Accurate information on urban tree species distribution can reveal insights into how street trees provide ecosystem services like mitigating air pollution and cooling surfaces. Here, we used LiDAR-derived structural properties of individual tree crowns (ITCs) and digital aerial images to classify urban tree species. We fused LiDAR features with RGB-NIR digital aerial images using a fully convolutional neural network. The fusion strategy consisted in stacking one LiDAR feature at a time with RGB-NIR bands. The results show that surface normals of tree leaves improve the F1-score of all species, with the highest increase reaching 13.7 percentage points.

When it comes to technology in agriculture, one of the most important aspects is farmland crop monitoring. However, in most cases, only the main crops are needed to be monitored by satellite images, due to their high territorial extension. Therefore, a semantic segmentation model for identifying plantations should correctly classify the majority classes and also automatically identify other unknown crops. Open set recognition (OSR) aims to embrace both of these causes, so that the model can be more robust in the wild. This work adapts the framework of outlier exposure (OE) for open set image segmentation. OE was evaluated by adding it to three distinct methods for open set segmentation: softmax thresholding, OpenPCS and OpenPCS++. We conducted several experiments to enrich the discussion of the impact of OE on the semantic segmentation of crop imagery. Our methodology achieved a consistent increase for OpenPCS and OpenPCS++ methods, with an improvement of up to 7.5% in terms of area under the receiver operating characteristic (AUROC) curve if compared to previous work.

Most of the current deforestation detection systems rely on cloud-free optical images, which are difficult to obtain in tropical regions. A synthetic aperture radar (SAR) is nearly unaffected by clouds, thus providing valuable insights for deforestation detection. In cloud-free conditions, the use of optical images usually provides better results than the use of SAR data alone. Optical-SAR fusion has been hailed as a promising way to improve deforestation detection. However, it was poorly investigated, particularly when optical images are affected by clouds. This letter employs optical-SAR fusion strategies to improve the classification accuracy of clear-cut deforestation in the Brazilian Amazon under diverse cloud conditions. Sentinel-1 and Sentinel-2 images were fused using fully convolutional networks (FCNs) and early, joint, and late fusion (LF) strategies. Experiments showed that the optical-SAR fusion outperforms the single-modality (optical or SAR) variants for deforestation detection on pixels affected by clouds. The joint fusion strategy provided the best results in all cloud cover scenarios.

Multitemporal remote-sensing images play a key role as a source of information for automated crop mapping and monitoring. The spatial/spectral pattern evolution along time provides information about the dynamics of the crops and are very useful for productivity estimation. Although the multitemporal mapping of crops has progressed considerably with the advent of deep learning in recent years, the classification models obtained still have limitations when exposed to unknown classes in the prediction phase, reducing their usefulness. In other words, these models are trained to identify a closed set of crops (e.g., soy and sugar cane) and are therefore unable to recognize other types of crops (e.g., maize). In this letter, we deal with the challenges of multitemporal crop recognition by proposing a new approach called OpenPCS++ that is not only able to learn known classes but is also capable of identifying new crops in the predicting phase. The proposed approach was evaluated in two challenging public datasets located in tropical climates in Brazil. Results showed that OpenPCS++ achieved increases of up to 0.19 in terms of area under the receiver-operating characteristic (ROC) curve in comparison with baselines. Code is available athttps://github.com/DiMorten/osss-mcr.

Unmanned aerial vehicles (UAVs) allow on-demand imaging of orchards at an unprecedented level of detail. The automated detection of plantation rows in the images helps in the successive analysis steps, such as the detection of individual fruit trees and planting gaps, aiding producers with inventory and planting operations. Citrus trees can be planted in curved rows that form intricate geometric patterns in aerial images, requiring robust detection approaches. While deep learning methods rank among state-of-the-art methods for segmenting images with particular geometrical patterns, they struggle to hold their performance when testing data differs much from training data (e.g., image intensity differences, image artifacts, vegetation characteristics, and landscape conditions). In this letter, we propose a method to learn geometric features of orchards in UAV images and use them to improve the detection of plantation rows. First, we train a detection encoder–decoder network (DetED) to segment planting rows in RGB images. Then, with labeled data, we train an encoder–decoder correction network (CorrED) that learns to map binary masks with spurious row segmentation geometries into corrected ones. Finally, we use the CorrED network to fix geometric inconsistencies in DetED outcome. Our experiments with commercial plantations of orange trees show that the proposed CorrED postprocessing can restore missing segments of plantation rows and improve detection accuracy in testing data.

