Coral Reef Classification Research Articles

Comparison of DenseNet-121 and MobileNet for Coral Reef Classification

Coral reefs are a type of marine organism that has beauty and benefits for other sea creatures’ ecosystems. However, despite its beauty and usefulness, coral reefs are vulnerable to damage such as coral bleaching, which can impact other coral reef ecosystems. This research aims to classify digital images of healthy, bleached, and dead coral reefs. This research method is DenseNet-121 and MobileNet is based on Convolutional Neural Networks. This research uses a dataset from 1582 coral reef image data with three main classes: 720 were bleached, 150 were dead, and 712 were healthy. The testing process is carried out using several forms of split datasets, namely 60:10:30, 50:10:40, and 70:10:20. The test results obtained with a data sharing percentage of 60:10:30 show that MobileNet architecture achieved 88.00% accuracy, and DenseNet-121 achieved 91.57% accuracy. Using a data split percentage of 50:10:40, MobileNet achieved 84.51% accuracy, and DenseNet- 121 achieved 90.52% accuracy. Meanwhile, with a data separation percentage of 70:10:20, MobileNet achieved 85.48% accuracy, and DenseNet-121 achieved 92.74% accuracy.

A Novel Coral Reef Classification Method Combining Radiative Transfer Model With Deep Learning

A Novel Coral Reef Classification Method Combining Radiative Transfer Model With Deep Learning

The Impact of Image Enhancement and Transfer Learning Techniques on Marine Habitat Mapping

Marine habitat mapping is primarily done to monitor and preserve underwater ecosystems. Images captured in a marine environment suffer from acidification, pollutions, waste chemicals, and lighting conditions. Human beings are progressing fast in terms of technology and are also responsible for the degradations of ecosystems, both marine and land habitats. Marine biologists possess a lot of data for the underwater environment, but it is hard to analyze, and the task becomes tiresome. Automating this process would help marine biologists quickly monitor the environment and preserve it. Our research focuses on coral reef classification and two critical aspects, i.e., Image enhancement and recognition of coral reefs. Image enhancement plays an essential role in marine habitat mapping because of the environment in which images are taken. The literature contains many image enhancement techniques for underwater. The authors want to determine whether a single image enhancement technique is suitable for coral reefs. Four image enhancement techniques based on an extensive literature review are selected. We have used DenseNet-169 and MobileNet for image classification. It has been reported that DenseNet-169 has excellent results for coral reefs classification. Histogram techniques combined with DenseNet-169 for classification resulted in higher classification rates.

Applying Transfer Learning ResNet-50 for Tracking and Classification of A Coral Reef in Development The Mobile Application with Scrum Framework

A Coral Reef has an important role in keeping a balanced ecosystem in the ocean. This research aims to develop a mobile application to track and classify to identify Coral Reefs with a presentation of information names of species, the type of classification of coral (hard or soft coral), accuracy, and area distribution that has been recorded before in certain areas. This research, utilized Flutter for deploying a mobile application, Deep Learning of Transfer Learning (TL) ResNet-50 for the classification of Coral Reefs, and the Google Cloud Platform used to deploy Rest API and deploy model TL ResNet-50 has been created, then the MongoDB database for storing data of the Coral Reef. A species of Coral Reef that could be classified to identify i.e. Acropora Clathrata, Acropora Florida, Cyphastrea Microphthalma, Diploastrea Heliopora, and Pachyseris Speciosa. A total dataset of a Coral Reef has 2197 images and is trained with TL ResNet-50 and it has resulted in an accuracy of 0.9594 or 96% for the classification of Corel Reef. We used the Scrum framework to produce a product in a short time and assist to manage process application development.

