A5002170017- Remote Sensing and LiDAR Applications
- Remote Sensing in Agriculture
- Smart Agriculture and AI
- Species Distribution and Climate Change
- Land Use and Ecosystem Services
- Advanced Neural Network Applications
- Speech and Audio Processing
- Additive Manufacturing and 3D Printing Technologies
- Speech Recognition and Synthesis
- Music and Audio Processing
- Industrial Vision Systems and Defect Detection
- Water Quality Monitoring Technologies
- Material Selection and Properties
- Advanced Image Fusion Techniques
- Identification and Quantification in Food
- Spectroscopy and Chemometric Analyses
- Advanced Chemical Sensor Technologies
- Blind Source Separation Techniques
- Visual Attention and Saliency Detection
- Recycling and utilization of industrial and municipal waste in materials production
- Fire Detection and Safety Systems
- Advanced Image and Video Retrieval Techniques
- Image and Object Detection Techniques
- Retinal Imaging and Analysis
- Wildlife-Road Interactions and Conservation
Universidade Federal de Mato Grosso do Sul
2019-2024
Centro de Tecnologia da Informação Renato Archer
2020
Fraunhofer Portugal Research
2020
Centre for Research and Development in Telecommunications (Brazil)
1991-2002
This study proposes and evaluates five deep fully convolutional networks (FCNs) for the semantic segmentation of a single tree species: SegNet, U-Net, FC-DenseNet, two DeepLabv3+ variants. The performance FCN designs is evaluated experimentally in terms classification accuracy computational load. We also verify benefits connected conditional random fields (CRFs) as post-processing step to improve maps. analysis conducted on set images captured by an RGB camera aboard UAV flying over urban...
Urban forests are an important part of any city, given that they provide several environmental benefits, such as improving urban drainage, climate regulation, public health, biodiversity, and others. However, tree detection in cities is challenging, the irregular shape, size, occlusion, complexity areas. With advance technologies, deep learning segmentation mapping methods can map accurately. We applied a region-based CNN object instance algorithm for semantic canopies environments based on...
This research arose from the need to aggregate computer vision technology and machine learning in sheep weight control facilitate weighing process of animals farms. The experiment was conducted collect images their weights, later, annotations were made, generating a mask image dataset. We selected attribute extraction algorithms that extracted shape, size, angles with k-curvature. With these data, we used stratified five-fold cross-validation. Also, eight techniques aimed at regression,...
Recently, methods based on Convolutional Neural Networks (CNN) achieved impressive success in semantic segmentation tasks. However, challenges such as the class imbalance and uncertainty pixel-labeling process are not completely addressed. As such, we present a new approach that calculates weight for each pixel considering its during labeling process. The pixel-wise weights used training to increase or decrease importance of pixels. Experimental results show proposed leads significant...
Pantanal is the largest continuous wetland in world, but its biodiversity currently endangered by catastrophic wildfires that occurred last three years. The information available for area only refers to location and extent of burned areas based on medium low-spatial resolution imagery, ranging from 30 m up 1 km. However, improve measurements assist environmental actions, robust methods are required provide a detailed mapping higher-spatial scale areas, such as PlanetScope imagery with 3–5...
Tree species mapping is an important type of information demanded in different study fields. However, this task can be expensive and time-consuming, making it difficult to monitor extensive areas. Hence, automatic methods are required optimize tree mapping. Here, we propose a deep learning-based mobile application tool for classification high-spatial-resolution RGB images. Several learning architectures were evaluated, including networks traditional models. A total 2,349 images used, which...
This paper presents a semi-supervised learning method based on superpixels and convolutional neural networks (CNNs) to assign improve the identification of weeds in soybean crops. Despite its promising results, CNNs require massive amounts labeled training data learn, so we intend manual labeling phase with an automated pseudo-labeling process. We propose that uses additional mini-batch processing fine-tune pseudo labels images previously annotated SLIC segmentation during algorithm phase....