Conditional (cGAN)-based methods proposed so far for synthetic aperture radar (SAR)-to-optical image synthesis tend to produce noisy and unsharp optical outcomes. In this work, we propose the atrous-cGAN, a novel cGAN architecture that improves the SAR-to-optical image translation. The proposed generator and discriminator networks rely on atrous convolutions and incorporate an atrous spatial pyramid pooling (ASPP) module to enhance fine details in the generated optical image by exploiting spatial context at multiple scales. This letter reports experiments carried out to assess the performance of atrous-cGAN for the synthesis of Landsat-8 images from Sentinel-1A data based on three public data sets. The experimental analysis indicated that the atrous-cGAN consistently outperformed the classical pix2pix counterpart in terms of visual quality, similar to the true optical image, and as a feature learning tool for semantic segmentation.

Many deep-learning (DL)-based, domain adaptation (DA) methods for remote sensing (RS) applications rely on adversarial training strategies to align features extracted from images of different domains in a shared latent space. However, the performance of such representation matching techniques is negatively impacted when class occurrences in the target domain, for which no labeled data are available during training, are highly imbalanced. In this work, we propose a DL-based representation matching approach for DA in the context of change detection tasks. We further evaluate the approach in a deforestation mapping application, characterized by a high-class imbalance between the deforestation and no-deforestation classes. The domains represent different sites in the Amazon and Brazilian Cerrado biomes. To mitigate the class imbalance problem, we devised an unsupervised pseudolabeling scheme based on change vector analysis (CVA) that prevents the feature alignment to be biased toward the overrepresented class. The experimental results indicate that the proposed approach can improve the accuracy of cross-domain deforestation detection.

Early detection of deforestation processes is vital to maintain and regulate tropical rainforests, such as in the Amazon region. Most of them rely on optical imagery. Approaches based on Synthetic Aperture Radar (SAR) data are comparatively unexplored, in particular for deforestation detection in tropical rainforests. This work addresses this gap and evaluates Fully Convolutional Networks based on the U-Net, Res-Unet and Siamese Network, for deforestation detection using images from three different sensors, Landsat-8, Sentinel-2, and Sentinel-1. Experiments conducted on a dataset of the Amazon rainforest indicated that Fully Convolutional Networks working on Sentinel-1 data can achieve sufficient accuracy for detecting deforestation in tropical rainforests when clouds prevent the use of optical data11The source code is available in https://github.zcom/MabelOrtega/Comparison-of-Optical-and-SAR-data-for-deforestation-mapping-in-the-Amazon-Forest-with-FCN.

This paper introduces a novel hybrid N-to-N bidirectional ConvLSTM for multi-temporal crop recognition in areas characterized by highly complex crop dynamics as in tropical regions. The proposed method seeks to explore both spatial and temporal context of the data by adding a bidirectional ConvLSTM to every skip connection in the U-Net and applying the fully connected conditional random fields (CRF) methodology as a post-processing of the predictions. We evaluated our method on a publicly available tropical region dataset, from Sentinel-1 data, obtaining an average improvement of almost 4% in the Average F1 Score and 1.7% in Overall Accuracy, when comparing with a state-of-the art method.

Remote sensing imagery from different sensors enables accurate crop monitoring and mapping, supporting efficient and sustainable agricultural practices. This paper proposes a flexible multi-modal Encoder-Decoder architecture to implement and investigate different fusion strategies for crop segmentation using synthetic-aperture radar (SAR) and optical image sequences. The trained model handles each modal individually or both, allowing us to investigate scenarios where one modal is not available. Our architecture consists of two modality-specific encoders, a shared decoder, a fusion module, and three classifiers, one for each modal and one for the fusion output. Also, we propose using a partial loss function that allows training the network with scarce ground truths. The proposed approach was evaluated in a public dataset comprising seven multitemporal SAR images and four multitemporal optical images from a tropical agricultural region in Brazil. We report results for feature and decision fusion strategies and discuss the benefits of using each of them for multi-modal crop segmentation.

