Applications of Remote Sensing in Agriculture

In the 1960s, the period of satellite remote sensing was opened when the electronic sensors and cameras were mounted on the spacecraft.  Currently, a massive collection of satellite structures exists, which effectively records information about the earth’s surface. A vast scope of pictures is accessible from the satellites.  Passive and active sensors, functioning from the ultraviolet regions to microwave regions of the electromagnetic spectrum, gather a tremendous amount of info regarding the realm each day. Each of the systems differs in regards to their ranges, radiometric, spatial, and temporal resolution (Sharma, Kamble & Gunasekaran, 2018).

Farming plays a crucial part in the countries’ economies. The food production is essential for everyone, and generating food more economically is the objective of each farmer and agricultural bodies.  The satellite can image regions, fields, and nations on a regular basis. The pictures can assist in determining the intensity and area of crop strains. Therefore, it can be applied to advance and execute a spot handling plan that maximizes the application of agricultural chemicals. The critical agricultural use of remote sensing encompasses the following: crop yield projections, crop conditions evaluations, and crop type classification (Pohl & Van Genderen, 2016). 

Remote sensing knowledge can be utilized to prepare crop types maps and delineating their scope. Conventional approaches to getting this info are ground surveying and census. The applications of satellites are crucial as they can create a repetitive and organized coverage of colossal zones and offer information regarding crop health. The crop statistics are required for agricultural bodies to plan an inventory of how to cultivate in specific areas and what time. The above info serves to project grain crop output, gathering crop yield statistics, aiding the crop rotation inventory, soil production mapping, pinpoint of elements affecting cross stress, evaluating the vegetation damage, and monitoring cultivation action (Salmon, Friedl, Frolking, Wisser & Douglas, 2015).

Many types of remote sensing structures utilized in agriculture exist; however, the most commonly used is a passive structure that senses the electromagnetic (EM) power reflected from the vegetation. The crop spectral reflection depends on the phenology or growth modification, stage type, and crop health. Therefore, it can be gauged and monitored by multispectral sensors. Several remote sensing sensors function on the red, green, and near-infrared zones of the EM spectrum, and they weigh both the reflectance and absorption influence linked with the crop. Multispectral changes aid accurate identification, detection, and monitoring of the plant. The crop phenology observation needs a multi-temporal pictures. Diverse sensors regularly offer complimentary info, and when linked together, can facilitate interpretation and categorize pictures. For instance, a combination of high-resolution panchromatic photos with bristly resolution multispectral images, or blending passively and actively sensed info (Wang, Zheng, Lei & Bai, 2019).

Types of Remote Sensing Sensors Used in Agriculture

Remote sensing has several characteristics that make them monitor crop health.  The optical sensing devices makes it be observed in the infrared regions, where the wavelength is highly sensitive to vegetation vigor, crop damage, and stress.  Remote sensing imagery also provides the spatial overview needed for the land (Szuster, Chen & Borger, 2011).  It can also aid in identifying vegetation affected by states that are too wet or dry, weed infestations, affected by insects or weather connected damages. Pictures can be got over the growing period not merely to detect concerns, but also to monitor victory of the handling. Monitoring plant’s health and identifying images need a high resolution, multi-temporal, and multispectral abilities. One of the most significant elements in creating imagery critical to the cultivators is a swift turnaround time from the acquisition of the data to crop info dispersion (Hu, Xia, Hu & Zhang, 2015).

It is obvious that farming can be reflected as the “vertebral pillar” of human existence and has significant regulator over the nations’ economy.  It shows the routine requirement for standard checking of the plant condition. There are a lot of characteristics to monitor vegetation conditions beginning from plant vigor states, soil humidity, humidity accessibility, and strain offered by abiotic influences (for instance, rain, temperatures, and moisture) and also biotic reasons such as illness and pest.  Any further deferment from the healthy growth parameters affects plant development and thus diminishes the production. Therefore, it is vital to evaluate vegetation conditions to assess crop condition for whole cycles of growth.