Calculating the Effectiveness of RCNN-based Coral Reef Classification

Calculating the Effectiveness of RCNN-based Coral Reef Classification

Aplikasi Teknologi Drone Sebagai Pelengkap Data Survei Lapang Untuk Pemetaan Ekosistem Terumbu Karang Menggunakan Citra Worldview-2

Coral reefs are one of the important ecosystems that need data inventory in order to manage marine resources both spatially and temporally. Remote sensing with drone technology is one of the vehicles for collecting data related to coral reef ecosystems in real time. Along with the times, drones are widely used in research both in terrestrial and in coastal areas. This research was conducted on December 4 – 10, 2018 in the Karang Lebar, Seribu Islands, Jakarta. The purpose of this research was to explored the ability of drones as a vehicle to complement field survey data for coral reef mapping. In addition, conducting a pixel-based analysis for coral reef classification based on WorldView-2 satellite imagery using field survey data and additional data from drone data. Based on the results of the study, information was obtained that drone imagery can help visual interpretation in detecting coral reefs. Thus, field survey data that are not covered in the research area can be assisted by the use of drones as a complementary vehicle for field survey data. The coral reef classification process using WorldView-2 imagery can be mapped well. Drone imagery has proven to be an alternative in field data collection as a training area in identifying coral reef objects, such as with massive coral lifeforms that can be clearly seen from drone images. It's just that visual interpretation is limited because of the water column, making it difficult to identify objects.

Using ensemble methods to improve the robustness of deep learning for image classification in marine environments

Abstract Deep learning has been shown to be effective for classification of coral in benthic imagery, and is becoming a tool for use in many monitoring programs around the world. Although deep learning accuracy for coral reef classification has been well published for studies where validation metrics are generated from data within surveys, little research has been done on the transferability and generalisability of deep learned models to data never seen by the trained model. Examples include data across multiple capture methods, camera systems, habitat types, water quality conditions and temporal studies. In this paper we investigate the use of deep ensembling to measure the reliability of predictions in new or unseen environments. In this paper we show that ensemble methods are more stable in their calibration across dataset shifts compared with other approaches. Ensembles show more robust uncertainty quantification in unseen environments compared to alternative methods, thus providing more confidence in the use of pre‐trained models in unconstrained environments. These results show that ensembles should be the de facto standard for any practitioner using deep learning for benthic image automation, and applying deep learning approaches to coral classification.

Assessing optimal UAV-data pre-processing workflows for quality ortho-image generation to support coral reef mapping

Since ortho-image constructed from unmanned aerial vehicles (UAV) acquired images has significant misalignment effect, an optimized and precise workflow (WF) is proposed in this study. Based on the parameters of photo alignment, sparse point cloud model, blending and seamline refinement, the 24 combinations of ortho-image producing methods were tested. The optimal WF is evaluated from the aspects of coral mapping accuracy, geometric fidelity, completeness, and efficiency. Statistical error analysis shows that blending and seamline refinement are the most relevant WF components that influence accuracy of orthorectification and consequently coral reef classification. The optimal WF found to be when ‘highest’ photo alignment, with ‘high’ tie points in the sparse cloud model are applied, in presence of blending and seamline refinements and can achieve 87.9% overall mapping accuracy. With the available photogrammetric software packages, the proposed WF can be used in mosaicking and mapping large scale macroalgae, seagrass and seaweed.

Coral Reef Classification Using Improved WLD Feature Extraction with Convolution Neural Network Classification

Background: Coral reefs are one of the most oldest and dynamic ecosystems of the world. Manual annotation of coral reef is not possible due to lacking consistency and objectivity in human labeling. Objective: Manual annotation consumes an enormous number of hours to annotate every coral image and video frames as well as human resource. An emblematic survey states that more than 400 person hours are required to annotate 1000 images. Incidentally, some coral species have different shapes, sizes and colors, most of the corals seem indistinguishable to the human eye. In order to avoid the contradictory classifications, an expert system that can automatically annotate the corals is essential to improve the accuracy of classification. Method: The proposed improved WLD extract texture features from six combinations of color channels like (1) R Channel, (2) G Channel (3) B Channel (4) RG Channel (5) GB Channel and (6) BR Channel of an image in a holistic way while preserving their relations. The extracted features are analyzed, and classified using CNN Classifier. Results: Experiments are carried out with EILAT, RSMAS, EILAT 2 and MLC2008 datasets and the proposed improved WLD based coral reef classification is found to be appropriate. From the accuracy point of view, the improved WLD demonstrate higher accuracy compared to other state-of-the-art techniques. Conclusion: This paper analyzes the role of Improved WLD for feature extraction to classify coral reefs. For this purpose EILAT, RSMAS, EILAT 2 and MLC2008 datasets have been used. It is observed that the proposed IWLD based classifier gives promising results for coral reef classification.