Reliable crop mapping is an essential tool for agricultural monitoring and food security. In the tropics, where cloud cover massively affects optical imagery, synthetic-aperture radar (SAR) imagery emerged as a cost-effective alternative for discriminating crops in large scale agricultural regions. Recently, unsupervised deep clustering approaches have emerged as a competitive alternative in several different applications with labeling restrictions. This paper explores this literature and evaluates the feasibility of such methods applied to crop classification in a tropical region from multi-temporal SAR image sequences. We focus on the k-Means-related deep clustering methods, specifically, on Deep Embedding Clustering and Deep K-Means. We report experiments conducted on a public dataset from a tropical region with highly complex crop dynamics. In our experiments the tested unsupervised approaches managed to deliver nearly 78% of supervised counterparts for this task11The source codes are available at https:/github.com/DLoboT/Project_DL_2020.

The synthesis of realistic data using deep learning techniques has greatly improved the performance of classifiers in handling incomplete data. Remote sensing applications that have profited from those techniques include translating images of different sensors, improving the image resolution and completing missing temporal or spatial data such as in cloudy optical images. In this context, this letter proposes a new deep-learning-based framework to synthesize missing or corrupted multispectral optical images using multimodal/multitemporal data. Specifically, we use conditional generative adversarial networks (cGANs) to generate the missing optical image by exploiting the correspondent synthetic aperture radar (SAR) data with a SAR-optical data from the same area at a different acquisition date. The proposed framework was evaluated in two land-cover applications over tropical regions, where cloud coverage is a major problem: crop recognition and wildfire detection. In both applications, our proposal was superior to alternative approaches tested in our experiments. In particular, our approach outperformed recent cGAN-based proposals for cloud removal, on average, by 7.7% and 8.6% in terms of overall accuracy and F1-score, respectively.

Agriculture monitoring is a key task for producers, governments and decision makers. The analysis of multitemporal remote sensing data provides a cost-effective way to perform this task. Recurrent Neural Networks (RNNs) have been successfully used in temporal modeling problems, while Convolutional Neural Networks (CNNS) are the state-of-the-art in image classification, mainly due to their ability to capture spatial context. In this work, we propose the use of a hybrid network architecture for crop mapping that combines RNNs and CNNs. We evaluate this architecture experimentally upon a Sentinel-1A database from a tropical region in Brazil. The ability of recurrent networks to model temporal context is compared with the conventional image stacking approach. The impact of using CNN learned features rather than context aware handcrafted features is also investigated. In our analysis the hybrid architecture achieved better average class accuracy than alternative approaches based on image stacking and GLCM features.

This work presents a dense fully convolutional architecture for crop type recognition from multitemporal RS images. Basically, we adapted a dense fully convolutional net to deal with stacks of multitemporal data. The proposed approach was tested upon a public dataset comprising two Sentinel-1A sequences from a tropical region in South America. We took as baseline a dense convolutional network designed for patch classification. Thematic and spatial accuracy, as well as the computational load were evaluated experimentally. The proposed architecture matched the baseline in terms of recognition rates and proved to be very efficient computationally in the inference phase.

In tropical/subtropical regions, the favorable climate associated with the use of agricultural technologies, such as no tillage, minimum cultivation, irrigation, early varieties, desiccants, flowering inducing, and crop rotation, makes agriculture highly dynamic. In this letter, we present the Campo Verde agricultural database. The purpose of creating and sharing these data is to foster advancement of remote sensing technology in areas of tropical agriculture, primarily the development and testing of methods for crop recognition and agricultural mapping. Campo Verde is a municipality of Mato Grosso state, localized in the Cerrado (Brazilian Savanna) biome, in central west Brazil. Soybean, maize, and cotton are the primary crops cultivated in this region. Double cropping systems are widely adopted in this area. There is also livestock and forestry production. Our database provides the land-use classes for 513 fields by month for one Brazilian crop year (between October 2015 and July 2016). This information was gathered during two field campaigns in Campo Verde (December 2015 and May 2016) and by visual interpretation of a time series of Landsat-8/Operational Land Imager (OLI) images using an experienced interpreter. A set of 14 preprocessed synthetic aperture radar Sentinel-1 and 15 Landsat-8/OLI mosaic images is also made available. It is important to promote the use of radar data for tropical agricultural applications, especially because the use of optical remote sensing in these regions is hindered by the high frequency of cloud cover. To demonstrate the utility of our database, results of an experiment conducted using the Sentinel-1 data set are presented.

