Image Data Sources Research Articles

Monitoring of Vegetation Drought Index in Laibin City Based on Landsat Multispectral Remote Sensing Data

Due to the impact of global warming, drought has caused serious damage to China’s ecological environment and social status. This article selects Laibin City in the Guangxi Zhuang Autonomous Region as the research area, utilizing multispectral remote sensing data as the data source and Landsat series image data for relevant preprocessing. It calculates the monthly normalized vegetation index (NDVI) and surface temperature (LST) data for Laibin City. Based on the ecological environment and surface coverage conditions of the research area, the ratio vegetation index (RVI), normalized vegetation moisture index (NDWI), temperature vegetation drought index (TVDI), and conditional vegetation temperature drought index (VTCI) were selected to calculate and invert the drought monitoring results of Laibin City. The drought monitoring results were obtained and overlaid with the vegetation area map to generate the vegetation drought monitoring results of Laibin City. Based on the climate, geography, and ecological characteristics of the monitored area in Laibin City, a specific analysis will be conducted to develop an appropriate TVDI index drought level, and generate vegetation drought level result maps for Laibin City in 2021, 2022, and 2023. Then, a detailed analysis of the vegetation drought situation in Laibin City is conducted according to time and space. Among them, in the past three years, the vegetation areas in Laibin City have experienced drought seasons mostly in summer and autumn. The interannual drought is mainly mild drought, and the proportion of areas with mild drought shows a relatively stable trend. In conclusion, TVDI proves to be a valuable tool for monitoring vegetation drought in Laibin City, offering insights for efficient water resource management strategies.

Towards real-time monitoring of insect species populations

Insect biodiversity and abundance are in global decline, potentially leading to a crisis with profound ecological and economic consequences. Methods and technologies to monitor insect species to aid in preservation efforts are rapidly being developed yet their adoption has been slow and focused on specific use cases. We propose a computer vision model that works towards multi-objective insect species identification in real-time and on a large scale. We leverage an image data source with 16 million instances and a recent improvement in the YOLO computer vision architecture to present a quick and open-access method to develop visual AI models to monitor insect species across climatic regions.

Non-destructive classification of sturgeon stress using cross-modal data fusion and multi-input deep learning models

In the aquaculture phase, ensuring the safe transportation of sturgeon is crucial. The stress levels experienced during transit directly impact the fish quality and the economic returns for farmers. To address this, distributors enlist fishery farming experts to evaluate sturgeon stress. Our investigation identified three critical parameters for grading: sturgeon physiological, environmental, and visual characteristics. This study aims to develop a Cross-Modal Stress Classification Network (CM-SCN) model. It integrates information from three sensing systems to assess sturgeon stress levels. The model is built upon the architectures of AlexNet and ANN, skillfully combining both image and non-image data sources. This integration enables the model to effectively categorize sturgeon stress into four classes: A, B, C, D (representing mild, minor, moderate, and severe stress levels). The results demonstrated the model's high performance with an accuracy of 88.96%, precision of 90.06%, recall of 89.43%, and an F1 score of 89.49%. Notably, the CM-SCN model surpassed both the unimodal visual stress model and the bimodal physiological-environmental stress model. This study introduces an efficient and dependable method for monitoring sturgeon health, offering promising advancements in the field.

Machine Learning Approaches for the Image-Based Identification of Surgical Wound Infections: Scoping Review.