Statistical Modeling based Directional Pattern Design (SMDPD) feature extraction for coral reef classification.

Coral reefs are a precious ecosystem that supports majority of marine life. The identification of coral species is essential in the conservation and monitoring process. Distinguishing the coral species among the coral reef family is really a challenging task since they have analogous characteristics and have complex spatial borders between the coral classes. This requires experts to identify corals. But due to inconsistency and biasing of manual labelling, the manual annotations of coral reefs are not feasible. The objective of this research work is to identify various types of corals present in the given input video. This work is aimed at identifying thirty-six types of coral by employing a new feature extraction method called Statistical Modeling based Directional Pattern Design (SMDPD) using a new directional pattern. The proposed work outperforms the state-of-art techniques for four coral datasets namely EILAT, EILAT 2 Red Sea, MLC 2010 and RSMAS. Another advantage of this work is the reduction in feature bin size from 255 to just 16 bins.

Reef Cover, a coral reef classification for global habitat mapping from remote sensing

Coral reef management and conservation stand to benefit from improved high-resolution global mapping. Yet classifications underpinning large-scale reef mapping to date are typically poorly defined, not shared or region-specific, limiting end-users’ ability to interpret outputs. Here we present Reef Cover, a coral reef geomorphic zone classification, developed to support both producers and end-users of global-scale coral reef habitat maps, in a transparent and version-based framework. Scalable classes were created by focusing on attributes that can be observed remotely, but whose membership rules also reflect deep knowledge of reef form and functioning. Bridging the divide between earth observation data and geo-ecological knowledge of reefs, Reef Cover maximises the trade-off between applicability at global scales, and relevance and accuracy at local scales. Two case studies demonstrate application of the Reef Cover classification scheme and its scientific and conservation benefits: 1) detailed mapping of the Cairns Management Region of the Great Barrier Reef to support management and 2) mapping of the Caroline and Mariana Island chains in the Pacific for conservation purposes.

Status of health and conservation classification of tropical coral reefs in Lakshadweep archipelago

Coral reefs of Lakshadweep perform a range of vital ecosystem functions and sustain the livelihoods of island communities. Coral reefs provide ecosystem services that are vital to human societies and industries through fisheries, coastal protection and tourism. The present study was undertaken during September and October 2015 to assess the status of reef health in 10 inhabited Islands of Lakshadweep following reef classification to imply conservation measures and also for the sustainable management by the dependent communities. Live coral cover, habitat complexity index, coral species richness, reef condition index, mortality index, morphological dominance, morphological diversity and coral conservation class were used to get an aggregate value of ‘composite condition’ index for determining the overall conservation values of coral reefs in these Islands. Kadmat reef had the highest live coral cover of 64.50% among the 10 inhabited islands. Bitra and Kiltan reef had good live coral cover of 62.75 and 52.25% respectively. The major benthic component in all the studied islands is live coral except in Amini and Andrott Islands. The recovery of corals has increased significantly in the last 5 years in Kadmat, Bitra, Kiltan and Bangarm after the 2010 mass bleaching event. Among the eight attributes, significant positive correlations were found between coral reef condition and live coral cover (r = 0.93, p 60% of stress tolerators. CC3 was assigned to Bangaram and Kadmat due to the presence of > 75% of ruderals. Mixed coral morphology was recorded in Agatti, Andrott, Kalpeni, Kavaratti, and Kiltan with approximately equal proportions of ruderals, competitors and stress tolerators, and these coral reefs were designated as CC4. Protection and conservation management measures, as well as sustainable resource use, have been suggested based on the live coral cover, conservation class, and coral reef health status.

