Object-based Image Analysis Classification Research Articles

Quantification of species composition in grass-clover swards using RGB and multispectral UAV imagery and machine learning.

Growing grass-legume mixtures for forage production improves both yield productivity and nutritional quality, while also benefiting the environment by promoting species biodiversity and enhancing soil fertility (through nitrogen fixation). Consequently, assessing legume proportions in grass-legume mixed swards is essential for breeding and cultivation. This study introduces an approach for automated classification and mapping of species in mixed grass-clover swards using object-based image analysis (OBIA). The OBIA procedure was established for both RGB and ten band multispectral (MS) images capturedby an unmanned aerial vehicle (UAV). The workflow integrated structural (canopy heights) and spectral variables (bands, vegetation indices) along with a machine learning algorithm (Random Forest) to perform image segmentation and classification. Spatial k-fold cross-validation was employed to assess accuracy. Results demonstrated good performance, achieving an overall accuracy of approximately 70%, for both RGB and MS-based imagery, with grass and clover classes yielding similar F1 scores, exceeding 0.7 values. The effectiveness of the OBIA procedure and classification was examined by analyzing correlations between predicted clover fractions and dry matter yield (DMY) proportions. This quantification revealed a positive and strong relationship, with R2 values exceeding 0.8 for RGB and MS-based classification outcomes. This indicates the potential of estimating (relative) clover coverage, which could assist breeders but also farmers in a precision agriculture context.

Shallow sea waters habitat mapping in 2018 and 2022 using Sentinel-2 Imagery with OBIA Method at Kelapa Dua Island, Seribu Island, Jakarta Province, Indonesia

Shallow sea habitat has a very important role as a place to live and shelter for various types of organisms. Mapping this area needs to be done regularly considering the many benefits it has and the tendency for this area to change, one of them is by using remote sensing technology with the object-based image analysis (OBIA) method. OBIA classification includes the process of image segmentation and classification of all segmentation results. In the segmentation process using scale 100 at level 1 and scale 10 at level 2. Furthermore, at the classification process, 6 (six) habitat classes are obtained, namely sand, coral, macroalgae, sea grass, rubble and dead coral algae. The overall accuracy value in 2018 showed a value of 70%, and in 2022, it showed a value of 70.77%. The kappa coefficient value is 0.617 in the 2018 image acquisition and 0.612 in the 2022 image acquisition. The analysis carried out showed that there was a change in each habitat with the habitat that experienced the greatest changes are sand and dead coral algae class.

Overcoming the UAS limitations in the coastal environment for accurate habitat mapping

Unmanned Aerial Systems (UAS) have been recently used as a close-range remote sensing tool, providing high spatial and temporal resolutions compared to other remote sensing methods. Their abilities have constantly increased their use in many environmental studies with very satisfactory results. However, the acquisition of reliable information for accurate marine habitat mapping is still challenging due to the plethora of parameters (mostly environmental conditions) that affect the accurate depiction and delineation of habitats in aerial imagery.In this study we are proposing an efficient methodology for marine habitat mapping that includes three main steps: i) data acquisition and processing of UAS imagery, ii) Object-Based Image Analysis (OBIA) and classification of marine habitats, iii) image quality and accuracy assessment of high-resolution thematic maps. This methodology overcomes the UAS limitations in the coastal environment by identifying the optimal acquisition times through the UASea toolbox and fully exploits the potentials of UAS in environmental applications. Moreover, the OBIA workflow and classification combined with the image quality and accuracy assessment demonstrates the importance of such methods in producing very high-resolution and reliable maps of marine habitats.The proposed methodology was implemented in four coastal areas, in Greece, with different characteristics and prevalent environmental conditions, using different UAS with RGB sensors. In particular, the efficiency of the methodology was tested by acquiring data in several conditions in each site and by comparing the results as to the image quality and the classification accuracies. The results showed that the imagery acquired in the proposed acquisition times was of higher quality along with high classification accuracies. This methodology offers an efficient solution for the acquisition of high-resolution and reliable marine information and accurate habitat delineation that can be adaptable in a range of ecological and environmental studies.

