# US 12,118,784 B2 · Methods and systems for image processing > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,118,784 B2 - Document ID: US12118784B2 - Status: Granted - Kind: B2 - Application number: 18/321,547 - Filed: 2023-05-22 - Published: 2024-09-05 - Granted: 2024-10-15 - Earliest claimed: 2023-03-02 - Assignees: RPS Group, Inc. - Applicants: RPS Group, Inc. - Inventors: Dave Houghton; Patrick Rath - Web page: https://patentgrep.com/patent/US12118784B2 ## CPC - G06V 20/194: {using hyperspectral data, i.e. more or other wavelengths than RGB} - G06V 10/36: Applying a local operator, i.e. means to operate on image points situated in the vicinity of a given point; Non-linear local filtering operations, e.g. median filtering - G06V 10/764: using classification, e.g. of video objects - G06V 20/13: Satellite images - G06V 20/17: taken from planes or by drones - G06V 10/16: {using multiple overlapping images; Image stitching} - G06V 10/7788: {the supervisor being a human, e.g. interactive learning with a human teacher} - G06V 10/82: using neural networks - G06V 2201/05: Recognition of patterns representing particular kinds of hidden objects, e.g. weapons, explosives, drugs - G06V 2201/10: Recognition assisted with metadata - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract Methods and systems for image processing for detection of devices are disclosed. Image data can be received. The image data can be filtered to provide output image data. The output image data can be classified. A pallet can be applied for displaying the classified image data ## Claims ### Claim 1 A computer implemented method of image processing for detection of explosive devices, the method comprising: receiving image data comprising aerial multispectral image data and/or aerial image data covering an area of terrain, wherein the image data comprises plural spectral channels in an electromagnetic spectrum; filtering the image data to provide output image data according to at least one filter in which each pixel value is a function of plural input image pixel values, wherein the at least one filter passes high reflectance from buried explosive devices and/or surface explosive devices; classifying the output image data into plural classes according to a classification scheme; applying a palette for displaying a classified image data, such that at least one class indicates a suspected explosive device with high contrast relative to surrounding areas; and interrogating the classified image data to detect explosive devices wherein detection is based on at least a shape of high contrast areas in the classified image data. ### Claim 2 (depends on 1) The method of claim 1, wherein the classification scheme comprises one or more of: natural breaks, equal counts, equal intervals, or any combination thereof. ### Claim 3 (depends on 1) The method of claim 1, wherein a number of classes is user adjustable. ### Claim 4 (depends on 1) The method of claim 1, wherein the palette is selectable from plural palettes. ### Claim 5 (depends on 1) The method of claim 1, wherein input data is receivable from an Unmanned Aircraft System flown at an altitude of between 5 m and 120 m. ### Claim 6 (depends on 5) The method of claim 5, comprising interrogating an output image to detect explosive devices based on a classified output image, wherein detection is based on: output pixel values, proximity to other areas of interest, contrast with surrounding areas, or other GIS landscape data, or any combination thereof. ### Claim 7 (depends on 6) The method of claim 6, comprising: iteratively training an AI system based on labelled input data to detect explosive devices. ### Claim 8 (depends on 7) The method of claim 7, wherein the AI system comprises a neural network. ### Claim 9 (depends on 7) The method of claim 7, wherein labelling comprises: a data ground truth mask indicating whether an image or portion of an image contains a landmine; and/or metadata comprising: mine type, time of day, weather, or GIS landscape data, or any combination thereof. ### Claim 10 (depends on 1) The method of claim 1, wherein the aerial multispectral image data comprises at least five channels comprising Red, Green, Blue, red edge, and near infrared. ### Claim 11 (depends on 1) The method of claim 1, comprising: receiving geotagged input images according to GPS data captured by a drone and/or triangulation position data of the drone; and creating an orthomosaic from plural images to obtain a larger image comprising a geotiff image covering the area of interest. ### Claim 12 (depends on 11) The method of claim 11, comprising receiving calibration images from drone cameras and calibrating a landscape images prior to processing. ### Claim 13 A system for detecting explosive devices, comprising: a processing device which when executing programming instructions is arranged to: receive image data comprising aerial multispectral image data and/or aerial image data covering an area of terrain, wherein the image data comprises plural spectral channels in an electromagnetic spectrum; filter the image data to provide output image data according to at least one filter in which each pixel value is a function of plural input image pixel values, wherein the at least one filter passes high reflectance from buried explosive devices and/or surface explosive devices; classify the output image data into plural classes according to a classification scheme; apply a palette for displaying classified image data, such that at least one class indicates a suspected explosive device with high contrast relative to surrounding areas; and interrogate the classified image data to detect explosive devices wherein detection is based on at least a shape of high contrast areas in the classified image data. ### Claim 14 (depends on 13) The system of claim 13, comprising an AI module arranged to interrogate the output image data to detect explosive devices based on the classified image data, wherein the AI module iteratively trains an AI system based on labelled input data to detect explosive devices. ### Claim 15 (depends on 13) The system of claim 13, comprising a drone arranged to capture geotagged image data and to transfer the geotagged image data to the processing device. ### Claim 16 (depends on 1) The method of claim 1 comprising: inputting filtered image data into a model trained to detect a presence or an absence of an explosive device in the image data, wherein the model comprises a convolutional neural network which automatically outputs a heat map prediction of whether or not an explosive device is present; and providing output to a user indicating the heat map prediction. ### Claim 17 (depends on 16) The method of claim 16, wherein a neural network is iteratively trained to perform one or both of: semantic segmentation of the image data on a training data set, the semantic segmentation comprising image samples and corresponding ground truth masks labelling whether or not the explosive device is present in a sample; and bounding box detection on the training data set, the bounding box detection comprising the image samples and corresponding coordinate labels of a location of the explosive device present in the sample. ### Claim 18 (depends on 16) The method of claim 16, wherein the image data comprises plural spectral channels in the electromagnetic spectrum, and wherein the method further comprises: prior to inputting image data to the model, filtering the image data to provide output image data according to a filter in which each pixel value is a function of plural input image pixel values, wherein the filter passes high reflectance from buried explosive devices and/or surface explosive devices. ### Claim 19 (depends on 16) The method of claim 16, wherein preprocessing comprises one or more of: downsampling, sharpening, and image augmentation operations. ### Claim 20 (depends on 1) Software stored on a computer readable medium comprising program instructions which when executed by a processing device cause the processing device to perform the method of claim 1.