Search

Predicting wildfire spread behavior is an extremely important task for many countries. On a small scale, it is possible to ensure constant monitoring of the natural landscape through ground means. However, on the scale of large countries, this becomes practically impossible due to remote and vast forest territories. The most promising source of data in this case that can provide global monitoring is remote sensing data. Currently, the main challenge is the development of an effective pipeline that combines geospatial data collection and the application of advanced machine learning algorithms. Most approaches focus on short-term fire spreading prediction and utilize data from unmanned aerial vehicles (UAVs) for this purpose. In this study, we address the challenge of predicting fire spread on a large scale and consider a forecasting horizon ranging from 1 to 5 days. We train a neural network model based on the MA-Net architecture to predict wildfire spread based on environmental and climate data, taking into account spatial distribution features. Estimating the importance of features is another critical issue in fire behavior prediction, so we analyze their contribution to the model’s results. According to the experimental results, the most significant features are wind direction and land cover parameters. The F1-score for the predicted burned area varies from 0.64 to 0.68 depending on the day of prediction (from 1 to 5 days). The study was conducted in northern Russian regions and shows promise for further transfer and adaptation to other regions. This geospatial data-based artificial intelligence (AI) approach can be beneficial for supporting emergency systems and facilitating rapid decision-making. [ABSTRACT FROM AUTHOR]

Floods are natural events that can have a significant impacts on the economy and society of affected regions. To mitigate their effects, it is crucial to conduct a rapid and accurate assessment of the damage and take measures to restore critical infrastructure as quickly as possible. Remote sensing monitoring using artificial intelligence is a promising tool for estimating the extent of flooded areas. However, monitoring flood events still presents some challenges due to varying weather conditions and cloud cover that can limit the use of visible satellite data. Additionally, satellite observations may not always correspond to the flood peak, and it is essential to estimate both the extent and volume of the flood. To address these challenges, we propose a methodology that combines multispectral and radar data and utilizes a deep neural network pipeline to analyze the available remote sensing observations for different dates. This approach allows us to estimate the depth of the flood and calculate its volume. Our study uses Sentinel-1, Sentinel-2 data, and Digital Elevation Model (DEM) measurements to provide accurate and reliable flood monitoring results. To validate the developed approach, we consider a flood event occurred in 2021 in Ushmun. As a result, we succeeded to evaluate the volume of that flood event at 0.0087 km3. Overall, our proposed methodology offers a simple yet effective approach to monitoring flood events using satellite data and deep neural networks. It has the potential to improve the accuracy and speed of flood damage assessments, which can aid in the timely response and recovery efforts in affected regions. [ABSTRACT FROM AUTHOR]

Leaf area and biomass are important morphological parameters for in situ plant monitoring since a leaf is vital for perceiving and capturing the environmental light as well as represents the overall plant development. The traditional approach for leaf area and biomass measurements is destructive requiring manual labor and may cause damages for the plants. In this work, we report on the AI-based approach for assessing and predicting the leaf area and plant biomass. The proposed approach is able to estimate and predict the overall plants biomass at the early stage of growth in a non-destructive way. For this reason we equip an industrial greenhouse for cucumbers growing with the commercial off-the-shelf environmental sensors and video cameras. The data from sensors are used to monitor the environmental conditions in the greenhouse while the top-down images are used for training Fully Convolutional Neural Networks (FCNN). The FCNN performs the segmentation task for leaf area calculation resulting in 82% accuracy. Application of trained FCNNs to the sequences of camera images allowed the reconstruction of per-plant leaf area and their growth-dynamics. Then we established the dependency between the average leaf area and biomass using the direct measurements of the biomass. This in turn allowed for reconstruction and prediction of the dynamics of biomass growth in the greenhouse using the image data with 10% average relative error for the 12 days prediction horizon. The actual deployment showed the high potential of the proposed data-driven approaches for plant growth dynamics assessment and prediction. Moreover, it closes the gap towards constructing fully closed autonomous greenhouses for harvests and plants biological safety. [ABSTRACT FROM AUTHOR]

