drought conditions. Among these traits, SG has emerged as a key factor for improving grain quality and yield by prolonging photosynthetic activity during reproductive stages. This study aims to highlight the importance of growth dynamics in a wheat mapping population by using multispectral images obtained from uncrewed aerial vehicles as a high-throughput phenotyping technique to assess the effectiveness of using such images for determining correlations between vegetation indices and grain yield, particularly regarding the SG trait. Results show that the determinate group exhibited a positive correlation between NDVI and grain yield, indicating the effectiveness of these traits in yield improvement. In contrast, the indeterminate group, characterized by excessive vegetative growth, showed no significant NDVI–grain yield relationship, suggesting that NDVI values in this group were influenced by sterile tillers rather than contributing to yield. These findings provide valuable insights for crop breeders by offering a non-destructive approach to enhancing genetic gains through the improved selection of resilient wheat genotypes.

Reflectance spectrometry and multispectral data for mapping fractures and hydrothermal alterations in the northern edge of the eastern High Atlas, Morocco

The Marginal Folds, in the north-western axis of Moroccan eastern High Atlas, are mainly composed of the Infra-Cenomanian terrains. These latter contain continental detrital deposits (red beds) and are characterized by a network of fractures. These marginal folds host significant concentrations of metallic minerals, such as Cu, Pb, Ag, and Zn, particularly at the Merija (Cu) and Bou Sellam (Pb-Zn) sites.Several techniques were applied to the Landsat-8 Oli images to improve the clarity and visibility of the linear features. Radiometric and atmospheric corrections, color compositions, directional filters, and principal component analysis were adopted. The lineaments resulting from automatic extraction, as well as their statistical analysis were validated by visual interpretation and field investigations. Statistical analysis revealed three dominants fracture systems: N-S, NE-SW, and E-W; the first two systems show a high density in the study area.The mineral alteration of the outcropping rocks is well described by ASTER multispectral imaging techniques. In this work, the effectiveness of this techniques for mineral alteration mapping was demonstrated on the Merija area. Several occurrences of mineral alteration were highlighted, using a combination of false color and band ratios, the principal component analysis (PCA), and the band ratios. This research was completed by reflectance spectrometry, covering the visible, near-infrared (VNIR), and short-wave infrared (SWIR) domains, to identify mineral alteration.Hydrothermal alteration in the Merija region includes clay, phyllite, carbonate, and phyllosilicate alteration, such as kaolinite, illite, montmorillonite, palygorskite, and vermiculite. Regarding to oxidation, as evidenced by the presence of goethite, jarosite, and hematite, it is widespread in most of the analyzed samples, resulting from the alteration of pyrite, followed by jarosite and goethite.

A Multispectral Blood Smear Background Images Reconstruction for Malaria Unstained Images Normalization

ABSTRACTMultispectral and multimodal unstained blood smear images are obtained and evaluated to offer computer‐assisted automated diagnostic evidence for malaria. However, these images suffer from uneven lighting, contrast variability, and local luminosity due to the acquisition system. This limitation significantly impacts the diagnostic process and its overall outcomes. To overcome this limitation, it is crucial to perform normalization on the acquired multispectral images as a preprocessing step for malaria parasite detection. In this study, we propose a novel method for achieving this normalization, aiming to improve the accuracy and reliability of the diagnostic process. This method is based on estimating the Bright reference image, which captures the luminosity, and the contrast variability function from the background region of the image. This is achieved through two distinct resampling methodologies, namely Gaussian random field simulation by variogram analysis and Bootstrap resampling. A method for handling the intensity saturation issue of certain pixels is also proposed, which involves outlier imputation. Both of these proposed approaches for image normalization are demonstrated to outperform existing methods for multispectral and multimodal unstained blood smear images, as measured by the Structural Similarity Index Measure (SSIM), Mean Squared Error (MSE), Zero mean Sum of Absolute Differences (ZSAD), Peak Signal to Noise Ratio (PSNR), and Absolute Mean Brightness Error (AMBE). These methods not only improve the image contrast but also preserve its spectral footprint and natural appearance more accurately. The normalization technique employing Bootstrap resampling significantly reduces the acquisition time for multimodal and multispectral images by 66%. Moreover, the processing time for Bootstrap resampling is less than 4% of the processing time required for Gaussian random field simulation.