A Scalable, Supervised Classification of Seabed Sediment Waves Using an Object-Based Image Analysis Approach

National mapping programs (e.g., INFOMAR and MAREANO) and global efforts (Seabed 2030) acquire large volumes of multibeam echosounder data to map large areas of the seafloor. Developing an objective, automated and repeatable approach to extract meaningful information from such vast quantities of data is now essential. Many automated or semi-automated approaches have been defined to achieve this goal. However, such efforts have resulted in classification schemes that are isolated or bespoke, and therefore it is necessary to form a standardised classification method. Sediment wave fields are the ideal platform for this as they maintain consistent morphologies across various spatial scales and influence the distribution of biological assemblages. Here, we apply an object-based image analysis (OBIA) workflow to multibeam bathymetry to compare the accuracy of four classifiers (two multilayer perceptrons, support vector machine, and voting ensemble) in identifying seabed sediment waves across three separate study sites. The classifiers are trained on high-spatial-resolution (0.5 m) multibeam bathymetric data from Cork Harbour, Ireland and are then applied to lower-spatial-resolution EMODnet data (25 m) from the Hemptons Turbot Bank SAC and offshore of County Wexford, Ireland. A stratified 10-fold cross-validation was enacted to assess overfitting to the sample data. Samples were taken from the lower-resolution sites and examined separately to determine the efficacy of classification. Results showed that the voting ensemble classifier achieved the most consistent accuracy scores across the high-resolution and low-resolution sites. This is the first object-based image analysis classification of bathymetric data able to cope with significant disparity in spatial resolution. Applications for this approach include benthic current speed assessments, a geomorphological classification framework for benthic biota, and a baseline for monitoring of marine protected areas.

Mapping Mature Post-Agricultural Forests in the Polish Eastern Carpathians with Archival Remote Sensing Data

Post-WWII displacements in the Polish Carpathians resulted in widespread land abandonment. Most of the pre-war agricultural areas are now covered with secondary forests, which will soon reach the felling age. Mapping their exact cover is crucial to investigate succession–regeneration processes and to determine their role in the landscape, before making management decisions. Our goal was to map post-agricultural forests in the Polish Eastern Carpathians using archival remote sensing data, and to assess their connectivity with pre-displacement forests. We used German Flown Aerial Photography from 1944 to map agricultural lands and forests from before displacements, and Corona satellite images to map agricultural lands which converted into the forest as a result of this event. We classified archival images using Object-Based Image Analysis (OBIA) and compared the output with the current forest cover derived from Sentinel-2. Our results showed that mature (60–70 years old) post-agricultural forests comprise 27.6% of the total forest area, while younger post-agricultural forests comprise 9%. We also demonstrated that the secondary forests fill forest gaps more often than form isolated patches: 77.5% of patches are connected with the old-woods (forests that most likely have never been cleared for agriculture). Orthorectification and OBIA classification of German Flown Aerial Photographs and Corona satellite images made it possible to accurately determine the spatial extent of post-agricultural forest. This, in turn, paves the way for the implementation of site-specific forest management practices to support the regeneration of secondary forests and their biodiversity.

Residential Population Estimation in Small-Area using LiDAR and Aerial Photograph Data

Population data has an important role in various aspects, such as policy determination, urban planning, and disaster mitigation. But, the most accurate population data in Indonesia is obtained once every 10 years. In this research, population estimation is conducted by applying the Object-Based Image Analysis (OBIA) classification method to detect the residential area. The OBIA classification utilizes aerial photogrammetry data and DSM & DTM of LiDAR. Then the population estimation is generated by the calculation of mathematical demographics and linear regression. Based on the results, OBIA classification produces a high accuracy land use / land cover map, assigned from the accuracy assessment using confusion matrix with kappa coefficient of 0.929 and overall accuracy of 95.24%. While, the habitable surface area classification achieves a high accuracy map with kappa coefficient of 0.86 and overall accuracy of 92.86%. The population estimation results, reveal that the linear regression method has a smaller error than the mathematical demographic method. The MAE, MAPE, RMSE, and RRMSE in Wisma Menanggal values are 40, 21%, 45, dan 0.25, while the ones of Gayungsari Timur are 23, 18%, 34, dan 0.278. In addition to the small error value, the MAE, MAPE, RMSE, and RRMSE values indicate that the population estimation produced by the linear regression model is most optimal.

