Content uploaded by Paulo Drews-Jr
Author content
All content in this area was uploaded by Paulo Drews-Jr
Content may be subject to copyright.
A preview of the PDF is not available
Microalgae Classification using Semi-supervised and Active Learning based on Gaussian Mixture Models
Abstract and Figures
Microalgae are unicellular organisms that have different shapes, sizes and structures. Classifying these microalgae manually can be an expensive task, because thousands of microalgae can be found in even a small sample of water. This paper presents an approach for an automatic/semi-automatic classification of microalgae based on semi-supervised and active learning algorithms, using Gaussian mixture models. The results show that the approach has an excellent cost-benefit relation, classifying more than 90 % of microalgae in a well distributed way, overcoming the supervised algorithm SVM.
Figures - uploaded by Paulo Drews-Jr
Author content
All figure content in this area was uploaded by Paulo Drews-Jr
Content may be subject to copyright.
... The flow of pure semi-supervised and transductive learning is exhibited in Fig. 7. The commonly used semi-supervised learning models are self-training [100,101], Gaussian mixture model (GMM) [102,103], and semi-supervised CNN [99,104], etc. ...
... The findings show the benefits of image analysis employing deep learning for microalgal culture classification. Drews-et al. (2013) have used the Gaussian mixed semisupervised classification model and active learning to classify the microalgae using the data obtained from the FlowCAM. They have used microalgal features such as diameter, length, width, aspect ratio, etc., as the input data. ...
The increased water scarcity, depletion of freshwater resources, and rising environmental awareness are stressing for the development of sustainable wastewater treatment processes. Microalgae-based wastewater treatment has resulted in a paradigm shift in our approach toward nutrient removal and simultaneous resource recovery from wastewater. Wastewater treatment and the generation of biofuels and bioproducts from microalgae can be coupled to promote the circular economy synergistically. A microalgal biorefinery transforms microalgal biomass into biofuels, bioactive chemicals, and biomaterials. The large-scale cultivation of microalgae is essential for the commercialization and industrialization of microalgae biorefinery. However, the inherent complexity of microalgal cultivation parameters regarding physiological and illumination parameters renders it challenging to facilitate a smooth and cost-effective operation. Artificial intelligence (AI)/machine learning algorithms (MLA) offer innovative strategies for assessing, predicting, and regulating uncertainties in algal wastewater treatment and biorefinery. The current study presents a critical review of the most promising AI/MLAs that demonstrate a potential to be applied in microalgal technologies. The most commonly used MLAs include artificial neural networks, support vector machine, genetic algorithms, decision tree, and random forest algorithms. Recent developments in AI have made it possible to combine cutting-edge techniques from AI research fields with microalgae for accurate analysis of large datasets. MLAs have been extensively studied for their potential in microalgae detection and classification. However, the ML application in microalgal industries, such as optimizing microalgae cultivation for increased biomass productivity, is still in its infancy. Incorporating smart AI/ML-enabled Internet of Things (IoT) based technologies can help the microalgal industries to operate effectively with minimum resources. Future research directions are also highlighted, and some of the challenges and perspectives of AI/ML are outlined. As the world is entering the digitalized industrial era, this review provides an insightful discussion about intelligent microalgal wastewater treatment and biorefinery for researchers in the field of microalgae.
... Decision Tree -Could predict the growth of microalgae which are natively polycultured in an open raceway pond -Simple to use and assess -Can only be constructed with big enough dataset (Noguchi et al., 2018) Convolutional Neural Network -Required minimum preprocessing -No expert assistance -Present pertinent data with an accuracy of 88.59 %. (Correa, Drews, Botelho, de Souza, & Tavano, 2017) Support Vector Machine -Growth phases of algal can be assessed -Environmental changes with the microalgae Raman spectra can be tentatively predicted (He et al., 2018) Gaussian mixture models -More effective than SVM in experiments -Addition of the active learning algorithm increased recognition accuracy to 92 % (Drews et al., 2013) Deep Learning -Possible to forecast a species' identification and appropriate stage of the life cycle with an impressive 97 % accuracy (Dunker et al., 2018) Extreme learning machine -Addition of genetic algorithm, improve the model -Extreme learning machine's high accuracy for different numbers of hidden neurons (Purnomo et al., 2015) Logic Regression -The technique had high accuracy -Performance can be enhanced with techniques of deep learning (Wang et al., 2018) Imaging technique based on time-stretch technology (TS-QPI) ...
