Multi-temporal Classification Method Research Articles

Improving multi-crop area assessment through Bootstrapping: A focus on tomato fields

Accurate crop mapping and automatic identification of different crops using remote sensing data and methods are vital to modern agriculture and provide farmers with valuable real-time information on crop type, health and yield in a geographic area leading to improved crop management practices and ultimately increasing efficiency, productivity, and profitability. This paper presents a multi-temporal classification method for automatic identification of different crops, especially tomato fields, using optical PlanetScope satellite images. Random Forest algorithm was used to create crop map of the study area and it was combined with the area adjustment algorithms to identify tomato area with a satisfied accuracy. The reliability of the test data used in accuracy assessments was determined using the Bootstrapping algorithm which estimates classification accuracy by creating multiple samples and assessing accuracy on each, yielding a mean accuracy and its variability. The resulting data was then used to adjust the tomato cultivation area, with a 95% confidence interval between 4493.20 ha and 5336.44 ha. The study showed the effectiveness of Random Forest classification for producing precise crop maps from satellite imagery and the utility of Bootstrapping algorithm for ensuring that the test data accurately represents the region, while highlighting the significance of area estimation adjustments and standard deviations for making well-informed decisions about satellite image-based crop maps.

Review on Multitemporal Classification Methods of Satellite Images for Crop and Arable Land Recognition

This paper presents a review of the conducted research in the field of multitemporal classification methods used for the automatic identification of crops and arable land using optical satellite images. The review and systematization of these methods in terms of the effectiveness of the obtained results and their accuracy allows for the planning towards further development in this area. The state of the art analysis concerns various methodological approaches, including selection of data in terms of spatial resolution, selection of algorithms, as well as external conditions related to arable land use, especially the structure of crops. The results achieved with use of various approaches and classifiers and subsequently reported in the literature vary depending on the crops and area of analysis and the sources of satellite data. Hence, their review and systematic conclusions are needed, especially in the context of the growing interest in automatic processes of identifying crops for statistical purposes or monitoring changes in arable land. The results of this study show no significant difference between the accuracy achieved from different machine learning algorithms, yet on average artificial neural network classifiers have results that are better by a few percent than others. For very fragmented regions, better results were achieved using Sentinel-2, SPOT-5 rather than Landsat images, but the level of accuracy can still be improved. For areas with large plots there is no difference in the level of accuracy achieved from any HR images.

Mapping urban expansion using night-time light images from Luojia1-01 and International Space Station

ABSTRACTNight-time light (NTL) images have been proved as a type of reliable data source to map urban expansion. In this paper, to investigate the potential of using multi-source NTL images at near 100 m resolution to detect urban expansion, we use a Luojia1-01 (LJ1-01) image in 2018 and an International Space Station (ISS) night-time image in 2010 in Wuhan city as experiment images. Based on the multiple linear robust regression model, a process of intercalibration between LJ1-01 imagery and ISS imagery is proposed to build a comparable dataset. To detect urban expansion, using the above images at 130 m resolution, Jeffries-Matusita distance is used as an indicator to select the feature combination with the largest class separability. Among all the candidate combinations, the combination of the LJ1-01 image, the simulated LJ1-01 image, and their ratio best meets our requirements for classification. With this feature combination, a multi-temporal classification method based on Support Vector Machines and Back Propagating (BP) – Neural Network, respectively, is utilized to classify the study area into stable non-urban class, stable urban class, and expanding the urban class. The results of the multi-temporal classification show that the overall accuracy is around 90%, and the Kappa coefficients are larger than 0.84. For each class, the user’s accuracy is larger than 87%, and the producer’s accuracy is larger than 83%. The results of this study indicate that it is feasible to detect urban expansion by using multi-source NTL images at near 100 m resolution.

A New Crop Classification Method Based on the Time-Varying Feature Curves of Time Series Dual-Polarization Sentinel-1 Data Sets

Multitemporal Sentinel-1 data sets are suitable for high-precision agricultural classification mapping due to its short revisit period and dual-polarization channels. At present, more and more attention has been paid to the multitemporal classification methods with feature curve matching, because the time-varying polarimetric characteristics show great potential to crop classification. However, current methods only use the variation of single intensity feature, and the indicators for evaluating similarity have not considered the effect of the variable growing seasons of different parcels. Based on this, a new method with feature curve matching is proposed, which uses the combination of multiple features and applies the discrete Fréchet distance and the Pearson distance to evaluate the similarity between two curves. The proposed method applies time-series Sentinel-1 images for crop classification in two study areas of Gansu province, China. The results show that the overall accuracies in two study areas of the proposed method are 94.98% and 90.20%, respectively. This method achieves higher classification accuracies, compared with the SVM classification method and some other methods with feature curve matching.