An Object-Based Image Analysis Approach Using Bathymetry and Bathymetric Derivatives to Classify the Seafloor

In this paper, object-based image analysis classification methods are developed that do not rely on backscatter in order to classify the seafloor. Instead, these methods make use of bathymetry, bathymetric derivatives, and grab samples for classification. The classification is performed on image object statistics. One of the methods utilizes only texture-based features, that is, features that are related to the spatial arrangement of image characteristics. The second method is similar, but relies on a wider set of image object features. The methods were developed and tested using a dataset from Norwegian waters, specifically the Røstbanken area off the coast of Lofoten. The classification results were compared to backscatter-based classification and to grab sample ground-reference data. The algorithm that performed the best was then also applied to a dataset from the Borkumer Stones area close to the island of Schiermonnikoog in Dutch waters. This allowed testing the applicability of the algorithm for different datasets. Because the algorithms that were developed do not require backscatter, the availability of which is much more scarce than bathymetry, and because of the low computational requirements, they could be applied to any area where high-resolution bathymetry and grab samples are available.

Self-adaptive Image Segmentation Optimization for Hierarchal Object-based Classification of Drone-based Images

This study proposes an approach for the quality improvement of feature extraction in unmanned aerial vehicle (UAV)-based images through object-based image analysis (OBIA). A fixed-wing UAV system equipped with an optical (red–green–blue) camera was used to capture very high spatial resolution images over urban and agricultural areas in an arid environment. A self-adaptive image segmentation optimization aided by an orthogonal array from the experimental design was used to optimize and systematically evaluate how OBIA classification results are affected by different settings of image segmentation parameters, feature selection, and single and multiscale feature extraction approaches. The first phase encompassed data acquisition and preparation, which included the planning of the flight mission, data capturing, orthorectification, mosaicking, and derivation of a digital surface model. In the second phase, 25 settings of multiresolution image segmentation (MRS) parameters, namely, scale, shape, and compactness, were suggested through the adoption of an L25 orthogonal array. In the third phase, the correlation-based feature selection technique was used in each experiment to select the most significant features from a set of computed spectral, geometrical, and textural features. In the fourth phase, the ensemble adaptive boosting algorithm (AdaBoost) was used to classify the image objects of segmentation levels in the orthogonal array. The overall accuracy measure (OA) and kappa coefficient (K) were computed to represent a quality indicator of each experiment. The OA and K values ranged from 89% to 95%, whereas the K values ranged from 0.75 to 0.95. The MRS parameter settings that provided the highest classification results (>94%) were analyzed, and class-specific accuracy measures and F-measure were computed. Multiscale AdaBoost classification was conducted on the basis of the computed F-measure values. Results of the multiscale AdaBoost classification demonstrated an improvement in OA, K, and F-measure.

Semi-Automated Analysis of Digital Photographs for Monitoring East Antarctic Vegetation.

Climate change is affecting Antarctica and minimally destructive long-term monitoring of its unique ecosystems is vital to detect biodiversity trends, and to understand how change is affecting these communities. The use of automated or semi-automated methods is especially valuable in harsh polar environments, as access is limited and conditions extreme. We assessed moss health and cover at six time points between 2003 and 2014 at two East Antarctic sites. Semi-automatic object-based image analysis (OBIA) was used to classify digital photographs using a set of rules based on digital red, green, blue (RGB) and hue-saturation-intensity (HSI) value thresholds, assigning vegetation to categories of healthy, stressed or moribund moss and lichens. Comparison with traditional visual estimates showed that estimates of percent cover using semi-automated OBIA classification fell within the range of variation determined by visual methods. Overall moss health, as assessed using the mean percentages of healthy, stressed and moribund mosses within quadrats, changed over the 11 years at both sites. A marked increase in stress and decline in health was observed across both sites in 2008, followed by recovery to baseline levels of health by 2014 at one site, but with significantly more stressed or moribund moss remaining within the two communities at the other site. Our results confirm that vegetation cover can be reliably estimated using semi-automated OBIA, providing similar accuracy to visual estimation by experts. The resulting vegetation cover estimates provide a sensitive measure to assess change in vegetation health over time and have informed a conceptual framework for the changing condition of Antarctic mosses. In demonstrating that this method can be used to monitor ground cover vegetation at small scales, we suggest it may also be suitable for other extreme environments where repeat monitoring via images is required.