Unsupervised Hyperspectral and Multispectral Image Blind Fusion Based on Deep Tucker Decomposition Network With Spatial-Spectral Manifold Learning.

Hyperspectral image (HSI) and multispectral image (MSI) fusion aims to generate high spectral and spatial resolution hyperspectral image (HR-HSI) by fusing high-resolution multispectral image (HR-MSI) and low-resolution hyperspectral image (LR-HSI). However, existing fusion methods encounter challenges such as unknown degradation parameters, and incomplete exploitation of the correlation between high-dimensional structures and deep image features. To overcome these issues, in this article, an unsupervised blind fusion method for LR-HSI and HR-MSI based on Tucker decomposition and spatial-spectral manifold learning (DTDNML) is proposed. We design a novel deep Tucker decomposition network that maps LR-HSI and HR-MSI into a consistent feature space, achieving reconstruction through decoders with shared parameters. To better exploit and fuse spatial-spectral features in the data, we design a core tensor fusion network (CTFN) that incorporates a spatial-spectral attention mechanism for aligning and fusing features at different scales. Furthermore, to enhance the capacity to capture global information, a Laplacian-based spatial-spectral manifold constraint is introduced in shared-decoders. Sufficient experiments have validated that this method enhances the accuracy and efficiency of hyperspectral and multispectral fusion on different remote sensing datasets. The source code is available at https://github.com/Shawn-H-Wang/DTDNML.

The phenology and water level time-series mangrove index for improved mangrove monitoring

Mangroves face decline and degradation due to human activities and natural forces, making their accurate mapping and dynamic monitoring essential. However, most of the existing mangrove indices that rely on multispectral image spectral characteristics suffer from limitations in terms of recognition accuracy and universality. Therefore, this study aimed to develop a robust and efficient Phenology and Water level Time-series Mangrove Index (PWTMI) for mangrove monitoring. PWTMI is constructed by combining spectral and temporal characteristics from dense time-series multispectral data, wherein phenology and water level time-series characteristics are extracted from NDVI and MNDWI time series. The results show that PWTMI outperforms existing multispectral-based mangrove indices and has an accuracy similar to a hyperspectral-based mangrove index, with overall accuracy ranging from 91.49% to 98.83% and F1 score ranging from 0.91 to 0.98 in four typical areas in China, indicating great potential for long time-series and large-scale mangrove monitoring.

Towards scalable wide area UAS monitoring of forest disturbance using hydrogen powered airships

ABSTRACT Uncrewed Aircraft Systems (UAS) have been recognized as powerful tools for providing monitoring data supporting mitigating and adapting to various abiotic and biotic forest disturbances. However, their suitability and commercial viability have been significantly limited by the necessity for operation primarily within visual line of sight (VLOS). Efforts are currently ongoing to establish frameworks for beyond visual line of sight (BVLOS) UAS operation. This study represents the first successful application of hydrogen-powered BVLOS airship UAS technology for monitoring forest disturbances. The test area covered 1.3 km2 of conserved forest area in North Karelia, Finland, seriously affected by bark beetles and other disturbance agents. Monitoring flights were conducted in spring, summer, and autumn in BVLOS setting operated from a command centre situated approximately 75 km distance from the test area. Deep learning models were applied to study the tree health in the disturbance area, with the special focus on examining the scalability of machine learning models. Multispectral imagery outperformed RGB imagery in classifying trees as healthy, infested, and dead spruce trees, as well as non-spruce tree classes. With multispectral imagery, the F1-scores were 0.88–0.94, 0.80, 0.92–0.96, and 0.75–0.82 for healthy, infested, and dead spruce trees, and non-spruces, respectively. Transfer learning and fine-tuning, which adapted previously trained models to new study areas, improved results compared to training models from scratch and thus contributed to the development of scalable UAS remote sensing methods in mitigating forest disturbances on a larger scale.

