A Framework for Evaluating Land Use and Land Cover Classification Using Convolutional Neural Networks (original) (raw)
Related papers
Land use Land Cover Classifıcation using Deep Learning Classifiers for Remote Sensed Images
International Conference on IoT based Control Networks and Intelligent Systems (ICICNIS 2020), 2020
The image classification accuracy is enhanced by applying Deep Learn ing Models (DLM) which has a robus t learning ability by incorporating both feature extract ion and classification p rocedure into single image classificat ion test. Here a deep learn in g-based classification technique is applied to High Spatial Resolution Remote Sensing Images (HSRRSI) to ext ract mult i-layer features. The two networks i.e., residual network and inception network are co mb ined into one new model to obtain higher accuracy then said individual residual network and inception network. The new model designed was extensively weighed on data's from Remote Sensing Image Classificat ion Benchmark (RSI-CB). The dataset obtained from RSI-CB is split into 70:30 rat io for training and testing respectively. The performances of proposed approach are then assessed by kappa coefficient (K) and accuracy (A).
Deep learning approach for land use images classification
E3S Web of Conferences
CNN (convolutional neural networks) are a category of neural networks that are majorly used for image classification and recognition. This Deep Learning (DL) technique is used to solve complex problems, particularly for environmental protection, its approaches have affected several domains without exception, geospatial world is one vised domain. In this paper we aim to classify aerial images of Tangier region, city located in north of Morocco, by using pixel based image classification with convolutional Neural Networks. Flickr API is used to get our test images dataset. These images are used as input to a pretrained network Resnet18, a small convolution neural network architecture, which is able to recognize 21 land use classes of images. Our methodology is based on the following steps, first we set up the data, and then we re-train the cited Deep Learning model (Transfer Learning) and perform a quick and visual verification, by generating a labeled map from the geotagged images, la...
International Journal of Advanced Computer Science and Applications, 2022
Deep Learning algorithms have become more popular in computer vision, especially in the image classification field. This last has many applications such as moving object detection, cancer detection, and the classification of satellite images, also called images of land use-land cover (LULC), which are the scope of this paper. It represents the most commonly used method for decision making in the sustainable management of natural resources at various geographical levels. However, methods of satellite images analysis are expensive in the computational time and did not show good performance. Therefore, this paper, on the one hand, proposes a new CNN architecture called Modified MobileNet V1 (MMN) based on the fusion of MobileNet V1 and ResNet50. On the other hand, it presents a comparative study of the proposed model and the most used models based on transfer learning, i.e. MobileNet V1, VGG16, DenseNet201, and ResNet50. The experiments were conducted on the dataset Eurosat, and they show that ResNet50 results emulate the other models.
Radioelectronic and Computer Systems, 2023
Multispectral images acquired by satellites have been used in many fields such as agriculture, urban change detection, finding fire-hazardous forest areas, and real-time surface monitoring. The central issue in remote sensing analysis is land use and land cover classification. Land use and land cover classification (LULC) is the process of classification into meaningful classes based on the spectral characteristics of remote sensing data. Land use and land cover classification is a challenging task due to the complex nature of the Earth's surface. The accuracy of solving the issue using deep learning approaches depends on the quality of the remote sensing data, the choice of the classification algorithm. The ability to obtain high-resolution multispectral images periodically could dramatically improve remote sensing solutions. In this study, we propose a solution for the land cover and land classification problem of high-resolution remote sensing data by applying deep learning methods using EuroPlanet geo-referenced high-quality images with four bands and pixel resolution of 204x204 per image, and acquired by Planet platform in 2020-2022 years. The dataset consists of 25911 images with spatial resolution up to 3.125 meters per pixel and 10 different classes. In the past decade, artificial neural networks have shown great performance in solving complex image classification tasks. For the dataset evaluation, we have taken advantage of state-of-art pretrained convolutional neural network models Res-Net50v2, EfficientNetV2, Xception, VGG-16, and DenseNet201 with fine tuning. It has been established that DenseNet201 pretrained neural network outperformed other models. The accuracy of the test data was 92.01 % and the F1 metric was 91.63 %. In addition, bands evaluation for the dataset was carried out. Overall classification accuracy of 93.83 % and F1 score of 93.56 % were achieved by DenseNet201 model. The results could be used for area verification, real-time monitoring, and surface change detection. Nowadays, this is very helpful for Ukrainian territory because of the Russian invasion and the country's recovery in the future.
Land Use Classification using Convolutional Neural Networks Applied to Ground-Level images
ACM SIGSPATIAL, 2015
Land use mapping is a fundamental yet challenging task in geographic science. In contrast to land cover mapping, it is generally not possible using overhead imagery. The recent, explosive growth of online geo-referenced photo collections suggests an alternate approach to geographic knowledge discovery. In this work, we present a general framework that uses ground-level images from Flickr for land use mapping.