Evaluation of Classification Techniques for Land Use Change Mapping of Indian Cities

This study looks into the development of multi-level classification approach for land use change mapping in Indian cities using Landsat imageries. In this study, we mapped 47 Indian cities at different time frames 1990, 2000, 2010, and 2017. We started with traditional classification methods, but results provided unsatisfactory accuracy levels. Thus, we employed multiple classification techniques to achieve results with higher accuracy. The paper captures the evaluation of different classification techniques—hybrid, unsupervised, decision tree classification (DTC), and object-based image analysis (OBIA). The results suggest improvement in accuracy levels by using multi-level classification for different cities at different stages of the classification process. The most prominent is the hybrid classification technique; 14 cities out of 47 reached to accuracy above 72% through hybrid classification. For problematic classes, we used DTC, OBIA, and unsupervised classification techniques after masking the datasets. DTC was used in cities with a greater number of problems in datasets. For example, in the case of Kochi City, the accuracy at the initial level was reported 51% through unsupervised classification which improved to 77% (supervised classification), and finally, it reached 90% by DTC technique. The overall accuracy achieved through the multi-level classification approach described in this paper for the 47 Indian cities ranges from 81 to 93%.

Multi-Sensor UAV Tracking of Individual Seedlings and Seedling Communities at Millimetre Accuracy

The increasing spatial and temporal scales of ecological recovery projects demand more rapid and accurate methods of predicting restoration trajectory. Unmanned aerial vehicles (UAVs) offer greatly improved rapidity and efficiency compared to traditional biodiversity monitoring surveys and are increasingly employed in the monitoring of ecological restoration. However, the applicability of UAV-based remote sensing in the identification of small features of interest from captured imagery (e.g., small individual plants, <100 cm2) remains untested and the potential of UAVs to track the performance of individual plants or the development of seedlings remains unexplored. This study utilised low-altitude UAV imagery from multi-sensor flights (Red-Green-Blue and multispectral sensors) and an automated object-based image analysis software to detect target seedlings from among a matrix of non-target grasses in order to track the performance of individual target seedlings and the seedling community over a 14-week period. Object-based Image Analysis (OBIA) classification effectively and accurately discriminated among target and non-target seedling objects and these groups exhibited distinct spectral signatures (six different visible-spectrum and multispectral indices) that responded differently over a 24-day drying period. OBIA classification from captured imagery also allowed for the accurate tracking of individual target seedling objects through time, clearly illustrating the capacity of UAV-based monitoring to undertake plant performance monitoring of individual plants at very fine spatial scales.

Species discrimination and individual tree detection for predicting main dendrometric characteristics in mixed temperate forests by use of airborne laser scanning and ultra-high-resolution imagery

This study aims to investigate the combined use of two types of remote sensing data — ALS derived and digital aerial photogrammetry data (based on imagery collected by airborne UAV sensors) — along with intensive field measurements for extracting and predicting tree and stand parameters in even-aged mixed forests. The study is located in South West Romania and analyzes data collected from mixed-species plots. The main tree species within each plot are Norway spruce (Picea abies L. Karst.) and Beech (Fagus sylvatica L.).The ALS data were used to extract the digital terrain model (DTM), digital surface model (DSM) and normalized canopy height model (CHM). Object-Based Image Analysis (OBIA) classification was performed to automatically detect and separate the main tree species. A local filtering algorithm with a canopy-height based variable window size was applied to identify the position, height and crown diameter of the main tree species within each plot. The filter was separately applied for each of the plots and for the areas covered with Norway spruce and beech trees, respectively (i.e. as resulted from OBIA classification).The dbh was predicted based on ALS data by statistical Monte Carlo simulations and a linear regression model that relates field dbh for each tree species with their corresponding ALS-derived tree height and crown diameter. The overall RMSE for each of the tree species within all the plots was 5.8 cm for the Norway spruce trees, respectively 5.9 cm for the beech trees. The results indicate a higher individual tree detection rate and subsequently a more precise estimation of dendrometric parameters for Norway spruce compared to beech trees located in spruce-beech even-aged mixed stands. Further investigations are required, particularly in the case of choosing the best method for individual tree detection of beech trees located in temperate even-aged mixed stands.