Food quality control is an important task in the agricultural domain at the postharvest stage for avoiding food losses. The latest achievements in image processing with deep learning (DL) and computer vision (CV) approaches provide a number of effective tools based on the image colorization and image-to-image translation for plant quality control at the postharvest stage. In this article, we propose the approach based on Generative Adversarial Network (GAN) and Convolutional Neural Network (CNN) techniques to use synthesized and segmented VNIR imaging data for early postharvest decay and fungal zone predictions as well as the quality assessment of stored apples. The Pix2PixHD model achieved higher results in terms of VNIR images translation from RGB (SSIM = 0.972). Mask R-CNN model was selected as a CNN technique for VNIR images segmentation and achieved 58.861 for postharvest decay zones, 40.968 for fungal zones and 94.800 for both the decayed and fungal zones detection and prediction in stored apples in terms of F1-score metric. In order to verify the effectiveness of this approach, a unique paired dataset containing 1305 RGB and VNIR images of apples of four varieties was obtained. It is further utilized for a GAN model selection. Additionally, we acquired 1029 VNIR images of apples for training and testing a CNN model. We conducted validation on an embedded system equipped with a graphical processing unit. Using Pix2PixHD, 100 VNIR images from RGB images were generated at a rate of 17 frames per second (FPS). Subsequently, these images were segmented using Mask R-CNN at a rate of 0.42 FPS. The achieved results are promising for enhancing the food study and control during the postharvest stage. [ABSTRACT FROM AUTHOR]

Currently, we can solve a wide range of tasks using computer vision algorithms, which reduce manual labor and enable rapid analysis of the environment. The remote sensing domain provides vast amounts of satellite data, but it also poses challenges associated with processing this data. Baseline solutions with intermediate results are available for various tasks, such as forest species classification, infrastructure recognition, and emergency situation analysis using satellite data. Despite these advances, two major issues with high-performing artificial intelligence algorithms remain in the current decade. The first issue relates to the availability of data. To train a robust algorithm, a reasonable amount of well-annotated training data is required. The second issue is the availability of satellite data, which is another concern. Even though there are a number of data providers, high-resolution and up-to-date imagery is extremely expensive. This paper aims to address these challenges by proposing an effective pipeline for building segmentation that utilizes freely available Sentinel-2 data with 10 m spatial resolution. The approach we use combines a super-resolution (SR) component with a semantic segmentation component. As a result, we simultaneously consider and analyze SR and building segmentation tasks to improve the quality of the infrastructure analysis through medium-resolution satellite data. Additionally, we collected and made available a unique dataset for the Russian Federation covering area of 1091.2 square kilometers. The dataset provides Sentinel-2 imagery adjusted to the spatial resolution of 2.5 m and is accompanied by semantic segmentation masks. The building footprints were created using OpenStreetMap data that was manually checked and verified. Several experiments were conducted for the SR task, using advanced image SR methods such as the diffusion-based SR3 model, RCAN, SRGAN, and MCGR. The MCGR network produced the best result, with a PSNR of 27.54 and SSIM of 0.79. The obtained SR images were then used to tackle the building segmentation task with different neural network models, including DeepLabV3 with different encoders, SWIN, and Twins transformers. The SWIN transformer achieved the best results, with an F1-score of 79.60. [ABSTRACT FROM AUTHOR]

Large datasets catalyze the rapid expansion of deep learning and computer vision. At the same time, in many domains, there is a lack of training data, which may become an obstacle for the practical application of deep computer vision models. To overcome this problem, it is popular to apply image augmentation. When a dataset contains instance segmentation masks, it is possible to apply instance-level augmentation. It operates by cutting an instance from the original image and pasting to new backgrounds. This article challenges a dataset with the same objects present in various domains. We introduce the Context Substitution for Image Semantics Augmentation framework (CISA), which is focused on choosing good background images. We compare several ways to find backgrounds that match the context of the test set, including Contrastive Language–Image Pre-Training (CLIP) image retrieval and diffusion image generation. We prove that our augmentation method is effective for classification, segmentation, and object detection with different dataset complexity and different model types. The average percentage increase in accuracy across all the tasks on a fruits and vegetables recognition dataset is 4.95 % . Moreover, we show that the Fréchet Inception Distance (FID) metrics has a strong correlation with model accuracy, and it can help to choose better backgrounds without model training. The average negative correlation between model accuracy and the FID between the augmented and test datasets is 0.55 in our experiments. [ABSTRACT FROM AUTHOR]