Suitable crop assessment at a correct level needs checking broad areas by a robust scheme. Remote sensing capabilities provide this over nondestructive synoptic screening abilities. It is recognized that spectral reaction of the topsoil attributes is dissimilar from diverse parts of electromagnetic range. The above-sensed dealings help in the discovery and acknowledgement of the international surface attribute.

The most significant aim of the usage of remote sensing in farming is to determine crop features by assessing the info encompassed in the distributed signal. The first most significant functional execution in the cultivation application of remote sensing was that of the Large Area Crop Inventory Experiment, where an exertion was created by approximate nation-wise Wheatland and outputs through the LANDSAT digital info (Krishna et al., 2014). Landsat is a routine of the earth surveillance satellite developed under a mutual program of the National Aeronautics and Space Administration (NASA) and the United States Geological Survey (USGS) (Dong et al., 2014).  Assessment of the impact of meteorological and edaphic aspects to plant has remained to be a key theme. There is a broader feature such as climatic, biotic, edaphic, agronomic, and hydrologic, which manage crop growth and yield. Moreover, the weather has a consideration point over the advancement and crop productivity reactions. Remote sensing approach employing the prospects of multi-temporal, multispectral, and synoptic contact has shown an appropriate prospective in offering a massive position of plant state and generation potential at the local phase. With the remote sensing approach, the kinds of plants grown in the location, crop state, and productivity can be reflected.  Noting the plant condition by remote sensing can make the crop condition in addition to the progress and health of their growth. Getting the plant situation info at initial phases of plant growth is still more important than getting the fixed yield after the harvesting retro.  

Multispectral Imaging and Spectral Reflection

Plant condition projection necessitates a data input, for instance, ecological states such as relative humidity, air temperature, rainfall, and surface states such as soil temperature and soil moisture. Remote sensing indices such as Normalized Difference Vegetation Index (NDVI), Land Surface Water Index (LSWI), Water Deficit Index (WDI), and Temperature-Vegetation Dryness Index (TVDI) got from the satellite imagery are essential to driving plant growth states or soil humidity conditions (Rahimzadeh-Bajgiran, Omasa & Shimizu, 2012). NDVI computes plant density by assessing the differences between red luminosity (which plant attracts) and near-infrared (which crop powerfully reflects). Furthermore, LSWI uses the Near-infrared (NIR) and Short Infrared (SWIR) areas of the electromagnetic range. LSWI is well known to be vulnerable to the total amount of liquefied water in plants and its topsoil. At the same time, there is active light absorption by fluid water in the SWIR. Additionally, TVDI is attained from NDVI-spatial Land Surface Temperature and be used as an indicator of moistness in the soil and thus the plant water force. The Soil Adjusted Vegetation Index (SAVI) takes into account the optical soil features on the vegetation cover reflectance. The WDI denotes the comparative speed of concealed heat variations, so it shows a speed of “zero” for an utterly damp surface and a value of “one” regarding desiccated surfaces where there is no unseen heat variation (Rahimzadeh-Bajgiran, Omasa & Shimizu, 2012).  The remote sensing approach supports to build a time-based growth contour of crops over its advancement stage.  With the retrieval of ecological elements in addition to remote sensing directories, it is quick to acknowledge the growth framework and also their link among each other and the effects of concerned variables on plant growth. Based on preceding data and tests, the remote sensing approach is particularly crucial in projecting the crop growth at a land level. The integration of the crop conditions entails ecological and surface. However, remote sensing, in addition to the soil area, supports the advancement of the model to compute the crop conditions (Rahimzadeh-Bajgiran, Omasa & Shimizu, 2012).

Remote detecting offers a vast point to get a normal synoptic image of the earth. Many classes of choosy info on accessible capitals such as topsoil usage, soil humidity, crop natures and state, and soil types info can be extracted from the satellite broadcasting data (Pohl & Van Genderen, 2016). 