OBJECT-BASED IMAGE ANALYSIS FOR HISTORIC MAPS CLASSIFICATION

Abstract. Heritage maps represent fundamental information for the study of the evolution of a region, especially in terms of landscape and ecologic features. Historical maps present two kinds of hurdle before they can be used in a modern GIS: they must be geometrically corrected to correspond to the datum in use and they must be classified to exploit the information they contain. This study deals the latter problem: the Historical Cadaster Map, created between 1851 and 1861, for the Trentino region in the North of Italy is available as a collection of maps in the ETRS89/UTM 32N datum. The map is a high resolution scan (230 DPI, 24 bit) of the original map and has been used in several ecological studies, since it provides detailed information not only about land property but also about land use. In the past the cadaster map has been manually digitized and for each area a set of attributes has been recorded. Since this approach is time consuming and prone to errors, automatic and semi-automatic procedures have been tested. Traditional image classification techniques, such as maximum likelihood classification, supervised or un-supervised, pixelwise and contextual, do not provide satisfactory results for many reasons: map colors are very variable within the same area, symbols and characters are used to identify cadaster parcels and locations, lines, drawn by hand on the original map, have variable thickness and colors. The availability of FOSS tools for the Object-based Image Analysis (OBIA) has made possible the application of this technique to the cadaster map. This paper describes the use of GRASS GIS and R for the implementation of the OBIA approach for the supervised classification of the historic cadaster map. It describes the determination of the optimal segments, the choice of their attributes and relevant statistics, and their classification. The result has been evaluated with respect to a manually digitized map using Cohens Kappa and the analysis of the confusion matrix. The result of the OBIA classification has also been compared to the classification of the same map using maximum likelihood classification, un-supervised and supervised, both pixelwise and contextual. The OBIA approach has provided very satisfactory results with the ability to automatically remove the background and symbols and characters, creating a ready to be used classified map. This study highlights the effectiveness of the OBIA processing chain available in the FOSS4G ecosystem, and in particular the added value of the interoperability between GRASS GIS and R.

MAPPING TREES ON FARMLANDS USING OBIA METHOD AND HIGH RESOLUTION SATELLITE DATA: A CASE STUDY OF KORAPUT DISTRICT, ODISHA

Abstract. Agroforestry is an integrated self-sustainable land use management system that is not only capable of producing food from marginal agricultural land but also capable of maintaining and improving the quality of environment. Accurate assessment of trees on farmlands i.e. agroforestry would help in determining their contribution in meeting timber demand and also in carbon sequestration vis-a-vis climate change mitigation. In the present, high resolution multispectral satellite imagery (LISS-IV) has been used for mapping and estimating agroforestry area in Koraput district of Odisha. Both supervised and Object based Image Analysis (OBIA) classifications methods have been applied. In case of supervised maximum likelihood method, those pixels are fully captured where trees exist, whereas in OBIA captures trees according to their crown shapes. This proved OBIA method to be better in identification of trees on farmlands (scattered trees, boundary, and block plantations) than supervised method. This can lead to accurate estimation of area under trees in scattered form, in linear form and also in patch form. Improved results were obtained in case of OBIA classification with more than 90% accuracy. This research implies that remote sensing provide promising tools for evaluating and mapping of agroforestry at district level. Hence, the proposed approach of using high resolution remote sensing data in conjunction with OBIA method would be promising for mapping agroforestry area.

Mapping and monitoring of the land use/cover changes in the wider area of Itanos, Crete, using very high resolution EO imagery with specific interest in archaeological sites