This work presents a spatio-temporal Conditional Random Field (CRF) based model for crop recognition from multi-temporal remote sensing image sequences. The association potential at each image site is based on the class posterior probabilities computed by a Random Forest (RF) classifier given the features at the corresponding site. A contrast-sensitive Potts model is used as a label smoothing method in the spatial domain, whereas the interactions in the temporal domain are modeled based on expert knowledge about the possible transitions between adjacent epochs. The CRF based model was tested for crop mapping in two subtropical areas based on a sequences of 9 Landsat and 14 Sentinel-1 images from Ipuã, São Paulo and Campo Verde, Mato Grosso, respectively, two municipalities in Brazil. The experiments showed significant improvements of the accumulated F1 score per class against a mono-temporal CRF approach of up to 50% and 75% for a total of 8 and 11 classes using Optical and SAR images respectively.

The vast amount of data obtained from current remote sensing data acquisition technologies represents a wealth of useful and affordable geospatial data for policy and decision makers. However, the consequent computational cost of analyzing these data may become prohibitive. This letter extends previous efforts in exploiting distributed processing to speed up the image interpretation process. In this letter, we propose and evaluate a mechanism to exploit task parallelism in addition to data parallelism. Experiments conducted on cloud computing infrastructure, following an object-based interpretation model, demonstrated that substantial performance gains can be obtained with the proposed mechanism.

This letter evaluates metaheuristics for the supervised parameter tuning of multiresolution-region-growing segmentation. Three groups of metaheuristics are tested in terms of convergence speed and solution quality. Generalized pattern search, mesh adaptive direct search, and Nelder-Mead represent the single-solution group. Differential evolution (DE) represents the population group. DE followed by each of the aforementioned single-solution metaheuristics represents the hybrid metaheuristic group. This letter reveals that the optimization objective functions typically have countless local minima, many of them leading to very poor solutions. Experiments on three data sets demonstrated that single-solution-based methods often lead to a solution with unacceptable quality. DE was less susceptible to be stuck in local minima when compared to single-solution methods, but it was slower in reaching the minima. Moreover, hybrid methods presented the best tradeoff between accuracy and convergence speed.

The Segmentation Parameter Tuner (SPT) is a tool designed for automatic tuning of segmentation parameters. In SPT, the goodness of a set of parameter values is given by the level of agreement between the segmentation result and a given reference (representing the desired outcome) quantified by a metric selected by the user (empirical discrepancy methods). This metric is used as the fitness function of an optimization algorithm that searches the parameter space for the minimum value, which is expected to correspond to the segmentation outcome most similar to the reference. SPT 3.1 offers many interesting features such as: five segmentation algorithms (for Optical, Hyperspectral and SAR images), four optimization algorithms (stochastic and direct search optimization methods) and seven discrepancy metrics (pixel and object-based). This paper describes the optimization procedure underlying SPT 3.1, the features added to this version as well as an experiment that illustrates the operation of the tool.

Advances in remote sensors are providing exceptional quantities of large-scale data with increasing spatial, spectral and temporal resolutions, raising new challenges in its analysis, e.g. those presents in classification processes. This work presents the architecture of the InterIMAGE Cloud Platform (ICP): Data Mining Package; a tool able to perform supervised classification procedures on huge amounts of data, on a distributed infrastructure. The architecture is implemented on top of the MapReduce framework. The tool has four classification algorithms implemented taken from WEKA's machine learning library, namely: Decision Trees, Naïve Bayes, Random Forest and Support Vector Machines. The SVM classifier was applied on datasets of different sizes (2 GB, 4 GB and 10 GB) for different cluster configurations (5, 10, 20, 50 nodes). The results show the tool as a potential approach to parallelize classification processes on big data.