Surgical site infections (SSIs) occur frequently and impact patients and health care systems. Remote surveillance of surgical wounds is currently limited by the need for manual assessment by clinicians. Machine learning (ML)-based methods have recently been used to address various aspects of the postoperative wound healing process and may be used to improve the scalability and cost-effectiveness of remote surgical wound assessment. The objective of this review was to provide an overview of the ML methods that have been used to identify surgical wound infections from images. We conducted a scoping review of ML approaches for visual detection of SSIs following the JBI (Joanna Briggs Institute) methodology. Reports of participants in any postoperative context focusing on identification of surgical wound infections were included. Studies that did not address SSI identification, surgical wounds, or did not use image or video data were excluded. We searched MEDLINE, Embase, CINAHL, CENTRAL, Web of Science Core Collection, IEEE Xplore, Compendex, and arXiv for relevant studies in November 2022. The records retrieved were double screened for eligibility. A data extraction tool was used to chart the relevant data, which was described narratively and presented using tables. Employment of TRIPOD (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis) guidelines was evaluated and PROBAST (Prediction Model Risk of Bias Assessment Tool) was used to assess risk of bias (RoB). In total, 10 of the 715 unique records screened met the eligibility criteria. In these studies, the clinical contexts and surgical procedures were diverse. All papers developed diagnostic models, though none performed external validation. Both traditional ML and deep learning methods were used to identify SSIs from mostly color images, and the volume of images used ranged from under 50 to thousands. Further, 10 TRIPOD items were reported in at least 4 studies, though 15 items were reported in fewer than 4 studies. PROBAST assessment led to 9 studies being identified as having an overall high RoB, with 1 study having overall unclear RoB. Research on the image-based identification of surgical wound infections using ML remains novel, and there is a need for standardized reporting. Limitations related to variability in image capture, model building, and data sources should be addressed in the future.

Most endovascular thrombectomy patients have Target Mismatch despite absence of formal CT perfusion selection criteria.

Endovascular thrombectomy (EVT) is the standard of care for large vessel occlusion stroke. Use of Computed Tomographic Perfusion (CTP) to select EVT candidates is variable. The frequency of treatment and outcome in patients with unfavourable CTP patterns is unknown. A retrospective analysis of CTP utilisation prior to EVT was conducted. All CTP data were analysed centrally and a Target Mismatch was defined as an infarct core ≤70 ml, penumbral volume ≥15ml, and a total hypoperfused volume:core volume ratio >1.8. The primary outcome was good functional outcome at 90 days, defined as a modified Rankin Scale (mRS) score 0-2. follow-up infarct volume, core expansion and penumbral salvage volumes were secondary outcomes. Of 572 anterior circulation EVT patients, CTP source image data required to generate objective maps were available in 170, and a Target Mismatch was present in 151 (89%). The rate of 90-day good functional outcome was similar between Target Mismatch (53%) and Large Core Non-Mismatch groups (46%, p = 0.629). Median follow-up infarct volume in the Large Core Non-Mismatch group (104ml [IQR 25ml-189ml]) was larger than that in the Target Mismatch patients (16ml [8ml-47ml], p<0.001). Despite a lack of formal CTP selection criteria, the majority of patients treated at our centres had a Target Mismatch. Patients without Target Mismatch had larger follow-up infarct volumes, but the functional recovery rate was similar to that in Target Mismatch patients. Infarct volumes should be included as objective assessment criteria in the evaluation of the efficacy of EVT in non-Target Mismatch patients.

OPTIMIZING THE IMPLEMENTATION OF THE YOLO AND DATA ALGORITHM AUGMENTATION IN HANACARAKA JAVANESE SCRIPT LANGUAGE CLASSIFICATION

The Javanese language and Javanese script are one of the rich cultural heritages in Indonesia, but there are still many people who do not have sufficient understanding of the Javanese language and culture, including the Javanese script. This can have an impact on the loss of language and cultural diversity, as well as the loss of the local identity of the Javanese people. The lack of knowledge about Javanese script language culture can be caused by various factors, such as the lack of access to adequate learning resources, the minimum use of Javanese language and script in daily life, and the lack of attention from the government and society towards the preservation of Javanese language and culture. This study aims to optimize the application of the YOLO (You Only Look Once) algorithm and data augmentation techniques in the Javanese language classification Hanacaraka script. The method used in this study was collecting data on Javanese Hanacaraka script images, data labeling, data augmentation, and model training using the YOLO algorithm. The results showed that the Javanese script pattern recognition method used the YOLO algorithm which had gone through the data augmentation process, showing good results with an accuracy of 96.4% using 3021 image data sources for the hanacaraka letters.

