A single image deraining method is proposed. A wavelet transforming step processes an initial rain image to generate an i-th stage low-frequency rain image and a plurality of i-th stage high-frequency rain images. An image deraining step inputs the i-th stage low-frequency rain image to a low-frequency deraining model to output an i-th stage low-frequency derain image. A first inverse wavelet transforming step recombines the n-th stage low-frequency derain image with the n-th stage high-frequency derain images to form an n-th stage derain image. A weighted blending step blends a (n−1)-th stage low-frequency derain image with the n-th stage derain image to generate a (n−1)-th stage blended derain image. A second inverse wavelet transforming step recombines the (n−1)-th stage high-frequency derain images with the (n−1)-th stage blended derain image to form a (n−1)-th stage derain image, and sets n to n−1 and repeats the last two steps.

A system and method for assessing categorized pixels of satellite images associated with agriculture land parcel using an artificial intelligence (AI) model are provided. The method includes, (i) obtaining satellite images associated with agriculture land parcel; (ii) pre-processing the satellite images to generate pre-processed satellite images, (iii) training the AI model by categorizing historical plurality of pixels from historical plurality of satellite images based on historical satellite data and correlating historical scores to historical categorized pixels to obtain trained AI model, (iv) classifying pixels of pre-processed satellite images into crop area-pixels and non-crop area pixels by determining a profile of time series data that corresponds to at least one of normalized difference vegetation index, normalized difference water index, land surface temperature, modified normalized difference water index, or land surface water index, (v) determining, using trained AI model, categorized pixels based on classification, (vi) assessing categorized pixels with a score.

A system for detecting clouds and cloud shadows is described. In one approach, clouds and cloud shadows within a remote sensing image are detected through a three step process. In the first stage a high-precision low-recall classifier is used to identify cloud seed pixels within the image. In the second stage, a low-precision high-recall classifier is used to identify potential cloud pixels within the image. Additionally, in the second stage, the cloud seed pixels are grown into the potential cloud pixels to identify clusters of pixels which have a high likelihood of representing clouds. In the third stage, a geometric technique is used to determine pixels which likely represent shadows cast by the clouds identified in the second stage. The clouds identified in the second stage and the shadows identified in the third stage are then exported as a cloud mask and shadow mask of the remote sensing image.

A method for characterizing the atmospheric turbulence, includes acquiring images of a celestial object by means of a camera coupled to a small telescope; analyzing the acquired images to determine angle of arrival fluctuations of wavefronts from positions of spots formed by the celestial object in the acquired images; determining variances of the angle of arrival fluctuations; and estimating the Fried parameter from the variances of the angle of arrival fluctuations, by setting an outer scale parameter of the atmospheric turbulence to a fixed median value.

A method includes identifying a plurality of images corresponding to sky conditions and isolating cloud pixels from sky pixels in each of the plurality of images. Responsive to determining percentage of cloud pixels in one or more of the plurality of images meets a threshold value, the method further includes determining predicted cloud movement relative to sun position. The method further includes causing a tint level of an electrochromic device to be controlled based on the predicted cloud movement relative to the sun position.

A computer, including a processor and a memory, the memory including instructions to be executed by the processor to determine a second convolutional neural network (CNN) training dataset by determining an underrepresented object configuration and an underrepresented noise factor corresponding to an object in a first CNN training dataset, generate one or more simulated images including the object corresponding to the underrepresented object configuration in the first CNN training dataset by inputting ground truth data corresponding to the object into a photorealistic rendering engine and generate one or more synthetic images including the object corresponding to the underrepresented noise factor in the first CNN training dataset by processing the simulated images with a generative adversarial network (GAN) to determine a second CNN training dataset. The instructions can include further instructions to train a CNN to using the first and the second CNN training datasets.

Implementations relate to diagnosis of crop yield predictions and/or crop yields at the field- and pixel-level. In various implementations, a first temporal sequence of high-elevation digital images may be obtained that captures a geographic area over a given time interval through a crop cycle of a first type of crop. Ground truth operational data generated through the given time interval and that influences a final crop yield of the first geographic area after the crop cycle may also be obtained. Based on these data, a ground truth-based crop yield prediction may be generated for the first geographic area at the crop cycle's end. Recommended operational change(s) may be identified based on distinct hypothetical crop yield prediction(s) for the first geographic area. Each distinct hypothetical crop yield prediction may be generated based on hypothetical operational data that includes altered data point(s) of the ground truth operational data.

In a computer-implemented method, one or more digital aerial images of a property are received, and processed according to one or more image analysis techniques to determine one or more features of the property, including one or more features of a tree located on or near the property. A time of year corresponding to at least one of the digital aerial images is determined, and the feature(s) of the property and the time of year are analyzed to determine a risk of damage to a structure located on the property. Based at least in part on the risk of damage, a risk output is generated that includes an indication of whether action should be taken to mitigate the risk of damage to the structure, an indication of whether insurance coverage associated with the structure should be offered, and/or a measure of the risk of damage to the structure.

A method for measuring a particles' precipitation rate includes the steps of acquiring at least one first image during a precipitation event through an image acquisition device having a sensor and lens; detecting the particles of the precipitation in the at least one first image by subtracting a background of the first image and setting a brightness threshold for detecting the particles, the particles being visible as a plurality of streaks in the image, wherein a first portion of the plurality of streaks comprises blurred streaks, and a second portion of the plurality of streaks comprises focused streaks; determining an apparent diameter and an apparent length for the plurality of streaks; estimating an actual diameter and an actual length for the plurality of streaks by resolving an equations' system having three equations and three unknowns, namely the actual diameter, the actual length and a depth position of the plurality of streaks, the depth position being the position of each particle from the lens, in which a first equation has the actual diameter as unknown in function of the depth position, a second equation has the actual length as unknown in function of the depth position and a third equation equals the theoretical terminal velocity of the particles with an estimated velocity of the particles in function of the depth position; estimating the velocity of the particles based on the ratio between a net streak length and an exposure time used to take at least one first image; estimating the particles' precipitation rate based on the actual diameter and the velocity (v).

In a method for training temporal precipitation interpolation models, the method may include receiving an initial image, a first intermediate image, and a final image, computing a first preliminary forward optical flow vector field from the initial image, and a first preliminary backward optical flow vector field, computing a first refined forward optical flow vector field and a first refined backward optical flow vector field using a terrain factor, among other things, and computing backpropagation losses to train neural networks by comparing the first intermediate image to an interpolated frame calculated using the first refined forward optical flow vector field and the first refined backward optical flow vector field.