Production and extraction systems of algal protein and handling process of functional food ingredients need to control several parameters such as temperature, pH, intensity, and turbidity. Many researchers have investigated the Internet of Things (IoT) approach for enhancing the yield of microalgae biomass and machine learning for identifying and classifying microalgae. However, there have been few specific studies on using IoT and artificial intelligence (AI) for production and extraction of algal protein as well as functional food ingredients processing. In order to improve the production of algal protein and functional food ingredients, the implementation of smart system is a must to have real-time monitoring, remote control system, quick response to sudden events, prediction and characterisation. Techniques of IoT and AI are expected to help functional food industries to have a big breakthrough in the future. Manufacturing and implementation of beneficial smart systems are important to provide convenience and to increase the efficiency of work by using the interconnectivity of IoT devices to have good capturing, processing, archiving, analyzing, and automation. This review investigates the possibilities of implementation of IoT and AI in production and extraction of algal protein and processing of functional food ingredients.
... To be specific, there is a sudden increase in 2013 which interprets that a great deal of attention and effort is given among researchers to study the classification of microalgae with ML. In 2013, most microalgae classification studies are based on leveraging models such as semisupervised, active learning, artificial neural network (ANN), support vector machine (SVM), and multilayer perceptron (MLP), however still lacked methods capable of dealing with a larger dataset and automatic feature extraction which results in a drastic drop in 2014 (Drews et al., 2013;Santhi et al., 2013). Most likely, the advancement of DL, computational power/ intelligence along with other sophisticated algorithms can be seen as the trend increases from 2014 to 2020. ...
Identification of microalgae species is of great importance due to the uprising of Harmful Algae Blooms (HABs), affecting both the aquatic habitat and human health. On the contrary, microalgae have been identified as future green biomass and alternatives due to their promising bioactive compounds activities that play a significant role in many industrial applications. Recently, microalgae species identification has been conducted through DNA analysis and various microscopy techniques such as light, scanning electron, transmission electron, and atomic force microscopy. The aforementioned procedure has encouraged researchers to consider alternate ways due to several limitations such as costly validation, requiring skilled taxonomists, prolonged analysis, and low accuracy. Nonetheless, this review highlights the potential and innovation of digital microscopy with the incorporation of both hardware and software that could produce a reliable recognition, detection, enumeration, and real-time acquisition of microalgae species. Several steps such as image acquisition, processing, feature extraction, and selection for generating high image quality by removing unwanted artifacts and noise from the background. The identification of microalgae species is then followed by robust and reliable image classification machine learning as well as deep learning algorithms such as artificial neural networks, support vector machines, and convolutional neural networks. Overall, this review paper aims to explore and provide comprehensive insights into numerous possibilities of microalgae image identification, image pre-processing, and machine learning techniques to develop a robust digital classification tool in near future.
... Microalgae is a type of unicellular microorganism that appears on the Earth in a variety of sizes, shapes, and structures [20]. They are easily found in marine, freshwater, and some ground structures and can be further classified into photosynthetic, parasitic, or symbiotic [36]. ...
Chemical fertilizer is the most ubiquitous nutrient source used to cultivate microalgae. However, the bottlenecks associated to its vast usage such as high cost and environmental hazards are evident, whose solution is to include a search for alternative nutrient sources. Thus, nutrient-rich wastewater has been utilized in recent years for microalgae cultivation as it is widely available and able to minimize the usage of freshwater. However, it contains a large amount of heavy metal ions and microorganisms which in turn can inhibit the growth of microalgae cells. This scenario led to the development of compost derived from animal manure as a feasible and economic substitute for existing nutrient sources to grow microalgae. It constitutes a copious amount of nutrient elements, cheap, omnipresent, and environmental-friendly, which makes it as a sustainable option for the cultivation of microalgae in the commercial stage. Nevertheless, its full-scale application is limited by several challenges such as transparency problems, variability in nutrient content, selectivity over some microalgae species, and formation of other microorganisms during the composting process as well as nitrogen leakage into the atmosphere. To overcome these limitations, pre-treatment methods such as decolourization, dilution, and zeolite addition have been incorporated and discussed to increase the potential usage of animal manure for large-scale production of microalgae biomass.
... The goal of ML processing is to automatically distinguish between living, dead, and debris cells. Semisupervised and active learning based on Gaussian mixture models can also be used to classify microalgae (Drews et al., 2013). The development of models and the implementation of advanced control mechanisms are critical. ...
Valorization of Microalgal Biomass and Wastewater Treatment provides tools, techniques, data and case studies to demonstrate the use of algal biomass in the production of valuable products like biofuels, food and fertilizers, etc. Valorization has several advantages over conventional bioremediation processes as it helps reduce the costs of bioprocesses. Examples of several successfully commercialized technologies are provided throughout the book, giving insights into developing potential processes for valorization of different biomasses. Wastewater treatment by microalgae generates the biomass, which could be utilized for developing various other products, such as fertilizers and biofuels.