Modeling the impact of land use/land cover change on soil erosion: in Suluh River Basin, Northern Ethiopia

&lt;p&gt;&lt;span lang="EN-US"&gt;The &lt;em&gt;Suluh&lt;/em&gt; river basin is subjected to soil erosion due to land use and land cover change. Yet, the impact of land use/land cover change soil erosion has not been applied in the study area. Thus, the current study focused on the modeling of the impact of land use/land cover changes on soil erosion in the &lt;em&gt;Suluh&lt;/em&gt; river basin, the northern highland part of Ethiopia. Landsat image data sources were used to achieve the objectives. Ancillary data was also used. The nearest neighbor's fuzzy way of classification and the land change modeler for modeling, and the revised universal soil loss equation model for estimating the soil erosion rate were employed. Both qualitative and quantitative data were analyzed qualitatively and quantitatively. The study's findings confirmed that cultivated land, bare land, and built-up areas significantly increased while areas occupied by natural vegetation such as forest land, shrub lands, and grazing lands decreased at a rapid rate in between 1990 to 2018. The predicted results suggest the continuation of the trend up to 2048 if business as usual is continued. The annual mean of soil lost in the study area was about 36.31, 43.32, and 47.78 in the years 1990, 2002, 2018, and will be 56.54, and 71.62 tons per hectare per year in 2028, and 2048, respectively. When we consider 15 t ha&lt;sup&gt;−1&lt;/sup&gt; year&lt;sup&gt;−1&lt;/sup&gt; as the maximum tolerable soil loss (TSL) rate for ease of presentation, the areas above the TSL have increased from 88.3% in 1990 to 88.6% in 2002 and to 89.6% in 2018, and are expected to increase to 89.9% and 99.8% in 2028 and 2048 periods, respectively. Thus, land use and land cover change information and its impact on soil erosion should be taken under consideration by land use planners to apply sustainable land management activities in the &lt;em&gt;Suluh&lt;/em&gt; river basin.&lt;/span&gt;&lt;/p&gt;

Analysis of reconstructed multisource and multiscale 3-D digital rocks based on the cycle-consistent generative adversarial network method

SUMMARY Digital rock physics (DRP) is important for characterizing the pore characteristics and petrophysical parameters of rocks from a microscopic perspective. Among the digital rock reconstruction methods, the most widely used is the computer tomography (CT) scan method. However, the pore structure of rocks has multiscale features, and CT scan images with a single resolution cannot completely describe the pore structure characteristics of rocks. In this paper, we analysed reconstructed multisource and multiscale 3-D digital rocks based on the cycle-consistent generative adversarial network (CycleGAN) method. This method fully integrates the advantages of the large field of view of low-resolution images and the high-precision features of high-resolution images. To investigate the performance of the method, two sequences of CT scan images of a sandstone (sample A) and a carbonate rock (sample B) collected from oilfields were studied. Moreover, due to the high resolution of scanning electron microscope (SEM) images, we also constructed 3-D digital rocks with different resolutions of the Berea sandstone based on the combination of CT scan images and SEM images. Finally, the statistical properties and absolute permeabilities were calculated to evaluate the accuracies of reconstructed multisource and multiscale 3-D digital rocks. The results show that the reconstructed multiscale digital rocks based on the CycleGAN method have good accuracy in terms of statistical properties and petrophysical properties. Considering the computational cost and computational accuracy, for high- and low-resolution CT scan images and CT scan images with SEM images as training image data sources, we suggest that the resolution of the reconstructed multiscale digital cores is 4–8 times and 4–16 times higher than that of the low-resolution CT scan images, respectively. The findings of our research will be helpful in gaining insight into the petrophysical properties of heterogeneous rocks.

ABT-GAMNet: A novel adaptive Boundary-aware transformer with Gated attention mechanism for automated skin lesion segmentation