Remote identifying as a tool will offer data regularly and at a costly value to allow right time for crop states interference. Satellite assemblies provides a temporally and spatially lasting cover of the world.   The Moderate Resolution Imaging Spectroradiometer (MODIS) also well recognized as an imaging appliance was initiated into the earth orbits through NASA. The tools capture information in 36 spectral differing in wavelength, beginning from 0.4×10⁻⁶m to 14.4 ×10⁻⁶m and at changing in three-dimensional movements. The Atmospheric Infrared Sounder, AIRS, is a facility device whose aim is to aid weather study and improve meteorological conditions forecast.

Monitoring Plant Health with Remote Sensing

Conversely, the Sentinel-2 process, supported by the European Space Agency (ESA), comprises binary polar-orbiting satellites: sentinel-2A and Sentinel-2B. It provides organized general handling of land analysis and surfaces between an altitude of 56 degrees South and 83 degrees North.  Many functions such as agriculture, terrestrial cover variations, and planning of biological elements such as foliage chlorophyll quantity, leaf area indicators, and leaf water quantity can be projected. Sentinel-2A, which was initiated in 2015, supplies wide-ranging reporting of the globe’s land every ten days and, when it is linked with Sentinel-2B launched in 2017, the handling time has minimized.  The two satellites are the same and transmit one blend of broad coverage:  systematic and controlled gaining of high-resolution images raised reexamination speed of five days, extension vision area of about 290km, and high resolution of 10 meters. It is possible due to high-tech multispectral images, 13 spectral bands, of which three are encompassed into the ‘red edge’ portion of the spectral area. As a result of improved features assessed to previous functions, Sentinel-2 is capable of pinpointing fundamental changes in crop healthiness, distinguishing between vegetation and disperses suitable info on many biophysical elements (Gascon, 2014). The above components can aid the chores of operators and experts to pinpoint food deficiencies signs in nations. 

Furthermore, the crop yields projections, modeling productions, and crop strain acknowledgment have been assessed through remote detecting data. Crop illness discovery and acknowledgement are crucial to suitable crop yield. One of the prospects of usages of remote detecting in cultivation is the projection of plant land and plant condition recognition. Vigorous crops offer a high reflectance in the near-infrared zones and an inferior in the observable locations. The above concept can be used in discerning active and infected plants.

Once crops are infected with a sickness, there is an absorption of an incident solar array. It is perhaps caused by a minimized chlorophyll amount and changes in inner structure. The variations of uptake consequently influence the infected plant’s reflectance. As a result, in assessing the scope variance of diseased and strong crops, experts are capable of acknowledging the stress potency of green plants. When the chlorophyll amount incline to reduce under diseases, the incent solar beam assimilation by green plants falloffs in the Near-Infrared scope relying on the contamination strength. Its foliar internal composition primarily provokes the robust spectral reflectance of green vegetation in the Near-infrared ray.  Nevertheless, crops with malady stress display various scopes of internal morphological changes, which leads to a reduction of spectral reflectance in the Near-Infrared display. The above spectral aspects of plants are the bases for remote detecting of malady-strained plants (Krishna et al., 2014).