Multitemporal UAV study of phenolic compounds in slash pine canopies

Phenolic compounds (PC) are important secondary metabolites in plants, playing a crucial role in plant defense mechanisms against pathogens and other plants. Monitoring PC levels is important for understanding tree stress and implementing effective breeding programs. However, traditional methods for monitoring PC are time-consuming, prone to altering the phenolic composition, and mostly applicable only on a small scale. In this study, we evaluated the performance of Unoccupied Aerial Vehicles (UAV) multispectral imaging in estimating the canopy phenolic content in slash pine over an 11-month period in 2021 and a seven-month period in 2022. Three machine learning models including Partial least squares regression (PLSR), Random forest (RF) and Support Vector Machine (SVM) were compared to determine the optimal predictive model for canopy PC. The RF model provided the best predictive results, with R2 values of 0.82 for the validation set and 0.94 for the calibration set. Additionally, the study assesses the heritable variation in canopy PC over time, with the monthly heritability (h2) of PC ranging from 0 to 0.26 in 2021 and from 0 to 0.35 in 2022; The highest h2 levels were observed in July and September 2021and July 2022. The findings demonstrate significant genetic control over the variation of PC. Furthermore, we observed higher breeding values and genetic gains in July and November, which further supports the strong correlation between PC and environmental factors such as temperature and light intensity. To the best of our knowledge, this is the first study to employ time-series UAV multispectral imaging to predict secondary metabolites in pine trees and estimate their genetic variation over time. As a proof of concept, these findings provide more reliable information for tree breeding programs, ultimately enhancing their overall performance.

A Comprehensive Compilation of Spectral Libraries for Petroleum Hydrocarbons (PHC) Encompassing VNIR-SWIR-TIR Ranges

Remote detection and mapping of surface materials using optical sensors relies predominantly on analyzing multispectral and hyperspectral imagery employing classification algorithms. The classification process involves comparing the spectra of individual pixels within the image to spectra from reference databases, commonly referred to as spectral libraries. Here, we introduce a comprehensive compilation of spectral libraries specifically tailored for petroleum hydrocarbons (PHC), meticulously crafted under controlled laboratory conditions. This compilation includes reference spectral libraries for various PHC forms, including crude oils, mineral substrate-PHC mixtures (comprising crude oils and fuels), oil-film on water, and oil-water emulsions. Data collection was conducted within the visible, near, and shortwave IR (VNIR-SWIR – 0.35–2.5 µm) spectra and thermal IR (TIR – 3–15 µm) range. The openly accessible spectral libraries presented herein support the scientific community and industry in characterizing field samples or spectral data from onshore and offshore sites. Furthermore, these libraries are instrumental in developing and applying classification algorithms designed for processing spectral images captured by cameras coupled to multiple platforms (e.g., tripods, drones, airborne, orbital satellites).

Aerosol optical depth retrieval using scaled digital number (DN) values of multi-spectral satellite and a generating adversarial model based on deep learning application

ABSTRACT The current quantitative retrieval of Aerosol Optical Depth (AOD) typically uses Top-of-atmosphere (TOA) reflectance data obtained by the scale and offset terms of radiometric calibration. Errors can be introduced during the conversion of Digital Number (DN) values to TOA reflectance, restraining the accurate retrieval of AOD. Particularly when surface reflectance is high, conversion errors can significantly distort the retrieval of Aerosol Optical Depth (AOD), given the minimal impact of aerosols on the radiation detected by satellite sensors. To counteract this, we introduce an innovative retrieval algorithm that harnesses a Generative Adversarial Network (GANN), a deep learning (DL) model, to accurately derive AOD from the scaled digital number (DN) values of multi-spectral satellite imagery. The DL technology enables the use of scaled DN values for AOD retrieval, bypassing the conversion errors associated with TOA reflectance since it relies on a sample learning approach instead of radiative transfer calculations. To ensure the number and representativeness of training samples, a total of 37,103 samples from different underlying surfaces from 2014 to 2018 were obtained using a space-time matching strategy. Our K-cross validation demonstrates that employing DN values for AOD retrieval promises to yield higher accuracy compared to using TOA reflectance. Independent 2019 data were used to estimate GANN, MCD19A2 and MOD04_3K aerosol products. Our model showed superior performance compared to the other products in terms of a higher correlation (R = 0.8184) with AERONET AOD and obtaining more reliable retrievals (7601). The sensitivity experiment suggests that GANN exhibits limitations in areas characterized by high aerosol loads and heterogeneity. Our study can give a novel insight into AOD retrieval based on multi-spectral satellite and DL models.