Image segmentation is a critical step in image analysis, and usually involves a high computational cost, especially when dealing with large volumes of data. Given the significant increase in the spatial, spectral and temporal resolutions of remote sensing imagery in the last years, current sequential and parallel solutions fail to deliver the expected performance and scalability. This work proposes a scalable and efficient segmentation method, capable of handling efficiently very large high resolution images. The proposed solution is based on the MapReduce model, which offers a highly scalable and reliable framework for storing and processing massive data in cloud computing environments. The solution was implemented and validated using the Hadoop platform. Experimental results attest the viability of performing region growing segmentation in the MapReduce framework.

Satellite acquisitions from LANDSAT (LS) and CBERS programs are widely used in monitoring land cover dynamics. In the acquired products, clouds form opaque objects are obscuring parts of the scene and preventing a reliable extraction of information from these areas. Consequently, cloud shadows create similar problems, as the reflected intensity of the shadowed areas is highly reduced, generating additional info gaps. The problem can be handled by replacing clouds/shadows pixels from other close-date acquisitions, but that would assume a prior knowledge of the spatial distribution of clouds and their corresponding shadows in a scene. This research introduces a method that provides the clouds/shadows layers and their percentage in LS (TM & ETM+) and CBERS (HRCC) scenes. The approach relies on a set of literature indicators to create a composite image that enhances the visual differentiation of clouds/shadows from other objects. The created composite RGB are then warped to a relative luminance raster calculated from the linear bands components. Afterwards, the raster is processed by a K-means unsupervised classifier with a definite number of classes in order to isolate the target-layer pixels. Next, the statistical mode for the population of each class is calculated, compared and used to select the cloud/shadow class automatically, and finally the results are refined by a set of morphological filters. The processing chain avoids the usage of thresholds and highly reduces the user intervention. The achieved outcomes on various test cases are promising and stable, and encourage further developments.

Hyperspectral imaging is a new technique in remote sensing that collects hundreds of images at differents wavelength values for the same area of the Earth. For instance the Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) sensor of NASA capable to obtain 224 spectral channels in a wavelength range between 40 and 250 nanometers. As a result each pixel of the image can be represented as a spectral signature. Image segmentation is the process of dividing a digital image into groups of pixels or objects. Hyperspectral image classification is an important and active area dedicated to identifying each pixel in the image with an exclusive material/object class. Several efforts had been done in this field using spectral and spatial information separately or simultaneously in order to improve the performance of the classification techniques. In this work we have developed a new technique that uses a segmentation algorithm to post-process the classification results obtained using a widely used classifier such as the support vector machine (SVM). Experimental results with a real hyperspectral data set collected over the city of Pavia, Italy, are provided.

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.

In this work we compare empirically five cascade classification schemes based on Support Vector Machines. Data fusion as well as decision fusion variants are considered. Data fusion is implemented by simply stacking feature vectors, whereas decision fusion is performed by a multitemporal SVM classifier, which classifies input patterns consisting of probability vectors produced by monotemporal SVMs. The exploitation of prior knowledge in terms of possible class transitions is a further aspect investigated in the present paper. The analysis is conducted upon a pair of IKONOS images from Rio de Janeiro, Brazil. The study reveals that a considerable accuracy improvement may be brought by the multitemporal approaches regarding their monotemporal counterparts. In particular, for the decision fusion schemes, the improvement is highly dependent on the relative accuracy of the monotemporal classifiers, whose individual decisions are combined to produce a consensual decision.

The characterization of urban areas can be improved considerably by combining spectral and spatial features. As a matter of fact, depending on objects of interest in a specific application, the exploitation of both types of features at multiple spatial resolutions is required. This paper proposes a decision fusion method that relies on both spectral and textural features. The proposed approach is able to produce different classification results based on distinct partitions of the same input data set. Experiments conducted on CBERS-2B data demonstrate a significant performance improvement brought by the combination of spectral and textural features in comparison to the use of only spectral features to describe the image objects.

The key component of a fuzzy Markov chain (FMC)-based multitemporal cascade classifier is the transition possibility matrix (TPM). Such matrix represents the temporal dynamics of the land use/land cover classes in the target site in a given time period. The choice of the TPM estimation approach is a crucial step in the design of FMC-based classifiers, as it strongly influences the final classification accuracy. Moreover, the task of collecting training data may involve considerable effort, since the number of transitions to be represented grows with the square of the number of classes in the application. In spite of their relevance, the TPM estimation has only been addressed superficially in previous publications about FCM-based classification methods. In this letter, we concern some of those aspects and investigate alternative ways of the TPM estimation. Experimental analysis on a multitemporal data set covering a 20-year period sheds light on the conditions under which those alternative estimation approaches may be used, as well as on their impact over the classification performance.