Abstract Recently, methods based on Convolutional Neural Networks (CNN) achieved impressive success in semantic segmentation tasks. However, challenges such as the class imbalance and uncertainty pixel-labeling process are not completely addressed. As such, we present a new approach that calculates weight for each pixel considering its during labeling process. The pixel-wise weights used training to increase or decrease importance of pixels. Experimental results using different datasets like...
Abstract The use of uncrewed aerial vehicle to map the environment increased significantly in last decade enabling a finer assessment land cover. However, creating accurate maps is still complex and costly task. Deep learning (DL) new generation artificial neural network research that, combined with remote sensing techniques, allows refined understanding our can help solve challenging cover mapping issues. This focuses on vegetation segmentation kettle holes. Kettle holes are small,...
Unmanned aerial vehicles (UAVs) are platforms suitable for obtaining information utilizing sensors in a great variety of areas and enviroments, thus this context, paper objective was to identify trees images collected using UAV high-resolution imagery, with the digital approach superpixel segmentation convolutional neural networks. A forest environment analyzed form an orthomosaic that produced 423 images. Superpixels were generated Simple Linear Iterative Clustering (SLIC) method,...
Knowing the spatial distribution of endangered tree species in a forest ecosystem or remnants is valuable information to support environmental conservation practices. The use Unmanned Aerial Vehicles (UAVs) offers suitable alternative for this task, providing very high-resolution images at low costs. In parallel, recent advances computer vision field have led development effective deep learning techniques end-to-end semantic image segmentation. scenario, DeepLabv3+ well established as...
Abstract. Knowing the spatial distribution of endangered tree species in a forest ecosystem or remnants is valuable information to support environmental conservation practices. The use Unmanned Aerial Vehicles (UAVs) offers suitable alternative for this task, providing very high-resolution images at low costs. In parallel, recent advances computer vision field have led development effective deep learning techniques end-to-end semantic image segmentation. scenario, DeepLabv3+ well established...
This paper presents a semi-supervised learning method based on superpixels and convolutional neural networks (CNNs) to assign improve the identification of weeds in soybean crops. Despite its promising results, CNNs require massive amounts labeled training data learn, so we intend manual labeling phase with an automated pseudo process. We propose that uses just additional mini-batch processing fine-tune labels images previously annotated SLIC segmentation during algorithm phase. research...
Object detection and image segmentation are essential for environmental monitoring. This task can be performed automated using machines with the processing capabilities of convolutional neural networks (CNN), achieving outstanding performance thanks to current computing capacity data available. Still, there advances perceive questions on how get optimal improve results. With this aim, we assessed a state-of-the-art CNN, Dynamic Dilated Convolution Neural Network (DDCN), segment trees inside...
Crop segmentation, the process of identifying and delineating agricultural fields or specific crops within an image, plays a crucial role in precision agriculture, enabling farmers public managers to make informed decisions regarding crop health, yield estimation, resource allocation Midwest Brazil. The (corn) this region are being damaged by wild pigs other diseases. For quantification corn fields, paper applies novel computer-vision techniques new dataset imagery composed 1416 256 × images...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Results of simulations are presented for LPC (linear predictive coding) vocoders using both scalar and vector quantization as well interpolation the quantized parameters. A conventional LPC-8 vocoder was initially carefully designed to be used a reference system. The use pitch detector based on skeleton speech signal well-designed table log area ratio coefficients provided high level quality intelligibility at 2200 b/s. By employing codebooks large database, similar performance achieved bit...
This paper presents a comparison among different parametric representations of speech signals employed in recognition. In order to make this comparison, isolated word recognizers were built using hidden Markov models (HMM) and multilayer perceptrons (MLP). All the implemented speaker-independent same database was used for their evaluation. The consisted 50-word vocabulary spoken Brazilian Portuguese.
Urban forests are crucial for the population well-being and improvement of quality life. For example, they contribute to rain damping local climate. Therefore a correct accurate mapping this resource is fundamental its management. We investigated method that combines machine learning SLIC superpixel techniques using different Superpixels (k) number map trees in metropolitan region municipality Campo Grande-MS, Brazil with aerial orthoimages GSD (Ground Sample Distance) 10 cm. The combination...