Unmanned aerial vehicle-based aerial survey of mines in Shanxi Province based on image data

Abstract Accurately monitoring the change of mine area in the mining process is beneficial to mine safety management. This paper briefly introduces the collection of remote sensing images by unmanned aerial vehicles (UAVs) and its application in measuring surface mining subsidence and surrounding vegetation in the mining area. A case study was carried out in some mining areas of Nanshan Mountain, Yuncheng City, Shanxi Province. The surface mining subsidence value and vegetation-related parameters were measured by comparing the digital elevation model and multi-spectral images collected on May 12 and June 12, 2023. The validity experiment verified that the UAV image data could be used to measure the mining subsidence and vegetation parameters. Moreover, it was found that mining underground coal could lead to significant ground subsidence and pollute the surrounding environment, reducing vegetation. The innovation of this article lies in using UAV-collected remote sensing images instead of manually collecting ground elevation data and vegetation distribution data, providing effective references for safe mining in mining areas.

Colorectal cancer histopathology image analysis: A comparative study of prognostic values of automatically extracted morphometric nuclear features in multispectral and red-blue-green imagery.

Multispectral imaging (MSI) has been utilized to predict the prognosis of colorectal cancer (CRC) patients, however, our understanding of the prognostic value of nuclear morphological parameters of bright-field MSI in CRC is still limited. This study was designed to compare the efficiency of MSI and standard red-green-blue (RGB) images in predicting the prognosis of CRC. We compared the efficiency of MS and conventional RGB images on the quantitative assessment of hematoxylin-eosin (HE) stained histopathology images. A pipeline was developed using a pixel-wise support vector machine (SVM) classifier for gland-stroma segmentation, and a marker-controlled watershed algorithm was used for nuclei segmentation. The correlation between extracted morphological parameters and the five-year disease-free survival (5-DFS) was analyzed. Forty-seven nuclear morphological parameters were extracted in total. Based on Kaplan-Meier analysis, eight features derived from MS images and seven featured derived from RGB images were significantly associated with 5-DFS, respectively. Compared with RGB images, MSI showed higher accuracy, precision, and Dice index in nuclei segmentation. Multivariate analysis indicated that both integrated parameters 1 (factors negatively correlated with CRC prognosis including nuclear number, circularity, eccentricity, major axis length) and 2 (factors positively correlated with CRC prognosis including nuclear average area, area perimeter, total area/total perimeter ratio, average area/perimeter ratio) in MS images were independent prognostic factors of 5-DFS, in contrast with only integrated parameter 1 (P<0.001) in RGB images. More importantly, the quantification of HE-stained MS images displayed higher accuracy in predicting 5-DFS compared with RGB images (76.9% vs 70.9%). Quantitative evaluation of HE-stained MS images could yield more information and better predictive performance for CRC prognosis than conventional RGB images, thereby contributing to precision oncology.

Interpretable Model-Driven Deep Network for Hyperspectral, Multispectral, and Panchromatic Image Fusion.

Simultaneously fusing hyperspectral (HS), multispectral (MS), and panchromatic (PAN) images brings a new paradigm to generate a high-resolution HS (HRHS) image. In this study, we propose an interpretable model-driven deep network for HS, MS, and PAN image fusion, called HMPNet. We first propose a new fusion model that utilizes a deep before describing the complicated relationship between the HRHS and PAN images owing to their large resolution difference. Consequently, the difficulty of traditional model-based approaches in designing suitable hand-crafted priors can be alleviated because this deep prior is learned from data. We further solve the optimization problem of this fusion model based on the proximal gradient descent (PGD) algorithm, achieved by a series of iterative steps. By unrolling these iterative steps into several network modules, we finally obtain the HMPNet. Therefore, all parameters besides the deep prior are learned in the deep network, simplifying the selection of optimal parameters in the fusion and achieving a favorable equilibrium between the spatial and spectral qualities. Meanwhile, all modules contained in the HMPNet have explainable physical meanings, which can improve its generalization capability. In the experiment, we exhibit the advantages of the HMPNet over other state-of-the-art methods from the aspects of visual comparison and quantitative analysis, where a series of simulated as well as real datasets are utilized for validation.