This letter proposes a parallel version for graphics processing units (GPU) of a region-growing image segmentation algorithm widely used by the geographic object-based image analysis (GEOBIA) community. Initially, all image pixels are considered as seeds or primitive segments. Fine-grained parallel threads assigned to individual pixels merge adjacent segments iteratively always ensuring to minimize the overall heterogeneity increase. Besides spectral features the merging criterion considers morphological features that can be efficiently computed in the underlying GPU architecture. Two alternatives using different merging criteria are proposed and tested. An experimental analysis upon five different test images has shown that the parallel algorithm may run up to 19 times faster than its sequential counterpart.

This letter investigates the use of rotation invariant descriptors based on Local Binary Patterns (LBP) and Local Phase Quantization (LPQ) for texture characterization in the context of land-cover and land-use classification of Remote Sensing (RS) optical image data. Very high resolution images from the IKONOS-2 and Quickbird-2 orbital sensor systems covering different urban study areas were subjected to classification through an object-based approach. The experiments showed that the discrimination capacity of LBP and LPQ descriptors substantially increased when combined with contrast information. This work also proposes a novel texture descriptors assembled through the concatenation of the histograms of either LBP or LPQ descriptors and of the local variance estimates. Experimental analysis demonstrated that the proposed descriptors, though more compact, preserved the discrimination capacity of bi-dimensional histograms representing the joint distribution of textural descriptors and contrast information. Finally, the paper compares the discrimination capacity of the LBP- and LPQ-based textural descriptors with that of features derived from the Gray Level Co-occurrence Matrices (GLCM). The related experiments revealed a noteworthy superiority of LBP and LPQ descriptors over the GLCM features in the context of RS image data classification.

This paper is committed to explore object-oriented methods for the classification of Quickbird images, aiming to support future urban population estimates. The study area concerns the southern sector of São José dos Campos city, located in the State of São Paulo, Brazil. By means of a multi-resolution segmentation approach and a six-layer hierarchical classification network, homogeneous residential areas were identified in terms of density of occupation and building standards (single dwelling units or high-rise buildings). The classification network was built upon spectral, geometrical and topological features of the objects in each level of segmentation as well as upon their contextual and semantic interrelationships in-between the hierarchical levels. The final classification of homogeneous residential units was subject to validation, using an object-based Kappa statistics.

In many biometric pattern-recognition problems, the number of training examples per class is limited, and consequently the sample group covariance matrices often used in parametric and nonparametric Bayesian classifiers are poorly estimated or singular. Thus, a considerable amount of effort has been devoted to the design of other covariance estimators, for use in limited-sample and high-dimensional classification problems. In this paper, a new covariance estimate, called the maximum entropy covariance selection (MECS) method, is proposed. It is based on combining covariance matrices under the principle of maximum uncertainty. In order to evaluate the MECS effectiveness in biometric problems, experiments on face, facial expression, and fingerprint classification were carried out and compared with popular covariance estimates, including the regularized discriminant analysis and leave-one-out covariance for the parametric classifier, and the Van Ness and Toeplitz covariance estimates for the nonparametric classifier. The results show that, in image recognition applications whenever the sample group covariance matrices are poorly estimated or ill posed, the MECS method is faster and usually more accurate than the aforementioned approaches in both parametric and nonparametric Bayesian classifiers.

The present paper presents the preliminary results of a research aiming at evaluating the potential of knowledge-based approaches for the interpretation of low-resolution satellite images. This work applies a knowledge-based image interpretation system, so-called GEOAIDA, developed at the University of Hannover, Germany, which hires semantic networks and external operators to model the knowledge basis as well as takes advantage of additional data from geographic information systems, GIS. GEOAIDA is used to perform automatically the post-editing, one of the steps of visual interpretation, aiming at mimicking the reasoning of a trained photo-interpreter when he refines the result of a pixel classification procedure. The results obtained hitherto show that the use of knowledge-based approaches for this purpose is promising and, in the future, can be used to automate the post-editing.