Archaeological site mapping is important for both understanding the history and protecting the sites from excavation during developmental activities. As archaeological sites are generally spread over a large area, use of high spatial resolution remote sensing imagery is becoming increasingly applicable in the world. The main objective of this study is to map the land cover of the Itanos area of Crete and of its changes, with specific focus on the detection of the landscape’s archaeological features. Six satellite images were acquired from the Pleiades and WorldView-2 satellites over a period of 3 years. In addition, digital imagery of two known archaeological sites was used for validation. An object-based image analysis classification was subsequently developed using the five acquired satellite images. Two rule sets were created, one using the standard four bands which both satellites have and another for the two WorldView-2 images with their four extra bands included. Validation of the thematic maps produced from the classification scenarios confirmed a difference in accuracy amongst the five images. Comparing the results of a 4-band rule set versus the 8-band rule set showed a slight increase in classification accuracy using extra bands. The resultant classifications showed a good level of accuracy exceeding 70%. Yet, separating the archaeological sites from the open spaces with little or no vegetation proved to be challenging. This was mainly due to the high spectral similarity between rocks and the archaeological ruins. The high resolution of the satellite data allowed for the accuracy in defining larger archaeological sites, but still there was difficulty in distinguishing smaller areas of interest. The digital image data provided a very good 3D representation for the archaeological sites, assisting as well as in validating the satellite-derived classification maps. To conclude, our study provides further evidence that use of high resolution imagery may allow for archaeological sites to be located, but only where the archaelogical features are of an adequate size.

Mapping Impervious Surfaces in Town–Rural Transition Belts Using China’s GF-2 Imagery and Object-Based Deep CNNs

Impervious surfaces play an important role in urban planning and sustainable environmental management. High-spatial-resolution (HSR) images containing pure pixels have significant potential for the detailed delineation of land surfaces. However, due to high intraclass variability and low interclass distance, the mapping and monitoring of impervious surfaces in complex town–rural areas using HSR images remains a challenge. The fully convolutional network (FCN) model, a variant of convolution neural networks (CNNs), recently achieved state-of-the-art performance in HSR image classification applications. However, due to the inherent nature of FCN processing, it is challenging for an FCN to precisely capture the detailed information of classification targets. To solve this problem, we propose an object-based deep CNN framework that integrates object-based image analysis (OBIA) with deep CNNs to accurately extract and estimate impervious surfaces. Specifically, we also adopted two widely used transfer learning technologies to expedite the training of deep CNNs. Finally, we compare our approach with conventional OBIA classification and state-of-the-art FCN-based methods, such as FCN-8s and the U-Net methods. Both of these FCN-based methods are well designed for pixel-wise classification applications and have achieved great success. Our results show that the proposed approach effectively identified impervious surfaces, with 93.9% overall accuracy. Compared with the existing methods, i.e., OBIA, FCN-8s and U-Net methods, it shows that our method achieves obviously improvement in accuracy. Our findings also suggest that the classification performance of our proposed method is related to training strategy, indicating that significantly higher accuracy can be achieved through transfer learning by fine-tuning rather than feature extraction. Our approach for the automatic extraction and mapping of impervious surfaces also lays a solid foundation for intelligent monitoring and the management of land use and land cover.

Assessment of Segmentation Parameters for Object-Based Land Cover Classification Using Color-Infrared Imagery

Using object-based image analysis (OBIA) techniques for land use-land cover classification (LULC) has become an area of interest due to the availability of high-resolution data and segmentation methods. Multi-resolution segmentation in particular, statistically seen as the most used algorithm, is able to produce non-identical segmentations depending on the required parameters. The total effect of segmentation parameters on the classification accuracy of high-resolution imagery is still an open question, though some studies were implemented to define the optimum segmentation parameters. However, recent studies have not properly considered the parameters and their consequences on LULC accuracy. The main objective of this study is to assess OBIA segmentation and classification accuracy according to the segmentation parameters using different overlap ratios during image object sampling for a predetermined scale. With this aim, we analyzed and compared (a) high-resolution color-infrared aerial images of a newly-developed urban area including different land use types; (b) combinations of multi-resolution segmentation with different shape, color, compactness, bands, and band-weights; and (c) accuracies of classifications based on varied segmentations. The results of various parameters in the study showed an explicit correlation between segmentation accuracies and classification accuracies. The effect of changes in segmentation parameters using different sample selection methods for five main LULC types was studied. Specifically, moderate shape and compactness values provided more consistency than lower and higher values; also, band weighting demonstrated substantial results due to the chosen bands. Differences in the variable importance of the classifications and changes in LULC maps were also explained.