Segmentation of skin lesions is the hotspot development process in the medical field. As skin cancer becomes most disastrous for humans, it requires framing an effective model for identifying the lesion region over the skin. As represented in the image format, implementing the decision-based system is quite challenging. Owing to the noise and artifacts in images, it becomes cumbersome to segment the exact lesion regions over the skin images. In recent years, the deep learning model is used to attain the expected results. Yet, it still subsists with overfitting issues and large-scale dimensional datasets, it mitigates the segmentation results. To meet up the pre-requisite, a novel adaptive-based segmentation technique is proposed for skin lesion segmentation. In the first stage, the source images are fetched from the reputed skin image data sources. It is followed by pre-processing stage. In the second phase, the pre-processing is done in two ways: (i) filtering methods and (ii) adaptive Contrast Limited Adaptive Histogram Equalization (CLAHE). In the adaptive CLAHE model, the threshold value is optimized by proposing the new hybrid algorithm as Hybrid Rat Electric Fish Swarm Optimization (HREFSO) for enhancing the contrast level in pre-processed images. Subsequently, the pre-processed images are fed into the novel Adaptive Boundary-aware Transformer with Gated Attention Mechanism (ABT-GAM) for segmenting the lesion regions in images. To improve the segmentation accuracy, the parameters are determined optimally using HREFSO. The evaluation is performed and validated across various metrics. Hence, the outcome elucidates that it acquires a higher segmentation accuracy for diagnosing skin cancer.

Research on elaborate image simulation method for close-range space target

In order to meet the urgent needs of target image data in the testing of close-range space target situational awareness algorithm, an elaborate image simulation method that fully considers the target surface texture structure is proposed. First, the physical mechanism is analysed by constructing a radiation transmission model. Then, Phong and five-parameters BRDF models are adopted to characterize the optical properties of the material. The geometric characteristic of texture structure is also characterized through displacement mapping. Finally, the sequence images of the space target are obtained by simulation based on the designed simulation process and given imaging conditions. The CPU multi-core parallel acceleration based on the Embree engine is taken to compensate for the speed reduction caused by the increase in the number of model faces. Compared with the measured image in orbit for verification, the image simulation similarity is above 80%. The results show that the simulation method in this article can better describe the characteristics of the target in the actual on-orbit environment and meet the near real-time simulation requirements of the system, providing a more realistic image data source for the ground test of the on-orbit service technology.

Intracranial arterial flow velocity mapping in quantitative time-of-flight MR angiography using deep machine learning

PurposeTo evaluate the application of deep machine learning (DML) to 3D quantitative time-of-flight (qTOF) magnetic resonance angiography (MRA) to measure blood flow velocity within the intracranial arteries. Materials and methodsIntracranial qTOF MRA was acquired in 15 subjects at 3 T. Blood flow velocity quantitation with qTOF MRA was done using a non-DML computer-vision procedure, and using convolutional DML neural networks. 3D phase contrast (PC) MRA was used as the comparator. Using PC velocity measures as the output target and qTOF two-echo source image data as inputs, DML neural networks were trained to predict component blood flow velocities. Total velocities and peak intracranial arterial blood flow velocities were computed from component velocities. ResultsCompared to non-DML image analysis, DML-based analysis of qTOF MRA image data improved agreement with PC for mean component velocity (intraclass correlation coefficient (ICC) = 0.966 versus 0.939), mean total velocity (ICC = 0.835 versus 0.723), and peak velocity (ICC = 0.816 versus 0.597), as well as narrowed the 95% Bland-Altman limits of agreement for mean component velocity ([−5.16, +4.31]cm/s versus [−6.86, +6.53]cm/s), mean total velocity ([−6.78,+3.59]cm/s versus [−9.39, +7.09]cm/s) and peak velocity ([−13.5,+10.2]cm/s versus [−21.3, +10.2]cm/s). Compared to non-DML analysis, DML image analysis reduced the root-mean-square deviation from PC velocity measures by 28%–36%, and shortened calculation times by 35-fold. ConclusionThe application of DML image analysis to intracranial qTOF MRA for velocity quantitation markedly shortened calculation times, substantially improved the agreement of component, total, and peak arterial blood flow velocities, and provided excellent agreement of hemodynamic measures with respect to 3D PC.