Estimation of terrestrial carbon balance is one of the key tasks in the understanding and prognosis of climate change impacts and the development of tools and policies according to carbon mitigation and adaptation strategies. Forest ecosystems are one of the major pools of carbon stocks affected by controversial processes influencing carbon stability. Therefore, monitoring forest ecosystems is a key to proper inventory management of resources and planning their sustainable use. In this survey, we discuss which computer vision techniques are applicable to the most important aspects of forest management actions, considering the wide availability of remote sensing (RS) data of different resolutions based both on satellite and unmanned aerial vehicle (UAV) observations. Our analysis applies to the most occurring tasks such as estimation of forest areas, tree species classification, and estimation of forest resources. Through the survey, we also provide a necessary technical background with a description of suitable data sources, algorithms' descriptions, and corresponding metrics for their evaluation. The implementation of the provided techniques into routine workflows is a significant step toward the development of systems of continuous actualization of forest data, including real-time monitoring. It is crucial for diverse purposes on both local and global scales. Among the most important are the implementation of improved forest management strategies and actions, carbon offset projects, and enhancement of the prediction accuracy of system changes under different land-use and climate scenarios. [ABSTRACT FROM AUTHOR]

Remote sensing tasks play a very important role in the domain of sensing and measuring, and can be very specific. Advances in computer vision techniques allow for the extraction of various information from remote sensing satellite imagery. This information is crucial in making quantitative and qualitative assessments for monitoring of forest clearing in protected areas for power lines, as well as for environmental analysis, in particular for making assessments of carbon footprint, which is a highly relevant task. Solving these problems requires precise segmentation of the forest mask. Although forest mask extraction from satellite data has been considered previously, no open-access applications are able to provide the high-detailed forest mask. Detailed forest masks are usually obtained using unmanned aerial vehicles (UAV) that set particular limitations such as cost and inapplicability for vast territories. In this study, we propose a novel neural network-based approach for high-detailed forest mask creation. We implement an object-based augmentation technique for a minimum amount of labeled high-detailed data. Using this augmented data we fine-tune the models, which are trained on a large forest dataset with less precise labeled masks. The provided algorithm is tested for multiple territories in Russia. The F 1 - s c o r e , for small details (such as individual trees) was improved to 0.929 compared to the baseline score of 0.856 . The developed model is available in an SAAS platform. The developed model allows a detailed and precise forest mask to be easily created, which then be used for solving various applied problems. [ABSTRACT FROM AUTHOR]

In this article, we share our experience in the scope of controlled-environment agriculture automation in the Antarctic station greenhouse facility called EDEN ISS. For remote plant monitoring, control, and maintenance, we solve the problem of plant classification. Due to the inherent communication limitations between Antarctica and Europe, we first propose the image compression mechanism for the data collection. We show that we can compress the images, on average, 7.2 times for efficient transmission over the weak channel. Moreover, we prove that decompressed images can be further used for computer vision applications. Upon decompressing images, we apply machine learning for the classification task. We achieve 92.6% accuracy on an 18-classes unbalanced dataset. The proposed approach is promising for a number of agriculture related applications, including the plant classification, identification of plant diseases, and deviation of plant phenology. [ABSTRACT FROM AUTHOR]