3D Classification of Cold-Water Coral Reefs: A Comparison of Classification Techniques for 3D Reconstructions of Cold-Water Coral Reefs and Seabed

Cold-water coral (CWC) reefs are complex structural habitats that are considered biodiversity “hotspots” in deep-sea environments and are subject to several climate and anthropogenic threats. As three-dimensional structural habitats, there is a need for robust and accessible technologies to enable more accurate reef assessments. Photogrammetry derived from remotely operated vehicle video data is an effective and non-destructive method that creates high-resolution reconstructions of CWC habitats. Here, three classification workflows [Multiscale Geometrical Classification (MGC), Colour and Geometrical Classification (CGC) and Object-Based Image Classification(OBIA)] are presented and applied to photogrammetric reconstructions of CWC habitats in the Porcupine Bank Canyon, NE Atlantic. In total, six point clouds, orthomosaics, and digital elevation models, generated from structure-from-motion photogrammetry, are used to evaluate each classification workflow. Our results show that 3D Multiscale Geometrical Classification outperforms the Colour and Geometrical Classification method. However, each method has advantages for specific applications pertinent to the wider marine scientific community. Results suggest that advancing from commonly employed 2D image analysis techniques to 3D photogrammetric classification methods is advantageous and provides a more realistic representation of CWC habitat composition.

Multi-scale Spectral-Spatial Remote Sensing Classification of Coral Reef Habitats Using CNN-SVM

Wan, J.X. and Ma, Y., 2020. Multi-scale spectral-spatial remote sensing classification of coral reef habitats using CNN-SVM. In: Jung, H.-S.; Lee, S.; Ryu, J.-H., and Cui, T. (eds.), Advances in Geospatial Research of Coastal Environments. Journal of Coastal Research, Special Issue No. 102, pp. 11-20. Coconut Creek (Florida), ISSN 0749-0208.In recent years, coral reefs have undergone serious degradation globally, prompting the use of remote sensing as an effective means to monitor these reefs on a large scale. Deep learning, which is a state-of-the-art image processing technique suitable for remote-sensing applications, can be used to learn nonlinear characteristics of images and is therefore applicable to the classification of small-scale coral reefs. This paper proposes a multi-scale method based on a convolutional neural network and support vector machine (CNN-SVM) to classify the coral reef habitats of Zhaoshu Island and Zhong Island in the Xisha Archipelago, China. This method combines spectrum, texture, and bidimensional empirical mode decomposition (BEMD) based scale separation algorithm to fully learn multi-scale information of coral reefs. Remote-sensing images captured by the WorldView-2 and Gaofen-2 (GF-2) satellites are used to evaluate the performance of the proposed CNN-SVM framework. The results indicate that the proposed method performs accurately and efficiently. Compared with SVM, random forest (RF), CNN, and CNN-RF, the overall accuracy is improved by 10.57 %, and the accuracy of classifying reef-clumping areas is improved by 17.44 %.

Habitat mapping using deep neural networks

Habitat mapping is an important and challenging task that helps in monitoring, managing, and preserving ecosystems. It becomes more challenging when marine habitats are mapped, as it is difficult to get quality images in an underwater environment. Moreover, achieving good location accuracy in underwater environments is an additional issue. Sonar imagery has good quality but is hard to be analyzed. Therefore, camera imagery is used for research purposes. Our research targets marine habitats - more specifically, coral reef marine habitats. Recognition of coral reef in underwater images poses a significant difficulty due to the nature of the data. Many species of coral reef have similar characteristics, i.e. higher inter-class similarity and lower intra-class similarity. Spatial borders between coral reef classes are hard to separate, as they tend to appear together in groups. For these reasons, the classification of coral reef species requires aid from marine biologists. This research work presents a technique for accurate coral reef classification using deep convolutional neural networks. The proposed approach has been validated on Moorea Labeled Corals (MLC), an imbalanced dataset, which is a subset of Moorea Coral Reef Long Term Ecological Research (MCR LTER) packaged for computer vision research. A custom valid patch dataset is extracted using the annotation files provided with the dataset. Two image enhancement algorithms and data-driven feature extraction techniques are employed using several pre-trained deep convolutional neural networks as feature extractors. Local-SPP technique is combined with feature extractors and followed by 2-layers multi-layer perceptron (MLP) classifier to achieve high classification rates.