Research on unbalanced mining of highway project key data based on knowledge graph and cloud model

Various stages of highway project construction process involve text, image, audio, video and other related data sources involving many participants, forming a huge amount of data. Accurately tracing the source of responsibility, refining and applying the unbalanced data in the highway project archives is of great significance for realizing the intelligent transformation of highway construction project management. This paper firstly sorts out the construction process of highway projects and the main data sources, constructs a data association network between construction entities and construction process, as well as a knowledge map of highway construction data. Then, according to the highway construction stage, an index system based on 12 key data is constructed by using the entropy weight cloud model method, and the importance of the data is evaluated. Thirdly, based on the unbalanced characteristics of highway project data, a method of mining big data in highway project archives using classification evaluation indexes is proposed, and the accuracy of this method is verified by case calculation. Finally, taking the Shizong Qiubei Expressway in China as an example, the intelligent management and control suggestions for key data of transportation projects are proposed. It is found that the key data with special importance rate in highway construction include construction data, supervision data and completion data. Boosting algorithm is more accurate than the traditional SMOTE algorithm for unbalanced data mining, which helps to save the project construction cost and improve the quality of data extraction in the project archives. This study provides a theoretical reference for key data traceability of highway project intelligent management and control platform and the improvement of intelligent management efficiency.

Multi Label Automatic Image Annotation Neural Network to Handle Multi Media Image Retrieval

The present-day business online web indexes have embraced electronic picture search to further develop precision in picture information recovery. However Re-positioning is expectedly considered as a successful cycle for deciding the situation with electronic picture web search tools, yet it experiences a lack of a couple of. Consequently, some grouping methods, particularly (Novel Image Re-positioning System) NIRS have to be proposed to carry out inquiry picture re-positioning with semantic marks in electronic picture information recovery, which naturally recovers results in view of visual semantic highlights for various question or catchphrase extensions. To get to productive picture with the annotation is an aggressive concept in present. So that in the present paper, we are going to propose the Unsupervised Multi Labeled Image Annotate Learning Approach (UMLIALA) to decrease complexity in indexing of image with mining of web related convex optimization and classify required image data from large image data sets. And also use group based approximation calculation to improve accuracy in retrieval of images from different image data sources. Experiments of proposed approach give better and efficient results when compare to traditional approaches in terms of different image exploration parameters studies on different large image data sets.

Diffractive optical elements for calibration of LIDAR systems: materials and fabrication

The promise of fully autonomous vehicles to replace the judgment of human drivers with real-time algorithmic decision-making based on optoelectronic systems relies fundamentally on the quality of the available data. Limitations imposed by sensor-resolution, available optical power, and achievable signal-to-noise ratios have been well studied in the light detection and ranging (LIDAR) application space. Additionally, the problem of integrating multiple sources of image data as well as the need to establish and maintain the system calibration over life are critically important to system reliability and safety. These latter concerns will receive even greater attention as self-driving vehicles begin to transition toward fully autonomous operation. Because of the importance of calibration to system performance and safety, the process of validating and recalibrating the system will ideally be integrated into the LIDAR system itself with calibration occurring automatically “anywhere and at any time,” without dedicated external infrastructure. Mass-market adoption is also being driven by the systems’ size and weight, as well as reliable manufacturability and resilience to environmental stresses. Due to their extreme stability, manufacturability, and small size, diffractive optical elements (DOEs) are well suited for use as optical calibration references. Current three-dimensional (3D) mapping systems based on structured light illumination already rely on DOEs as precision pattern generators to provide 3D depth sensing in a wide array mobile devices. We examine the potential use of DOEs as calibration elements in multicamera or LIDAR systems, including appropriate choices of materials, designs, and fabrication methods to ensure reliable long-term performance under automotive use conditions. We present simulations of the impact of DOE material properties on the accuracy of the generated dot patterns and consequently on the depth accuracy and lateral distortion of the 3D image. Additionally, we present requirements for DOE manufacture using conventional semiconductor fabrication technologies optimized for creating engineered surface nanostructures capable of transforming the output of a laser or other narrow-band source into a precise reference pattern.

Application and Prospect of Image Recognition in Intelligent Agriculture

Image recognition has been successfully applied in automatic driving and face recognition, and its application in agriculture can greatly promote the process of intelligent agriculture. This article firstly expounds the basic concepts of the intelligent agriculture and image recognition, and enumerates the part of the development of intelligent agriculture in China and abroad. Then, the image recognition in land classification and protection, accurate identification of animals as well as pest and disease detection are analyzed. It is concluded that the source of image data is relatively single, and it is difficult to deal with the identification problem of multi-agent and multi-variable quickly and accurately, as well as the hardware performance is defective. Finally, the prospect of the application of image recognition in intelligent agriculture is proposed, which can be used as a reference for future research in related fields.