The Hogweed of Sosnowskyi (lat. Heracleum sosnówskyi) is poisonous for humans, dangerous for farming crops, and local ecosystems. This plant is fast-growing and has already spread all over Eurasia: from Germany to the Siberian part of Russia, and its distribution expands year-by-year. In-situ detection of this harmful plant is a tremendous challenge for many countries. Meanwhile, there are no automatic systems for detection and localization of hogweed. In this article, we report on an approach for fast and accurate detection of hogweed. The approach includes the Unmanned Aerial Vehicle (UAV) with an embedded system on board running various Fully Convolutional Neural Networks (FCNN). We propose the optimal architecture of FCNN for the embedded system relying on the trade-off between the detection quality and frame rate. We propose a model that achieves ROC AUC 0.96 in the hogweed segmentation task, which can process 4K frames at 0.46 FPS on NVIDIA Jetson Nano. The developed system can recognize the hogweed on the scale of individual plants and leaves. This system opens up a wide vista for obtaining comprehensive and relevant data about the spreading of harmful plants allowing for the elimination of their expansion. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) has smoothly penetrated in a number of monitoring and control applications including agriculture. However, research efforts toward low-power sensing devices with fully functional AI on board are still fragmented. In this article, we present an embedded system enriched with the AI, ensuring the continuous analysis and in situ prediction of the growth dynamics of plant leaves. The embedded solution is grounded on a low-power embedded sensing system with a graphics processing unit (GPU) and is able to run the neural network-based AI on board. We use a recurrent neural network (RNN) called the long short-term memory network (LSTM) as a core of AI in our system. The proposed approach guarantees the system autonomous operation for 180 days using a standard Li-ion battery. We rely on the state-of-the-art mobile graphical chips for “smart” analysis and control of autonomous devices. This pilot study opens up wide vista for a variety of intelligent monitoring applications, especially in the agriculture domain. In addition, we share with the research community the Tomato Growth data set. [ABSTRACT FROM AUTHOR]

Artificial Intelligence (AI) has been recently applied to a number of sensing scenarios for realizing the prediction, control and/or recognition tasks. However, its integration to embedded systems is still limited. We propose a low-power sensing system with the AI on board with a special focus on the application in agriculture. For this reason we designed a Convolutional Neural Network (CNN) which achieves 83% of average Intersection over Union (IoU) score on the test dataset and 97% of seeds recognition accuracy on the validation dataset. The proposed solution is able to perform the seeds recognition, and germination detection through the images processing. For training the CNN we collect a dataset of images of seed germination process at different stages. The entire system is assessed in an industrial facility. The experimental results demonstrate that the proposed system opens up wide vista for smart applications in the context of Internet of Things requiring the intelligent and autonomous operation from ‘things’. [ABSTRACT FROM AUTHOR]

The near-infrared (NIR) spectral range (from 780 to 2500 nm) of the multispectral remote sensing imagery provides vital information for landcover classification, especially concerning vegetation assessment. Despite the usefulness of NIR, it does not always accomplish common RGB. Modern achievements in image processing via deep neural networks make it possible to generate artificial spectral information, for example, to solve the image colorization problem. In this research, we aim to investigate whether this approach can produce not only visually similar images but also an artificial spectral band that can improve the performance of computer vision algorithms for solving remote sensing tasks. We study the use of a generative adversarial network (GAN) approach in the task of the NIR band generation using only RGB channels of high-resolution satellite imagery. We evaluate the impact of a generated channel on the model performance to solve the forest segmentation task. Our results show an increase in model accuracy when using generated NIR compared to the baseline model, which uses only RGB ( 0.947 and 0.914 F1-scores, respectively). The presented study shows the advantages of generating the extra band such as the opportunity to reduce the required amount of labeled data. [ABSTRACT FROM AUTHOR]

Usage of multispectral satellite imaging data opens vast possibilities for monitoring and quantitatively assessing properties or objects of interest on a global scale. Machine learning and computer vision (CV) approaches show themselves as promising tools for automatizing satellite image analysis. However, there are limitations in using CV for satellite data. Mainly, the crucial one is the amount of data available for model training. This paper presents a novel image augmentation approach called MixChannel that helps to address this limitation and improve the accuracy of solving segmentation and classification tasks with multispectral satellite images. The core idea is to utilize the fact that there is usually more than one image for each location in remote sensing tasks, and this extra data can be mixed to achieve the more robust performance of the trained models. The proposed approach substitutes some channels of the original training image with channels from other images of the exact location to mix auxiliary data. This augmentation technique preserves the spatial features of the original image and adds natural color variability with some probability. We also show an efficient algorithm to tune channel substitution probabilities. We report that the MixChannel image augmentation method provides a noticeable increase in performance of all the considered models in the studied forest types classification problem. [ABSTRACT FROM AUTHOR]

