MATLAB Helper ®’s Post

Transform images from spatial to frequency domain in #MATLAB! Learn more in our blog: https://lnkd.in/ddjGbqWZ #ImageProcessing

A new study has been published concerning the determination of aerosol properties using images from an all-sky camera. The presented approach works in all sky conditions, but with focus partially cloudy scenes. A Gaussian Process Regression (GPR) machine learning has been used to associate RGB radiances with aerosol optical depth (AOD) and Angstrom exponent (AE) observed by AERONET. The method has been tested on a 2-year dataset with varying atmospheric conditions. AOD and AE predictions showed excellent agreement with AERONET measurements, particularly when using the quality assurance method.. Sensitivity analysis confirms the stability of our methodology, especially when using quality assurance criteria. This approach could potentially enhance the use of all-sky cameras in studying aerosol-cloud interactions in partially cloudy scenarios. https://lnkd.in/dbRSTzz4

Read how DUG Multi-parameter FWI Imaging breathes new life into vintage data, in this new feature on GEO EXPRO! https://lnkd.in/ge5cn2Jv

Lately I've been thinking about different ways to look at data spatially and through time. Most spatial methods/statistics don't include a time component. Time-series analysis focuses on changes through time. More recently with the advent of Generalized Additive Models (GAMs) and other analyses can look at change through time and space. But is there a way to integrate methods or share definitions/ontologies between traditional spatial statistics and newer time-series analysis? Here is my first try at formulating this question. #GAM #spatialstats #timeseries #trends https://lnkd.in/eDB7FWQQ

Create a Composite Image from different Satellite Images using addBands() Function in Google Earth Engine | Part 4 https://lnkd.in/gqQbvbS4

🔥 #hottopic Multiscale Pixel-Level and Superpixel-Level Method for Hyperspectral Image Classification: Adaptive Attention and Parallel Multi-Hop Graph Convolution by Junru Yin, et al. ➡️ https://brnw.ch/21wLoTt

📣 A one-page overview of LiveSat - the world's only automatic satellite image analysis tool.

Pretrained self-supervised models have greatly advanced many applications in remote sensing. However, satellite images have some unique differences compared to your everyday photograph. The inclusion of multiple spectral bands and images from different time points tell a more complete story than one RGB image alone. Xuyang Li et al. developed Spatial-SpectralMAE (S2MAE) to model these unique aspects with a masked autoencoder. By using both spectral and spatial information, the model learns to reconstruct the input image when 90% of it is masked out. Experiments demonstrated that S2MAE outperforms other alternatives for land cover/use and change detection. Ablation studies further revealed the most critical components of S2MAE: 1) A high masking ratio 2) The use of 3D random masking for spectral data 3) Progressive pretraining on diverse datasets https://lnkd.in/eipqjTyD #RemoteSensing #EarthObservation #MachineLearning #DeepLearning #ComputerVision __________________ Enjoyed this post? Like 👍, comment 💬, or re-post 🔄 to share with others. Click "View my newsletter" under my name ⬆️ to join 1500 readers.

Day 83 of the #365GenAIChallenge! 📅 Today's metric is hot stuff! 🌶️ SPICE (Semantic Propositional Image Caption Evaluation) is a metric used to evaluate the quality of image captions generated by computer vision models. It was developed to overcome some of the limitations of existing metrics like BLEU and METEOR, which primarily focus on surface-level similarity and may not capture the semantic accuracy and relevance of image captions.

Why have object tracking metrics evolved so drastically in the last 20 years? If you are interested in Computer Vision projects, especially in Object Detection and Tracking algorithms check out my overview of typical computer vision projects and the evolution of Detection and Tracking metrics https://lnkd.in/exB4Hdb9