This book will equip researchers and policymakers in the energy sector with the scientific methodology and metrics needed to develop strategies for a viable transition in the energy sector. It will be a key resource for students, researchers and practitioners seeking to deepen their knowledge on energy planning, wastewater treatment and current and future trends.
... The goal of machine learning processing is to automatically distinguish between living, dead, and debris cells. Semi-supervised and active learning based on Gaussian mixture models can also be used to classify microalgae [Drews et al. (2013)]. The development of models and the implementation of advanced control mechanisms are critical. ...
To address the energy crisis and environmental issues, algal farming has the potential to lead us toward sustainable development. Microalgal farming may be used to produce food, medication, cosmetics, and bioenergy. But, the development of a method that is both efficient and affordable is a major roadblock in the mass production of algae. Internet of Things (IoT) has the potential to improve the efficiency of microalgal biorefineries by saving time and money. A network of sensors based on IoT may allow operators to monitor the development and production of algae in real time, saving money and time. Online monitoring can help regulate and control biological processes. Genetic modification tools like RNAi, ZFNs, TALENs, and CRISPR-Cas9 that help us in the improvement of the algal strains can also benefit from the artificial intelligence algorithms.
... The goal of ML processing is to automatically distinguish between living, dead, and debris cells. Semisupervised and active learning based on Gaussian mixture models can also be used to classify microalgae (Drews et al., 2013). The development of models and the implementation of advanced control mechanisms are critical. ...
Life Cycle Assessment (LCA) is a widely used tool for estimation of environmental footprint of any products, technologies and services, throughout its whole lifecycle from cradle to grave. It is a standardized decision support system, for quantifying the different environmental impact categories and deciding upon the sustainability of each system employed. The use of LCA tools for wastewater treatment and their impact assessment is started very recently. In wastewater treatment the LCA tools compile and evaluate the inputs and the outputs, and consider their potential environmental impacts associated with the operation of the system for all types of wastewater treatment plants either for conventional or algal ponds, throughout its whole process chain. The LCA studies generally follow ISO standards (International Organization for Standardization) with baseline framework consisting of four phases’ viz. goal and scope determination, life cycle inventory analysis (LCI), life cycle impact assessment (LCIA) and interpretation of results. The inventory analysis accumulate the data or the database for analysis, using specific criteria or data quality matrices and the impact assessment is carried out with the help of different type of softwares viz. SimaPro®, Gabi®, OpenLCA®, Umberto® etc. The impact assessment transforms the mathematical data to environmental effect equivalent via the factor multiplication. The LCA studies has validated that the wastewater treatment with microalgae comparing to the conventional, can significantly reduced the negative environmental impacts, as well as the system has the advantage on low cost of operation, the possibility of recycling the nutrients in wastewater to high value products, reducing the emissions by absorption of CO2 present in the flue gases and the discharge of oxygenated effluent into the water body.
... The accuracy and effectiveness of the model were confirmed after the identification of a mixture of microalgae. Paulo Drews-Jr et al. applied semi-supervised learning in their work on microalgae classification (Drews et al., 2013). The dataset was microalgae data obtained through the FlowCAM device in the Atlantic Ocean. ...
Microalgae are essential parts of marine ecology, and they play a key role in species balance. Microalgae also have significant economic value. However, microalgae are too tiny, and there are many different kinds of microalgae in a single drop of seawater. It is challenging to identify microalgae species and monitor microalgae changes. Machine learning techniques have achieved massive success in object recognition and classification, and have attracted a wide range of attention. Many researchers have introduced machine learning algorithms into microalgae applications, and similarly significant effects are gained. The paper summarizes recent advances based on various machine learning algorithms in microalgae applications, such as microalgae classification, bioenergy generation from microalgae, environment purification with microalgae, and microalgae growth monitor. Finally, we prospect development of machine learning algorithms in microalgae treatment in the future.
Microalgae are unicellular organisms that have different shapes, sizes and structures. Classifying these microalgae manually can be a expensive task, because in a small sample of water, thousands of microalgae can be found. This paper presents an approach for an automatic/semi-automatic classification of microalgae based on semi-supervised and active learning algorithms, using Gaussians Mixtures Models. The results show that the approach has an excellent cost-benefit, correctly classifying approximately 94% microalgae in a well distributed way, overcoming the supervised algorithm SVM.