Currently many countries are struggling to rationalize water quality monitoring stations which is caused by economic demand. Though this process is essential indeed, the exact elements of the system to be optimized without a subsequent quality and accuracy loss still remain obscure. Therefore, accurate historical data on groundwater pollution is required to detect and monitor considerable environmental impacts. To collect such data appropriate sampling and assessment methodologies with an optimum spatial distribution augmented should be exploited. Thus, the configuration of water monitoring sampling points and the number of the points required are now considered as a fundamental optimization challenge. The paper offers and tests metaheuristic approaches for optimization of monitoring procedure and multi-factors assessment of water quality in "New Moscow" area. It is shown that the considered algorithms allow us to reduce the size of the training sample set, so that the number of points for monitoring water quality in the area can be halved. Moreover, reducing the dataset size improved the quality of prediction by 20%. The obtained results convincingly demonstrate that the proposed algorithms dramatically decrease the total cost of analysis without dampening the quality of monitoring and could be recommended for optimization purposes. [ABSTRACT FROM AUTHOR]

Sustainable management of the environment is based on the preservation of natural resources, first of all, freshwater—both surface and groundwater—from exhaustion and contamination. Thus, development of adequate monitoring solutions, including fast and adaptive modelling approaches, are of high importance. Recent progress in machine learning techniques provide an opportunity to improve the prediction accuracy of the spatial distribution of properties of natural objects and to automate all stages of this process to exclude uncertainties caused by handcrafting. We propose a technique to construct the weighted Water Quality Index (WQI) and the spatial prediction map of the WQI in tested area. In particular, WQI is calculated using dimensionality reduction technique (Principal Component Analysis), and spatial map of WQI is constructed using Gaussian Process Regression with automatic kernel structure selection using Bayesian Information Criterion (BIC). We validate our approach on a new dataset for groundwater quality in the New Moscow region, where groundwater is mostly used for drinking purposes. According to estimated WQI values, groundwater quality across the study region is relatively high, with few points, less than 0.5% of all observations, severely contaminated. Estimated WQIs then were used to construct spatial distribution models, GPR-BIC approach was compared with ordinary Kriging (OK), Universal Kriging (UK) with exponential, Gaussian, polynomial and periodic kernels. Quality of models was assessed using cross-validation scheme, according to which BIC-GPR approach showed better performance on average with 15% higher R 2 score comparing to other Kriging models. We show that the proposed geospatial interpolation is a potentially powerful and adaptable tool for predicting the spatial distribution of properties of natural resources. [ABSTRACT FROM AUTHOR]

Environment pollutants, especially those from total petroleum hydrocarbons (TPH), have a highly complex chemical, biological and physical impact on soils. Here we study this influence via modelling the TPH acute phytotoxicity effects on eleven samples of soils from Sakhalin island in greenhouse conditions. The soils were contaminated with crude oil in different doses ranging from the 3.0–100.0 g kg−1. Measuring the Hordeum vulgare root elongation, the crucial ecotoxicity parameter, we have estimated. We have also investigated the contrast effect in different soils. To predict TPH phytotoxicity different machine learning models were used, namely artificial neural network (ANN) and support vector machine (SVM). The models under discussion were proved to be valid using the mean absolute error method (MAE), the root mean square error method (RMSE), and the coefficient of determination (R2). We have shown that ANN and SVR can successfully predict barley response based on soil chemical properties (pH, LOI, N, P, K, clay, TPH). The best achieved accuracy was as following: MAE – 8.44, RMSE –11.05, and R2 –0.80. Image 1 • We studied TPH phytotoxicity in eleven soils from Sakhalin island. • Barley root bioassay response to TPH is highly complex and non-linear. • ANN and SVR are able to predict phytotoxicity in the TPH polluted soils. [ABSTRACT FROM AUTHOR]