Importance of Remote Sensing in Identifying Agricultural Issues

Remote detected info had been used for assessment of terrestrial cover since the remote identifying was established. Land cover projection using remotely detected records cannot further be recognized as complete in the logic of their 3-D and spectral evaluations (Gurney, 2014). Land cover classification improvement and categorization of satellite info can be prepared by methods such as Artificial Neural nets, K-nearest neighbors, Decision Tree analytical approach, and lastly, Clustering division and segmentation approaches. The Artificial Neural Networks structures the composition or operations of biological neural grids. The key aim is to realize prototype identical for the categorization and relapse concerns (Breiman, 2017). However, the approach copies the techniques used by the natural creatures rather than rigorously dependent on the correct math-centered method. One of the instances of the artificial neural network architecture algorithms is Perceptron, Radial Basis Function Network (RBBFN), Self-Organizing Map (SOM), Hopfield Network, and Feed-forward Neural Network (Abiodun et al., 2018).  The above algorithms are crucial in finding illustrations that are hard for being physically mined. From the viewpoint of this conformation, trials are established to the Artificial Neural Network through the input sheath that has merely one neuron for each component present in the input inventory and is related to one more concealed coats present in the assembly (Salmon, Friedl, Frolking, Wisser & Douglas, 2015). It is worth noting that the processing occurs in the hidden sheets by an organization of connections differentiated by the biases and weight. The input is gathered, and the neuron approximations a weighted quantity accumulation. Following the result and a preset activation function, it creates a verdict whether it ought to be castoff or activated. Consequently, the neuron moves the records downstream to other linked neurons. Finally, the last concealed layer is related to the output sheath, which has one neuron for every likely required output (Benza, Weeks, Stow, Lopez-Carr & Clarke, 2016).

The decision tree creates a framework of choices on the foundations of actual value seen in the archives (Jelinek, Abawajy, Kelarev, Chowdhury & Stranieri, 2014). The resulting tree structure allows creation of comparisons amid new and present information quickly. The above type of algorithm frequently finds the use of regression and categorization concerns. Generally, decision trees are machine learning algorithms that progressively divide data sets that can be, to less extent, be described through a label number. Therefore, they try to authenticate new data-centered on this comprehension. Thus, the decision tree algorithms are perfect for solving the classification and regressions challenges (Breiman, 2017).

Regression trees are used while the dependent number is reflected as continuous or quantitative. Classifications trees are used while the conditional value is contemplated as categorical or qualitative.  Some of the decision tree algorithms are Chi-squared Automatic Interaction Detection (CHAID), Iterative Dichotomiser 3 (ID3), and classification and Regression Tree (CART) (Zhong & Zhang, 2011). Clustering approaches express a concept for controlling info by a class. There are several instances of clustering construction algorithms: Expectation-Maximization (EM), Hierarchical, K-medians, and K-means (Breiman, 2017).

The imagery configuration of remote detecting recording either can be done in the way of realizing various spectral sets on the foundations of phantom illumination speed (Yang, Cao, Xing & Li, 2015). The soil-truth plan is prepared by means of help of user interface through the imagery handling approach and comprehending of classes on the land. Typically, the spectral concept present inside the info for each pixel is used as the numerical source for categorization.  Spectral framework identification creates a locus to the set of classification of design pixels on the foundation of the spatial linking among pixels adjacent to them. The above kinds of groups attempt to replicate the sort of spatial fusion prepared using qualified experts through the visual assessment operations found on the picture texture and quality, attribute dimension, pixel nearness, duplication, contour, in addition to the surrounding. The focus comprises determining spectral prototypes through the spatial links in a given remote identifying record rather than discovering by a temporal procedure (Yang, Cao, Xing & Li, 2015).

Spectral imagery categorization is typically divided into two key classes of approaches: supervised and unsupervised. The key difference among these approaches is that in the former, the classification comprises a training phase succeeded by a categorization phase (Zheng, Chen, Lin & Wang, 2013). In the latter classification, the image data are first categorized by linking them into the natural spectral linkages.  The spectral sets are classified by assessing them to soil reference records (Li & Ge, 2015).

One of the regular signs of defining categorization accuracy is the building of a sorting error table. Conclusions from the arrangement and organization of remotely detected records are typically recapped as confusion matrix or else contingency table. Inaccuracy matrices assess, on an unconditional base, the probable assembly among the reference accounts in addition to the connected results of an electronic structure. The above models are referred to as square, through the summation of columns and rows equal to the totality of categorization whose classification accuracy is being reflected (Ye, Luo, Dong, He & Min, 2019).