Classifying early apple scab infections in multispectral imagery using convolutional neural networks

Multispectral imaging systems combined with deep learning classification models can be cost-effective tools for the early detection of apple scab (Venturia inaequalis) disease in commercial orchards. Near-infrared (NIR) imagery can display apple scab symptoms earlier and at a greater severity than visible-spectrum (RGB) imagery. Early apple scab diagnosis based on NIR imagery may be automated using deep learning convolutional neural networks (CNNs). CNN models have previously been used to classify a range of apple diseases accurately but have primarily focused on identifying late-stage rather than early-stage detection. This study fine-tunes CNN models to classify apple scab symptoms as they progress from the early to late stages of infection using a novel multispectral (RGB-NIR) time series created especially for this purpose.This novel multispectral dataset was used in conjunction with a large Apple Disease Identification (ADID) dataset created from publicly available, pre-existing disease datasets. This ADID dataset contained 29,000 images of infection symptoms across six disease classes. Two CNN models, the lightweight MobileNetV2 and heavyweight EfficientNetV2L, were fine-tuned and used to classify each disease class in a testing dataset, with performance assessed through metrics derived from confusion matrices. The models achieved scab-prediction accuracies of 97.13 % and 97.57 % for MobileNetV2 and EfficientNetV2L, respectively, on the secondary data but only achieved accuracies of 74.12 % and 78.91 % when applied to the multispectral dataset in isolation. These lower performance scores were attributed to a higher proportion of false-positive scab predictions in the multispectral dataset. Time series analyses revealed that both models could classify apple scab infections earlier than the manual classification techniques, leading to more false-positive assessments, and could accurately distinguish between healthy and infected samples up to 7 days post-inoculation in NIR imagery.

Automatic pine wilt disease detection based on improved YOLOv8 UAV multispectral imagery

The pine wilt disease (PWD) can cause destructive death to pine trees in a short period. Utilizing unmanned aerial vehicle (UAV) remote sensing technology to promptly identify PWD-infected trees has become an effective and feasible method for precise PWD monitoring. In this study, UAV multispectral imagery was used to analyze the sensitive spectral bands and different vegetation indices for PWD discriminability. A dataset of optimal spectral combinations from visible light and multispectral images was constructed, along with an improved YOLOv8 deep learning model for rapid and accurate identification of PWD-infected trees. The improved YOLOv8 model used omni-dimensional dynamic convolution (ODConv) to enhance the performance of convolutional networks, designed a dynamic head (DyHead) module to capture PWD features more accurately, and applied MPDioU to improve the regression accuracy and model runtime efficiency. Experimental results showed that the mAP@0.5 of the improved YOLOv8 model increased to 89.1 %, with a user accuracy of 90 % and a recall rate of 93.1 %. This achieved rapid and accurate detection of PWD-infected trees, providing effective technical support for automatic identification of PWD epidemic areas and control of PWD outbreaks based on UAV multispectral imagery.

Middle-Level Fusion YOLO on Multispectral Image to Detect Unhealthy Oil Palm Trees

Abstract Locating the unhealthy oil palm trees is one of the essential things in precision agriculture. In several countries such as Indonesia, Malaysia, and Thailand, Ganoderma disease often attacks oil palm trees. Detection techniques using leaf sample pieces, the possibility of biological changes in the leaf, and the collection process are too tricky. Detection techniques using image samples captured by the drone can be more accessible, but they need to provide complete information related to plant vegetation. This research proposes detecting unhealthy oil palms by capturing top-view images using drone. The drone has dual cameras (RGB and OCN bands) to get more information on plant vegetation. Middle-level fusion YOLO is used to recognize the unhealthy oil palm trees. The data was collected at the Oil Palm Plantation in Bogor, which contains over 100 unhealthy objects. Using multispectral images can improve performance compared to using only RGB images. The proposed method provides better performance than using only RGB images and OCN images with mAP (mean Average Precision) is 0.919. The proposed method provides better performance in detecting unhealthy oil palm trees.