Gradient Based Aura Feature Extraction for Coral Reef Classification

Coral reefs, of all marine ecosystems, are the most diverse. They teem with life with almost a quarter of all ocean species depending upon reefs for food and shelter. Because of the immense diversity of coral reefs, they are often referred to as the ‘Rain-Forests of the Sea’. But due to the lack of consistency and objectivity in human labeling, the manual annotations of coral reefs are not conceivable. Multiple species of coral have coinciding characteristics and this makes automatic annotation, a challenging task. This research work is based on developing machine learning algorithms to employ Aura Matrix based on gradient based Cumulative Relative Difference which exploits the neighborhood relationship between pixels by comparing the relative pixel strength with respect to the center pixel. The objective of this paper is to show improvement of coral reef classification accuracy and that the proposed work outperforms the state-of-the art CNN classifiers on four coral datasets namely EILAT, EILAT 2, MLC 2010 and RSMAS.

Cloud detection algorithm for multi-modal satellite imagery using convolutional neural-networks (CNN)

Cloud detection algorithms are crucial in many remote-sensing applications to allow an optimized processing of the acquired data, without the interference of the cloud fields above the surfaces of interest (e.g., land, coral reefs, etc.). While this is a well-established area of research, replete with a number of cloud detection methodologies, many issues persist for detecting clouds over areas of high albedo surfaces (snow and sand), detecting cloud shadows, and transferring a given algorithm between observational platforms. Particularly for the latter, algorithms are often platform-specific with corresponding rule-based tests and thresholds inherent to instruments and applied corrections. Here, we present a convolutional neural network (CNN) algorithm for the detection of cloud and cloud shadow fields in multi-channel satellite imagery from World-View-2 (WV-2) and Sentinel-2 (S-2), using their Red, Green, Blue, and Near-Infrared (RGB, NIR) channels. This algorithm is developed within the NASA NeMO-Net project, a multi-modal CNN for global coral reef classification which utilizes imagery from multiple remote sensing aircraft and satellites with heterogeneous spatial resolution and spectral coverage. Our cloud detection algorithm is novel in that it attempts to learn deep invariant features for cloud detection utilizing both the spectral and the spatial information inherent in satellite imagery. The first part of our work presents the CNN cloud and cloud shadow algorithm development (trained using WV-2 data) and its application to WV-2 (with a cloud detection accuracy of 89%) and to S-2 imagery (referred to as augmented CNN). The second part presents a new domain adaptation CNN-based approach (domain adversarial NN) that allows for better adaptation between the two satellite platforms during the prediction step, without the need to train for each platform separately. Our augmented CNN algorithm results in better cloud prediction rates as compared to the original S-2 cloud mask (81% versus 48%), but still, clear pixels prediction rate is lower than S-2 (81% versus 91%). Nevertheless, the application of the domain adaptation approach shows promise in better transferring the knowledge gained from one trained domain (WV-2) to another (S-2), increasing the prediction accuracy of both clear and cloudy pixels when compared to a network trained only by WV-2. As such, domain adaptation may offer a novel means of additional augmentation for our CNN-based cloud detection algorithm, increasing robustness towards predictions from multiple remote sensing platforms. The approach presented here may be further developed and optimized for global and multi-modal (multi-channel and multi-platform) satellite cloud detection capability by utilizing a more global dataset.

Classification of Coral Reef Benthos around Ganquan Island Using WorldView-2 Satellite Imagery

Xu, H.; Liu, Z.; Zhu, J.; Lu, X., and Liu, Q., 2019. Classification of coral reef benthos around ganquan island using WorldView-2 satellite imagery. In: Guido-Aldana, P.A. and Mulahasan, S. (eds.), Advances in Water Resources and Exploration. Journal of Coastal Research, Special Issue No. 93, pp. 466–474. Coconut Creek (Florida), ISSN 0749-0208.Coral reefs are structures formed by animals known as stony corals, which have an important impact on the physical and ecological environment of their surroundings. Presently, coral reefs are severely damaged by a variety of human activities and by global climate change. This paper is intended to explore the potential of using WorldView-2 (WV2) satellite images for coral reef bottom types classification. The research area for this study is around Ganquan Island in the South China Sea. Sunny sky and cloudless WV2 satellite imagery is selected to reduce atmospheric impact and improve the accuracy of coral reef classification. For the preprocessed imagery data, training samples were manually established, and the seafloor reef bottom types in the sea near Ganquan Island were divided into five substratum categories: coral reef, sand, reef, spray and deep sea water. These are classified using a Support Vector Machine (SVM), a Neural Network and the Maximum Likelihood method. The classification results were verified by comparing the satellite image visual interpretation classification results and the field seafloor image data. Confusion matrixes was compared and the classification accuracy was evaluated. The results show that the classification based on the SVM method is the best, with a total classification accuracy of 93.33% and a Kappa coefficient of 0.89. These results prove that high-resolution images can provide detailed and accurate information for coral reef bottom types classification and thus provide a scientific basis and reliable data for the establishment of a more complete coral reef ecological monitoring system and marine ecological environmental protection system.