The categorization of pictures has regularly been realized through using standard arithmetical and machine learning approaches in the previous days. Mathematical approaches like Bayesian networks are perfect, while the statistics is contemplated as noise-free or normalized (Cao & Lin, 2015). The statistical concepts are accomplished more once the instruction concerning the categories is available; however, they have limits in the case of classification purposes and when the distribution of the logged place is not pinpointed, like the case with the remote detecting spatial annals. Data mining competence has become gradually more critical approaches to handle info from a broad pool of documents (Li, Zhang & Zhao, 2019).

The term “data mining” has been there for many years, while the machine learning statistics and routines such as Decision tree (DT) and Artificial Neural Networks currently linked with the data mining were advanced (Cao & Lin, 2015). Now, many image classification approaches have been used to mine much info from remote identifying imageries. Gathering of crucial classification tactics is particularly vital to extract important outcomes from images actively.  Investigative classifiers like the decision tree and artificial neural networks do not use arithmetical elements to acknowledge sessions. They are well reformed for assessing noisy, missing inventory or multimodal (Jelinek, Abawajy, Kelarev, Chowdhury & Stranieri, 2014).

Data mining for three-dimensional form recognition is an approach of assessing fascinating info, for example, relationships, configurations, irregularities, changes, and critical organizations, from colossal volume (Li, Zhang & Zhao, 2019). Because of the discarding of enormous amounts of records, data mining has fascinated important deliberation in the information management operations. Normally, data mining projects can be characterized into two forms: predictive and descriptive data mining. The latter denotes to the info set concisely and highpoints common statistics features; the former achieve explanation on the available figures set and exertions to estimate the new data comportment (Li, Zhang & Zhao, 2019).

A data mining assembly could accomplish at least one of the information mining tasks: connection, planning, forecasting, and clustering. Among many data mining approaches and methods, the Artificial Neural Network (ANN) procedure is one of the most lengthily used practices in engineering (Wang, Jia, Yao & Xu, 2019). It is specially used when info is available from many foundations, in addition to a priori understanding of descriptive advances that is available because of the ability of ANN to research sophisticated conformations quickly.   This technique was effectively used in many arenas, like biology, physics, chemistry. Moreover, the decision tree schemes of data mining methods are more directly adjusted for cataloguing, from the period when information symbolizing a specified distinct are categorized through the decision tree building to be ordered straight into a preprogrammed cluster (Jelinek, Abawajy, Kelarev, Chowdhury & Stranieri, 2014). They not merely signify an active organization system but also have the complementary advantage of simplicity of elucidation of the features used to categorize data sets to their appropriate groups, though simultaneously carrying to highpoint the comparative implication of varied variables in the concerned scheme. It is mostly challenging to recover explanations for incidences when ANN approaches are utilized because of the “black-box” procedure in ANN (Huang, Zhao, Wei, Wang & Du, 2015). 

The schemes are learned by ANN over iterative learning series of descriptive information, henceforth producing predictions of spectral sorts by sensing anonymous pixels. Decision trees, equally, use dichotomization to direct accessible statistics to the detailed group, as is witnessed in botanical basics. Even if the software produces the difference circumstances at each phase of the decision tree organization, it is probable to examine the circumstances to understand what origin difference has been accomplished (Jelinek, Abawajy, Kelarev, Chowdhury & Stranieri, 2014).

Conclusion

Remote detecting is the knowledge and art of obtaining info (spectral, three-dimensional, temporal) about substances, zones, or occurrences by the examination of information got by a gears from measurements completed at a space, without getting into bodily contact with the substances, space, or occasions under examination. Remote detecting skill makes usage of the full array of the electromagnetic spectrum (EMS). Most of the commercially accessible remote-sensing information are attained in the infrared, visible and microwave wavelength part of the EMS.