A co-training, mutual learning approach towards mapping snow cover from multi-temporal high-spatial resolution satellite imagery

High-spatial and -temporal resolution snow cover maps for mountain areas are needed for hydrological applications and snow hazard monitoring. The Chinese GF-1 satellite is potential to provide such information with a spatial resolution of 8m and a revisit of 4days. The main challenge for the extraction of multi-temporal snow cover from high-spatial resolution images is that the observed spectral signature of snow and snow-free areas is non-stationary in both spatial and temporal domains. As a result, successful extraction requires adequate labelled samples for each image, which is difficult to be achieved. To solve this problem, a semi-supervised multi-temporal classification method for snow cover extraction (MSCE) is proposed. This method extends the co-training based algorithms from single image classification to multi-temporal ones. Multi-temporal images in MSCE are treated as different descriptions of the same land surface, and consequently, each pixel has multiple sets of features. Independent classifiers are trained on each feature set using a few labelled samples, and then, they are iteratively re-trained in a mutual learning way using a great number of unlabelled samples. The main principle behind MSCE is that the multi-temporal difference of land surface in spectral space can be the source of mutual learning inspired by the co-training paradigm, providing a new strategy to deal with multi-temporal image classification. The experimental findings of multi-temporal GF-1 images confirm the effectiveness of the proposed method.

Conditional Random Fields for Multitemporal and Multiscale Classification of Optical Satellite Imagery

In this paper, we present a method for the multitemporal and contextual classification of georeferenced optical remote sensing images acquired at different epochs and having different geometrical resolutions. The method is based on Conditional Random Fields (CRFs) for contextual classification. The CRF model is expanded by temporal interaction terms that link neighboring epochs via transition probabilities between different classes. In order to be able to deal with data of different resolution, the class structure at different epochs may vary with the resolution. The goal of the multitemporal classification is an improved classification performance at all individual epochs, but also the detection of land-cover changes, possibly using lower resolution data. This paper also contains a comparison of the performance of different models for the interaction potentials. Results are given for two different test sites in Germany, where Ikonos, RapidEye, and Landsat images are available. Our results show that the multitemporal classification does indeed increase the overall accuracy of all epochs compared to a monotemporal classification and to a state-of-the-art multitemporal classification method, and that it is feasible to detect changes in lower resolution images.

Observing the Middle Elbe Biosphere in Germany by Means of TerraSAR-X Images

The Lower Saxonian Elbe Valley Biosphere Reserve is part of the UNESCO Biosphere Reserve “Elbe River Landscape”, and used mainly for agriculture. One of tasks of the Biosphere Reserve Administration is to develop sustainable forms of land use which requires comprehensive updated land cover maps. Land use maps are hard to produce because of surveying costs and time. Nevertheless, these large areas need to be monitored. TerraSAR-X images are used to establish agricultural land use maps. In this study, two areas are selected within the Elbe Biosphere Reserve situated around the oxbows Wehninger Werder and Walmsburger Werder. Multi temporal classification methods were used to identify the different crops using maximum likelihood classifier for the years 2010 and 2011. The crop classifications were used to evaluate the effect of the number of images, the necessity of polarizations, and the consequences of some missing images within the crop calendar. These classifications were analyzed to estimate producer accuracy and Kappa index for each crop besides the overall accuracy for each agricultural land use map. The study shows that using dual polarization imagery enhances producer accuracies for many crops over the single polarization imagery, and demonstrates the importance of using frequent images during the cultivation period.

A generalized fuzzy Markov chain-based model for classification of remote-sensing multitemporal images

In this work we propose a conceptual generalization of the cascade classification, fuzzy Markov chain-based method introduced in earlier studies. Such generalization, which is based on the assumption of the invertibility of the fuzzy Markov classification model with respect to time, leads to a model that can classify image objects at two points in time simultaneously. We start by defining two temporal modes of operation. In the forward mode a temporal transformation, supported by a transition possibility matrix T, projects an image object’s fuzzy classification for time t into time t + 1 and fuses the updated membership values with the object’s fuzzy classification for time t + 1. In the backward mode the transition matrix is inverted and the fuzzy classification for time t + 1 is updated backwards, i.e. projected into time t. Furthermore, we tackle a key problem with respect to the application of fuzzy Markov chains in remote-sensing data classification: the estimation of transition possibility values. Previously, transition possibilities estimation in the context of fuzzy Markov chain-based multitemporal classification methods has been carried out with the aid of stochastic methods – specifically, through genetic algorithms. In this work we propose an analytical, least squares-based estimation technique, as a more stable and computational efficient alternative to the stochastic approach. Finally, we report on the application of the multitemporal method in the classification of two different test sites – rural and urban – covered by images produced by medium and high resolution orbital, optical sensor systems.

An integrated spatio-temporal classification method for urban fringe change detection analysis

The urban fringe is the transition zone between urban land use and rural land use. It represents the most active part of the urban expansion process. Change detection using multi-temporal imagery is proven to be an efficient way to monitor land-use/land-cover change caused by urban expansion. In this study, we propose a new multi-temporal classification method for change detection in the urban fringe area. The proposed method extracts and integrates spatio-temporal contextual information into multi-temporal image classification. The spatial information is extracted by object-oriented image segmentation. The temporal information is modelled with temporal trajectory analysis with a two-step calibration. A probabilistic schema that employs a global membership function is then used to integrate the spectral, spatial and temporal information. A trajectory accuracy measurement is proposed to assist the comparison on the performances of the integrated spatio-temporal method and classical pixel- and ‘snapshot’-based classification methods. The experiment shows that the proposed method can significantly improve the accuracies of both single scene classification and temporal trajectory analysis.