Instruments for in vivo identification and quantification of marine organisms are becoming more common, and the interpretation of data from these instruments is still evolving. In the present study, we compare the sizing performance of 3 instruments: (1) a black and white (B/W) FlowCAM II; (2) a BeckmanCoulter Multisizer III (MIII); and (3) an inverted microscope. We applied 3 different particle sizing algorithms available from the FlowCAM to suspensions of 5 different particle morphotypes (4 different phytoplankton species and a spherical NIST calibration bead). The FlowCAM generated size distributions similar to those reported by the MIII for the spherical calibration beads. However, differences in reported sizes emerged among FlowCAM algorithms as well as among instruments when applied to morphologically more complex particles, such as diatom chains. There was an immediate and substantial loss of cell counts when live cells were fixed in Lugol’s solution, but only minor differences in cell size distributions among the different FlowCAM algorithms. The difference in sizing performance of the FlowCAM algorithms affects biovolume estimates of the natural plankton samples analysed. Species diversity was apparently higher in samples analysed by microscopy than with the FlowCAM, but the cell size distribution from the microscope was extremely narrow compared to FlowCAM and MIII. The present study demonstrates that the particle sizing algorithm has severe impact on the characteristics of the particle size distribution and on the total community biomass estimate.
Nowadays, the advance of the technology allows robots to acquire dense point clouds decreasing the price and increasing the performance. However, it is a hard task to deal with due to the large amount of points, the redundancy and the noise. This paper proposes an adaptable system to build a 3D feature model of point clouds using Gaussian Mixture Models. These 3D models are used in order to detect changes in the autonomous robot's working environment. The presented work describes an efficient change detection system based on two consecutive stages. First, a top-down approach estimates features using Gaussian Mixture Models. The presented new approach improves the performance of previous related works in terms of computational load and robustness, nevertheless the system is selection criteria dependent. Thus, the efficiency of different selection criteria are evaluated and compared in this paper. Experimental results demonstrate that the Minimum Distance Length (MDL) criteria outperforms the other studied methods. In the second stage, a change detection method is performed using the previously estimate Mixture of Gaussians. The proposed full system is able to detect changes using Gaussian Mixture Models with a reduced computational cost in relation to state-of-art algorithms.
Thesupport-vector network is a new learning machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very high-dimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data.High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition.
Semantic Information Theory (SIT) is concerned with studies in Logic and Philosophy on the use of the term information, “in the sense in which it is used of whatever it is that meaningful sentences and other comparable combinations of symbols convey to one who understands them” (Hintikka, 1970). Notwithstanding the large scope of this description, SIT has primarily to do with the question of how to weigh sentences according to their informative content. The main difference with conventional information theory is that information is not conveyed by an ordered sequence of binary symbols, but by means of a formal language in which logical statements are defined and explained by a semantics. The investigation of SIT concerns two research directions: the axiomatisation of the logical principles for assigning probabilities or similar weighting functions to logical sentences and the relationship between information content of a sentence and its probability.
The support-vector network is a new learning machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very high-dimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data. High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition.
High-resolution photomicrographs of phytoplankton cells and chains can now be acquired with imagingin-flow systems at rates that make manual identification impractical for many applications. To address the challenge for automated taxonomie identification of images generated by our custom-built submersible Imaging FlowCytobot, we developed an approach that relies on extraction of image features, which are then presented to a machine learning algorithm for classification. Our approach uses a combination of image feature types including size, shape, symmetry, and texture characteristics, plus orientation invariant moments, diffraction pattern sampling, and co-occurrence matrix statistics. Some of these features required preprocessing with image analysis techniques including edge detection after phase congruency calculations, morphological operations, boundary representation and simplification, and rotation. For the machine learning strategy, we developed an approach that combines a feature selection algorithm and use of a support vector machine specified with a rigorous parameter selection and training approach. After training, a 22-category classifier provides 88% overall accuracy for an independent test set, with individual category accuracies ranging from 68% to 99%. We demonstrate application of this classifier to a nearly uninterrupted 2-month time series of images acquired in Woods Hole Harbor, including use of statistical error correction to derive quantitative concentration estimates, which are shown to be unbiased with respect to manual estimates for random subsamples. Our approach, which provides taxonomically resolved estimates of phytoplankton abundance with fine temporal resolution (hours for many species), permits access to scales of variability from tidal to seasonal and longer. © 2007, by the American Society of Limnology and Oceanograpny, Inc.
In this paper, a novel approach for classifying algal images was presented, which is used in flow-cytometry-based real-time red tide monitoring system. Firstly, an ensemble of support vector machines (SVMs) was trained and the test samples were labeled by them based on the summation of negative probability (SNP). Secondly, those samples most likely mistakenly labeled were picked out and re-labeled by semi-supervised fuzzy c-means (FCM) clustering algorithm. Experiments show that this new method improves the accuracy of algal images classification for the same subject with SVMs of different kernels.