Remote identifying as a scheme will deliver data frequently and at a low-priced value to allow, in the suitable period, interference for retrieval of crop condition. Satellite assemblies provide spatially and temporally everlasting annals cover of the sphere. Along through the growth of remote detecting roles, satellite data has become the key data basis to control high-dimension plant condition. With the support of satellite broadcasting and digital image methods, it is simple and also prices effectual in designing and witnessing the plant condition.

References

Abiodun, O. I., Jantan, A., Omolara, A. E., Dada, K. V., Mohamed, N. A., & Arshad, H. (2018). State-of-the-art in artificial neural network applications: A survey. Heliyon, 4(11), e00938. [Online]. Available from: https://www.sciencedirect.com/science/article/pii/S2405844018332067

Benza, M., Weeks, J. R., Stow, D. A., Lopez-Carr, D., & Clarke, K. C. (2016). A pattern-based definition of urban context using remote sensing and GIS. Remote sensing of environment, 183, 250-264. [Online]. Available from:  https://www.sciencedirect.com/science/article/pii/S0034425716302504

Breiman, L. (2017). Classification and regression trees. Routledge. [Online]. Available from: https://content.taylorfrancis.com/books/download?dac=C2009-0-07054-X&isbn=9781351460491&format=googlePreviewPdf

Cao, J., & Lin, Z. (2015). Extreme learning machines on high dimensional and large data applications: a survey. Mathematical Problems in Engineering, 2015. [Online]. Available from: https://www.hindawi.com/journals/mpe/aa/103796/abs/

Dong, Z., Wang, Z., Liu, D., Song, K., Li, L., Jia, M., & Ding, Z. (2014). Mapping wetland areas using Landsat-derived NDVI and LSWI: A case study of West Songnen plain, Northeast China. Journal of the Indian Society of Remote Sensing, 42(3), 569-576. [Online]. Available from:  https://link.springer.com/article/10.1007/s12524-013-0357-1

Gascon, F. (2014, November). Sentinel-2 products and algorithms. In Workshop Preparations for Sentinel (Vol. 2, pp. 25-26). [Online]. Available from:  https://www.eurosdr.net/sites/default/files/images/inline/1-2dag_gascon_ws_sentinel.pdf

Gurney, K. (2014). An introduction to neural networks. CRC press. [Online]. Available from: https://content.taylorfrancis.com/books/download?dac=C2010-0-30102-3&isbn=9781482286991&format=googlePreviewPdf

Hu, F., Xia, G. S., Hu, J., & Zhang, L. (2015). Transferring deep convolutional neural networks for the scene classification of high-resolution remote sensing imagery. Remote Sensing, 7(11), 14680-14707. [Online]. Available from:  https://www.mdpi.com/2072-4292/7/11/14680

Huang, D., Zhao, D., Wei, L., Wang, Z., & Du, Y. (2015). Modeling and analysis in marine big data: advances and challenges. Mathematical Problems in Engineering, 2015. [Online]. Available from: https://www.hindawi.com/journals/mpe/aa/384742/abs/

Jelinek, H., Abawajy, J., Kelarev, A., Chowdhury, M., & Stranieri, A. (2014). Decision trees and multi-level ensemble classifiers for neurological diagnostics. Australian Journal of Medical Science, 1(1), 1-12. [Online]. Available from: https://dro.deakin.edu.au/eserv/DU:30071501/jelinek-decisiontreesand-2014.pdf

Krishna, G., Sahoo, R. N., Pargal, S., Gupta, V. K., Sinha, P., Bhagat, S., … & Chattopadhyay, C. (2014). Assessing wheat yellow rust disease through hyperspectral remote sensing. Int. Arch. Photogramm. Remote Sensing Spatial Inform. Sci, 1413-1416. [Online]. Available from: https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XL-8/1413/2014/isprsarchives-XL-8-1413-2014.pdf

Li, H., Zhang, Z., & Zhao, Z. Z. (2019). Data-Mining for Processes in Chemistry, Materials, and Engineering. Processes, 7(3), 151. [Online]. Available from:  https://www.mdpi.com/2227-9717/7/3/151