Satellite-Net: Automatic Extraction of Land Cover Indicators from Satellite Imagery by Deep Learning
ArXiv, 2019
In this paper we address the challenge of land cover classification for satellite images via Deep Learning (DL). Land Cover aims to detect the physical characteristics of the territory and estimate the percentage of land occupied by a certain category of entities: vegetation, residential buildings, industrial areas, forest areas, rivers, lakes, etc. DL is a new paradigm for Big Data analytics and in particular for Computer Vision. The application of DL in images classification for land cover purposes has a great potential owing to the high degree of automation and computing performance. In particular, the invention of Convolution Neural Networks (CNNs) was a fundament for the advancements in this field. In [1], the Satellite Task Team of the UN Global Working Group describes the results achieved so far with respect to the use of earth observation for Official Statistics. However, in that study, CNNs have not yet been explored for automatic classification of imagery. This work invest...
A hybrid deep convolutional neural network for accurate land cover classification
International Journal of Applied Earth Observation and Geoinformation, 2021
Land cover classification provides updated information regarding the Earth's resources, which is vital for agricultural investigation, urban management, and disaster monitoring. Current advances in sensor technology on satellite and aerial remote sensing (RS) devices have improved the spatial-spectral, radiometric, and temporal resolutions of images over time. These improvements offer invaluable chances of understanding land cover information. However, land cover classification from RS images is an intricate task because of the high intra-class disparities, low inter-class similarities, and image variation types. We propose a cascaded residual dilated network (CRD-Net) for land cover classification using very high spatial resolution (VHSR) images to address these challenges. The proposed hybrid network follows the encoder-decoder concept with a spatial attention block to guide the network on learnable discriminate features coupled with an intermediary loss to enhance the training process. Moreover, a cascaded residual dilated module increases the network's receptive field to enrich multi-contextual features further, thus boosting the resultant feature descriptor. Extensive experimental results demonstrate that the proposed CRD-Net outperformed state-of-the-art methods, achieving an overall accuracy (OA) of 90.73% and 90.51% on the ISPRS Potsdam land cover dataset and ISPRS Vaihingen dataset, respectively.
Springer in series Lecture Notes in Artificial Intelligence LNCS/LNAI, 2019
Regional land use planning and monitoring remain an issue in many developing countries. Efficient solution for both tasks depended on remote sensing technology to capture and analyze remotely sensed data of the region of interest. Although a plethora of methods for land cover classification have been reported, the problem remained a challenging task in computer vision field. The advent of deep learning method in the past decade has been very instrumental to develop a robust method for land cover classification using satellite imagery as input. The objective of this paper was to present empiric results on using CNN as a land cover classifier model using Sentinel-2 spatial satellite imagery. Prior to model training, the input image representation was extracted using eCognition to produce texture, brightness, shape, and vegetation index. Land cover labeling followed the Land Cover Class in Medium Resolution Optical Imagery Interpretation document provided by Indonesian National Standardization Agency. The training of CNN model achieved 0.98 mean training accuracy and 0.98 mean testing accuracy. As comparison, the same data and same feature were trained with another model: Gradient Boosting Model (GBM). The results revealed that the training accuracy and testing accuracy with GBMs were 0.98 and 0.95 respectively. CNN model showed small improvement of the accuracy to classify land cover with the image feature (NDVI, Brightness, GLCM homogeneity and Rectangular fit).
Remote Sensing, 2020
Land cover information plays an important role in mapping ecological and environmental changes in Earth’s diverse landscapes for ecosystem monitoring. Remote sensing data have been widely used for the study of land cover, enabling efficient mapping of changes of the Earth surface from Space. Although the availability of high-resolution remote sensing imagery increases significantly every year, traditional land cover analysis approaches based on pixel and object levels are not optimal. Recent advancement in deep learning has achieved remarkable success on image recognition field and has shown potential in high spatial resolution remote sensing applications, including classification and object detection. In this paper, a comprehensive review on land cover classification and object detection approaches using high resolution imagery is provided. Through two case studies, we demonstrated the applications of the state-of-the-art deep learning models to high spatial resolution remote sensi...
Research Square (Research Square), 2023
Land use and land cover (LULC) analysis is highly signi cant for various environmental and social applications. As remote sensing (RS) data becomes more accessible, LULC benchmark datasets have emerged as powerful tools for complex image classi cation tasks. These datasets are used to test stateof-the-art arti cial intelligence models, particularly convolutional neural networks (CNNs), which have demonstrated remarkable effectiveness in such tasks. Nonetheless, there are existing limitations, one of which is the scarcity of benchmark datasets from diverse settings, including those speci cally pertaining to the Indian scenario. This study addresses these challenges by generating medium-sized benchmark LULC datasets from two Indian states and evaluating state-of-the-art CNN models alongside traditional ML models. The evaluation focuses on achieving high accuracy in LULC classi cation, speci cally on the generated patches of LULC classes. The dataset comprises 4000 labelled images derived from Sentinel-2 satellite imagery, encompassing three visible spectral bands and four distinct LULC classes. Through quantitative experimental comparison, the study demonstrates that ML models outperform CNN models, exhibiting superior performance across various LULC classes with unique characteristics. Notably, using a traditional ML model, the proposed novel dataset achieves an impressive overall classi cation accuracy of 96.57%. This study contributes by introducing a standardized benchmark dataset and highlighting the comparative performance of deep CNNs and traditional ML models in the eld of LULC classi cation.