Hydrological differentiation and spatial distribution of high altitude wetlands in a semi-arid Andean region derived from satellite data

Abstract. High Altitude Wetlands of the Andes (HAWA) belong to a unique type of wetland within the semi-arid high Andean region. Knowledge about HAWA has been derived mainly from studies at single sites within different parts of the Andes at only small time scales. On the one hand, HAWA depend on water provided by glacier streams, snow melt or precipitation. On the other hand, they are suspected to influence hydrology through water retention and vegetation growth altering stream flow velocity. We derived HAWA land cover from satellite data at regional scale and analysed changes in connection with precipitation over the last decade. Perennial and temporal HAWA subtypes can be distinguished by seasonal changes of photosynthetically active vegetation (PAV) indicating the perennial or temporal availability of water during the year. HAWA have been delineated within a region of 12 800 km2 situated in the Northwest of Lake Titicaca. The multi-temporal classification method used Normalized Differenced Vegetation Index (NDVI) and Normalized Differenced Infrared Index (NDII) data derived from two Landsat ETM+ scenes at the end of austral winter (September 2000) and at the end of austral summer (May 2001). The mapping result indicates an unexpected high abundance of HAWA covering about 800 km2 of the study region (6 %). Annual HAWA mapping was computed using NDVI 16-day composites of Moderate Resolution Imaging Spectroradiometer (MODIS). Analyses on the relation between HAWA and precipitation was based on monthly precipitation data of the Tropical Rain Measurement Mission (TRMM 3B43) and MODIS Eight Day Maximum Snow Extent data (MOD10A2) from 2000 to 2010. We found HAWA subtype specific dependencies on precipitation conditions. A strong relation exists between perennial HAWA and snow fall (r2: 0.82) in dry austral winter months (June to August) and between temporal HAWA and precipitation (r2: 0.75) during austral summer (March to May). Annual changes in spatial extend of perennial HAWA indicate alterations in annual water supply generated from snow melt.

Cascade multitemporal classification based on fuzzy Markov chains

This paper proposes a new fuzzy cascade multitemporal classification method based on Fuzzy Markov Chains. This method differs from prior fuzzy multitemporal approaches proposed thus far, as the method does not require the knowledge of the true class at an earlier date; instead it uses the attributes of the image object being classified at the earlier date. This method combines the fuzzy, non-temporal, classification of a geographical region at two points in time to provide a single unified result. A transformation law based on class transition possibilities projects the earlier classification onto the later date before combining both results. Performance analysis was conducted upon a sequence of three LANDSAT images from the central region of Brazil using a Genetic Algorithm to estimate transition possibilities. The analysis showed that the increase in performance is highly dependent on whether or not a significant correlation exists between the temporal data sets, as well as on the accuracy of the monotemporal classifier at the earlier date. While the monotemporal approach used in the experiments attained an average class accuracy of approximately 55%, the multitemporal scheme achieved between 65% and 95%. Similar results in terms of overall accuracy were also observed. Furthermore, compared to two alternative cascade multitemporal classification approaches, the proposed method consistently showed better results.

Use of a dark object concept and support vector machines to automate forest cover change analysis

An automated method was developed for mapping forest cover change using satellite remote sensing data sets. This multi-temporal classification method consists of a training data automation (TDA) procedure and uses the advanced support vector machines (SVM) algorithm. The TDA procedure automatically generates training data using input satellite images and existing land cover products. The derived high quality training data allow the SVM to produce reliable forest cover change products. This approach was tested in 19 study areas selected from major forest biomes across the globe. In each area a forest cover change map was produced using a pair of Landsat images acquired around 1990 and 2000. High resolution IKONOS images and independently developed reference data sets were available for evaluating the derived change products in 7 of those areas. The overall accuracy values were over 90% for 5 areas, and were 89.4% and 89.6% for the remaining two areas. The user's and producer's accuracies of the forest loss class were over 80% for all 7 study areas, demonstrating that this method is especially effective for mapping major disturbances with low commission errors. IKONOS images were also available in the remaining 12 study areas but they were either located in non-forest areas or in forest areas that did not experience forest cover change between 1990 and 2000. For those areas the IKONOS images were used to assist visual interpretation of the Landsat images in assessing the derived change products. This visual assessment revealed that for most of those areas the derived change products likely were as reliable as those in the 7 areas where accuracy assessment was conducted. The results also suggest that images acquired during leaf-off seasons should not be used in forest cover change analysis in areas where deciduous forests exist. Being highly automatic and with demonstrated capability to produce reliable change products, the TDA-SVM method should be especially useful for quantifying forest cover change over large areas.