Classification of Coral Reefs in the South China Sea by Combining Airborne LiDAR Bathymetry Bottom Waveforms and Bathymetric Features

Geographic information describing coral reefs plays an important role in constructing electronic chart systems and protecting the ecological environment of the ocean. To derive geographic information of coral reefs more effectively, this paper proposes a methodology to detect coral reefs by combining airborne LiDAR bathymetry (ALB) bottom waveform and bathymetric feature data. A feature vector was established by deriving bottom waveform variables (the peak amplitude, pulsewidth, area, skewness, kurtosis, and backscatter cross section) and bathymetric variables (the depth standard deviation, slope, bathymetric position index, Gaussian curvature, mean curvature, and roughness). Using a support vector machine classifier, coral reefs were detected by distinguishing two classes (coral reefs and others) on the seafloor. To evaluate the classification performance of coral reefs, the developed method was applied to Yuanzhi Island, South China Sea surveys, and verified by field data (aerial digital camera images and underwater video images). The results showed that the classification overall accuracy of coral reefs can be greatly improved from 80.59%/90.31% when ALB bottom waveform or bathymetric variables features were used separately to 93.57% when using a combination of ALB bottom waveform and bathymetric features. In addition, the kappa coefficient can also be greatly improved from approximately 0.61/0.80 to 0.87. And the new proposed method performs better compared to the current classification method using ALB data to detect coral reefs with an overall accuracy of 90.92% and Kappa of 0.81. This highlights the potential of ALB data, combining waveform data and bathymetric data, for precisely detecting coral reefs in shallow water areas.

Implications of sensor design for coral reef detection: Upscaling ground hyperspectral imagery in spatial and spectral scales

Remote sensing offers a potential tool for large scale environmental surveying and monitoring. However, remote observations of coral reefs are difficult especially due to the spatial and spectral complexity of the target compared to sensor specifications as well as the environmental implications of the water medium above. The development of sensors is driven by technological advances and the desired products. Currently, spaceborne systems are technologically limited to a choice between high spectral resolution and high spatial resolution, but not both. The current study explores the dilemma of whether future sensor design for marine monitoring should prioritise on improving their spatial or spectral resolution. To address this question, a spatially and spectrally resampled ground-level hyperspectral image was used to test two classification elements: (1) how the tradeoff between spatial and spectral resolutions affects classification; and (2) how a noise reduction by majority filter might improve classification accuracy. The studied reef, in the Gulf of Aqaba (Eilat), Israel, is heterogeneous and complex so the local substrate patches are generally finer than currently available imagery. Therefore, the tested spatial resolution was broadly divided into four scale categories from five millimeters to one meter. Spectral resolution resampling aimed to mimic currently available and forthcoming spaceborne sensors such as (1) Environmental Mapping and Analysis Program (EnMAP) that is characterized by 25 bands of 6.5nm width; (2) VENμS with 12 narrow bands; and (3) the WorldView series with broadband multispectral resolution. Results suggest that spatial resolution should generally be prioritized for coral reef classification because the finer spatial scale tested (pixel size < 0.1m) may compensate for some low spectral resolution drawbacks. In this regard, it is shown that the post-classification majority filtering substantially improves the accuracy of all pixel sizes up to the point where the kernel size reaches the average unit size (pixel < 0.25m). However, careful investigation as to the effect of band distribution and choice could improve the sensor suitability for the marine environment task. This in mind, while the focus in this study was on the technologically limited spaceborne design, aerial sensors may presently provide an opportunity to implement the suggested setup.