A Random Forest Algorithm for Landsat Image Chromatic Aberration Restoration Based on GEE Cloud Platform—A Case Study of Yucatán Peninsula, Mexico

With the growth of cloud computing, the use of the Google Earth Engine (GEE) platform to conduct research on water inversion, natural disaster monitoring, and land use change using long time series of Landsat images has also gradually become mainstream. Landsat images are currently one of the most important image data sources for remote sensing inversion. As a result of changes in time and weather conditions in single-view images, varying image radiances are acquired; hence, using a monthly or annual time scale to mosaic multi-view images results in strip color variation. In this study, the NDWI and MNDWI within 50 km of the coastline of the Yucatán Peninsula from 1993 to 2021 are used as the object of study on GEE platform, and mosaic areas with chromatic aberrations are reconstructed using Landsat TOA (top of atmosphere reflectance) and SR (surface reflectance) images as the study data. The DN (digital number) values and probability distributions of the reference image and the image to be restored are classified and counted independently using the random forest algorithm, and the classification results of the reference image are mapped to the area of the image to be restored in a histogram-matching manner. MODIS and Sentinel-2 NDWI products are used for comparison and validation. The results demonstrate that the restored Landsat NDWI and MNDWI images do not exhibit obvious band chromatic aberration, and the image stacking is smoother; the Landsat TOA images provide improved results for the study of water bodies, and the correlation between the restored Landsat SR and TOA images with the Sentinel-2 data is as high as 0.5358 and 0.5269, respectively. In addition, none of the existing Landsat NDWI products in the GEE platform can effectively eliminate the chromatic aberration of image bands.

Performance Analysis of the YOLOv4 Algorithm for Pavement Damage Image Detection with Different Embedding Positions of CBAM Modules

One of the most critical tasks for pavement maintenance and road safety is the rapid and correct identification and classification of asphalt pavement damages. Nowadays, deep learning networks have become the popular method for detecting pavement cracks, and there is always a need to further improve the accuracy and precision of pavement damage recognition. An improved YOLOv4-based pavement damage detection model was proposed in this study to address the above problems. The model improves the saliency of pavement damage by introducing the convolutional block attention module (CBAM) to suppress background noise and explores the influence of the embedding position of the CBAM module in the YOLOv4 model on the detection accuracy. The K-means++ algorithm was used to optimize the anchor box parameters to improve the target detection accuracy and form a high-performance pavement crack detection model called YOLOv4-3. The training and test sets were constructed using the same image data sources, and the results showed the mAP (mean average precision) of the improved YOLOv4-3 network was 2.96% higher than that before the improvement. The experiments indicate that embedding CBAM into the Neck module and the Head module can effectively improve the detection accuracy of the YOLOv4 model.

System Analysis of the Learning Behavior Recognition System for Students in a Law Classroom: Based on the Improved SSD Behavior Recognition Algorithm

Classroom teaching activities have always been the focus of research in the field of pedagogy. The main body of classroom teaching activities is students, and students’ classroom behavior status can reflect classroom efficiency to a certain extent, making it an important reference index for classroom quality assessment. With the rapid development of artificial intelligence, school education is gradually becoming more intelligent. At present, most of the classrooms are equipped with video equipment. These videos record the real behavior status of the students in the classroom. For example, by analyzing the data, combining artificial intelligence, deep learning, and other related technologies with education to develop behavioral intelligence, the analysis system has a certain positive effect on helping the reform of classroom education. This study proposes an improved SSD behavior recognition model. The network model is optimized and the model convergence speed is accelerated based on the RMSProp optimization algorithm Through a database of 2,500 images of five behaviors, including raising hands, sitting up, writing, sleeping, and playing with mobile phones, and using them as object detection datasets, we use the OpenCV library to extract frames from classroom screen recording videos as image data sources for student behavior recognition and face recognition. Finally, an improved method is proposed to change the virtual network to MobileNet and complete the fusion function. The results show that compared with the traditional SSD method, the improved model has a significantly improved effect in recognizing small objects and the recognition speed is not significantly reduced.