DWELLING EXTRACTION IN REFUGEE CAMPS USING CNN – FIRST EXPERIENCES AND LESSONS LEARNT

Abstract. There is a growing use of Earth observation (EO) data for support planning in humanitarian crisis response. Information about number and dynamics of displaced population in camps is essential to humanitarian organizations for decision-making and action planning. Dwelling extraction and categorisation is a challenging task, due to the problems in separating different dwellings under different conditions, with wide range of sizes, colour and complex spatial patterns. Nowadays, so-called deep learning techniques such as deep convolutional neural network (CNN) are used for understanding image content and object recognition. Although recent developments in the field of computer vision have introduced CNN networks as a practical tool also in the field of remote sensing, the training step of these techniques is rather time-consuming and samples for the training process are rarely transferable to other application fields. These techniques also have not been fully explored for mapping camps. Our study analyses the potential of a CNN network for dwelling extraction to be embedded as initial step in a comprehensive object-based image analysis (OBIA) workflow. The results were compared to a semi-automated, i.e. combined knowledge-/sample-based, OBIA classification. The Minawao refugee camp in Cameroon served as a case study due to its well-organised, clearly distinguishable dwelling structure. We use manually delineated objects as initial input for the training samples, while the CNN network is structured with two convolution layers and one max pooling.

Harnessing the Temporal Dimension to Improve Object-Based Image Analysis Classification of Wetlands

Research related to object-based image analysis has typically relied on data inputs that provide information on the spectral and spatial characteristics of objects, but the temporal domain is far less explored. For some objects, which are spectrally similar to other landscape features, their temporal pattern may be their sole defining characteristic. When multiple images are used in object-based image analysis, it is often constrained to a specific number of images which are selected because they cover the perceived range of temporal variability of the features of interest. Here, we provide a method to identify wetlands using a time series of Landsat imagery by building a Random Forest model using each image observation as an explanatory variable. We tested our approach in Douglas County, Washington, USA. Our approach exploiting the temporal domain classified wetlands with a high level of accuracy and reduced the number of spectrally similar false positives. We explored how sampling design (i.e., random, stratified, purposive) and temporal resolution (i.e., number of image observations) affected classification accuracy. We found that sampling design introduced bias in different ways, but did not have a substantial impact on overall accuracy. We also found that a higher number of image observations up to a point improved classification accuracy dependent on the selection of images used in the model. While time series analysis has been part of pixel-based remote sensing for many decades, with improved computer processing and increased availability of time series datasets (e.g., Landsat archive), it is now much easier to incorporate time series into object-based image analysis classification.

A novel approach for detecting agricultural terraced landscapes from historical and contemporaneous photogrammetric aerial photos

Terraces are the most distinctive sign of human activity on the shape of the Earth surface. Their construction has increased the soils permeability and reduced the slope gradient of mountains since those territories could be exploited both for agricultural and habitable purposes. Over the last decades, they have been the subject of a quick abandonment due to their scarce competitiveness. This has caused some environmental problems, such as soil degradation and hydrological instability. Minori in Italy is one of the most ancient municipalities in the Mediterranean area characterized by the presence of terraces. This paper intends to develop a method for automatic extraction of terraces from historical and contemporaneous aerial photos using an Object-based Image Analysis (OBIA) approach. Historic photos from 1956, acquired by the Geographical Military Institute (IGM), and a contemporaneous block of RGB and multispectral images, taken in 2017 of the study area have been processed to generate a high resolution Digital Elevation Model (DEM) and detailed orthophotos. Subsequently, the OBIA classification has been applied for producing a binary map of terraced and not terraced landscapes for both datasets. Orthophoto resolution was equal to 240 mm, 7 mm and 15 mm for the historical, RGB and multispectral pictures, respectively. DEM resolution results equal to 480 mm and 0.19 mm for the historical and RGB set of data. The R2 between the check points and the estimated values, generated during the metric reconstructions of the two obtained DEMs, resulted equal to 0.99 for both datasets (1956 and 2017). The classification accuracy of the generated binary maps (terraced/not terraced landscapes) were equal to 93% and 98%, respectively. The developed approach looks promising for the historical and contemporaneous datasets. That outcome is essential because it allows to detect terraces position and to compare them over the years in order to analyse their evolution and geomorphological changes.