Li, P., & Ge, W. (2015). A soft shadow detection method based on MRF for remote sensing images. Mathematical Problems in Engineering, 2015. [Online]. Available from: https://www.hindawi.com/journals/mpe/aa/404095/abs/

Pohl, C., & Van Genderen, J. (2016). Remote sensing image fusion: A practical guide. Crc Press. [Online]. Available from: https://content.taylorfrancis.com/books/download?dac=C2014-0-58049-3&isbn=9781498730037&format=googlePreviewPdf

Rahimzadeh-Bajgiran, P., Omasa, K., & Shimizu, Y. (2012). Comparative evaluation of the Vegetation Dryness Index (VDI), the Temperature Vegetation Dryness Index (TVDI) and the improved TVDI (iTVDI) for water stress detection in semi-arid regions of Iran. ISPRS Journal of Photogrammetry and Remote Sensing, 68, 1-12. [Online]. Available from: https://www.sciencedirect.com/science/article/pii/S0924271611001572

Salmon, J. M., Friedl, M. A., Frolking, S., Wisser, D., & Douglas, E. M. (2015). Global rain-fed, irrigated, and paddy croplands: A new high resolution map derived from remote sensing, crop inventories and climate data. International Journal of Applied Earth Observation and Geoinformation, 38, 321-334. [Online]. Available from: https://www.sciencedirect.com/science/article/pii/S0303243415000240

Sharma, R., Kamble, S. S., & Gunasekaran, A. (2018). Big GIS analytics framework for agriculture supply chains: A literature review identifying the current trends and future perspectives. Computers and Electronics in Agriculture, 155, 103-120. [Online]. Available from: https://www.sciencedirect.com/science/article/pii/S0168169918311311

Szuster, B. W., Chen, Q., & Borger, M. (2011). A comparison of classification techniques to support land cover and land use analysis in tropical coastal zones. Applied Geography, 31(2), 525-532. [Online]. Available from:  https://www.sciencedirect.com/science/article/pii/S0143622810001530

Wang, J., Zheng, T., Lei, P., & Bai, X. (2019). A Hierarchical Convolution Neural Network (CNN)-Based Ship Target Detection Method in Spaceborne SAR Imagery. Remote Sensing, 11(6), 620. [Online]. Available from:  https://www.mdpi.com/2072-4292/11/6/620

Wang, L., Jia, Y., Yao, Y., & Xu, D. (2019). Accuracy Assessment of Land Use Classification Using Support Vector Machine and Neural Network for Coal Mining Area of Hegang City, China. Nature Environment and Pollution Technology, 18(1), 335-341. [Online]. Available from: https://search.proquest.com/openview/21c56e500e0eb5d93f3e4195b9b56408/1?pq-origsite=gscholar&cbl=1606374

Yang, B., Cao, C., Xing, Y., & Li, X. (2015). Automatic classification of remote sensing images using multiple classifier systems. Mathematical Problems in Engineering, 2015. [Online]. Available from: https://www.hindawi.com/journals/mpe/aa/954086/

Ye, F., Luo, W., Dong, M., He, H., & Min, W. (2019). SAR Image Retrieval Based on Unsupervised Domain Adaptation and Clustering. IEEE Geoscience and Remote Sensing Letters. [Online]. Available from: https://ieeexplore.ieee.org/abstract/document/8660481/

Zheng, Y. J., Chen, S. Y., Lin, Y., & Wang, W. L. (2013). Bio-inspired optimization of sustainable energy systems: a review. Mathematical Problems in Engineering, 2013. [Online]. Available from: https://www.hindawi.com/journals/mpe/2013/354523/abs/

Zhong, Y., & Zhang, L. (2011). A new fuzzy clustering algorithm based on clonal selection for land cover classification. Mathematical Problems in Engineering, 2011. [Online]. Available from: https://www.hindawi.com/journals/mpe/2011/708459/abs/