Analysis of Landscape Pattern Evolution and Driving Forces Based on Land-Use Changes: A Case Study of Yilong Lake Watershed on Yunnan-Guizhou Plateau

In order to explore the landscape pattern evolution and driving forces of the Yilong Lake watershed, the combined method of supervised classification with manual visual interpretation based on the landsat5TM/8OLI remote sensing image data sources was used to establish a high-precision spatial distribution information database of the Yilong Lake watershed. Landscape index was used to analyze the distribution and spatial pattern change characteristics of various land-use types. Based on correlation and principal component analysis, we discuss the relationship between the change characteristics of land-use type, distribution and spatial pattern, and the interference of local socio-economic development and natural factors. The results show that: (1) In the past 30 years, the land-use types of the Yilong Lake watershed are mainly forest, garden plot and cultivated land. The forest area decreased significantly by 30.45 km2, of which the fastest reduction stage was from 2000 to 2005, with a total reduction of 20.56 km2. The garden plot conversion is relatively large, with a total of 181.69 km2 transferred out, of which 28.84 km2 has become unused land, respectively. (2) In the past 30 years, the maximum patch index decreased by 9.94% and the patch density index increased by 14.25%, indicating that the landscape fragmentation in the whole basin increased. The Shannon diversity index showed an increasing process; the aggregation index showed a decreasing process. (3) The change in landscape pattern in the watershedwas closely related to economic growth, population growth, social affluence and agricultural development. Natural factors, social factors and economic indicators are significantly positively correlated with patch density, edge density, landscape shape index and Shannon diversity index, and significantly negatively correlated with the largest patch index and the contagion index. On the whole, the wetlands in the basin are shrinking and the landscape diversity is changing. Reducing the excessive impact of human activities on the watershed ecosystem is a key factor for the local protection of wetland resources and the maintenance of wetland ecological functions.

Performance comparison of RGB and multispectral vegetation indices based on machine learning for estimating Hopea hainanensis SPAD values under different shade conditions.

Reasonable cultivation is an important part of the protection work of endangered species. The timely and nondestructive monitoring of chlorophyll can provide a basis for the accurate management and intelligent development of cultivation. The image analysis method has been applied in the nutrient estimation of many economic crops, but information on endangered tree species is seldom reported. Moreover, shade control, as the common seedling management measure, has a significant impact on chlorophyll, but shade levels are rarely discussed in chlorophyll estimation and are used as variables to improve model accuracy. In this study, 2-year-old seedlings of tropical and endangered Hopea hainanensis were taken as the research object, and the SPAD value was used to represent the relative chlorophyll content. Based on the performance comparison of RGB and multispectral (MS) images using different algorithms, a low-cost SPAD estimation method combined with a machine learning algorithm that is adaptable to different shade conditions was proposed. The SPAD values changed significantly at different shade levels (p < 0.01), and 50% shade in the orthographic direction was conducive to chlorophyll accumulation in seedling leaves. The coefficient of determination (R2), root mean square error (RMSE), and average absolute percent error (MAPE) were used as indicators, and the models with dummy variables or random effects of shade greatly improved the goodness of fit, allowing better adaption to monitoring under different shade conditions. Most of the RGB and MS vegetation indices (VIs) were significantly correlated with the SPAD values, but some VIs exhibited multicollinearity (variance inflation factor (VIF) > 10). Among RGB VIs, RGRI had the strongest correlation, but multiple VIs filtered by the Lasso algorithm had a stronger ability to interpret the SPAD data, and there was no multicollinearity (VIF < 10). A comparison of the use of multiple VIs to estimate SPAD indicated that Random forest (RF) had the highest fitting ability, followed by Support vector regression (SVR), linear mixed effect model (LMM), and ordinary least squares regression (OLR). In addition, the performance of MS VIs was superior to that of RGB VIs. The R2 of the optimal model reached 0.9389 for the modeling samples and 0.8013 for the test samples. These findings reinforce the effectiveness of using VIs to estimate the SPAD value of H. hainanensis under different shade conditions based on machine learning and provide a reference for the selection of image data sources.