Abstract:In the last decades, the development of smartphones, drones, aerial patrols, and digital cameras enabled high-quality photographs available to large populations and, thus, provides an opportunity to collect massive data of the nature and society with global coverage. However, the data collected with new photography tools is usually oblique - they are difficult to be georeferenced, and huge amounts of data is often obsolete. Georeferencing oblique imagery data may be solved by a technique called monoplotting, which only requires a single image and Digital Elevation Model (DEM). In traditional monoplotting, a human user has to manually choose a series of ground control point (GCP) pairs in the image and DEM and then determine the extrinsic and intrinsic parameters of the camera to establish a pixel-level correspondence between photos and the DEM to enable the mapping and georeferencing of objects in photos. This traditional method is difficult to scale due to several challenges including the labor-intensive inputs, the need of rich experience to identify well-defined GCPs, and limitations in camera pose estimation. Therefore, existing monoplotting methods are rarely used in analyzing large-scale databases or near-real-time warning systems. In this paper, we propose and demonstrate a novel semi-automatic monoplotting framework that provides pixel-level correspondence between photos and DEMs requiring minimal human interventions. A pipeline of analyses was developed including key point detection in images and DEM rasters, retrieving georeferenced 3D DEM GCPs, regularized gradient-based optimization, pose estimation, ray tracing, and the correspondence identification between image pixels and real world coordinates. Two numerical experiments show that the framework is superior in georeferencing visual data in 3-D coordinates, paving a way toward fully automatic monoplotting methodology.

A new parallel tool for classification of remotely sensed imagery

A Novel Terminal-Processing Application for Utilizing Satellite Imagery in Mobile GIS

ParSymG: a Parallel Clustering Approach for Unsupervised Classification of Remotely Sensed Imagery

Design and implementation of a parallel geographically weighted -nearest neighbor classifier.

Rational Design and Development of Radiation‐Sensitizing Histone Deacetylase Inhibitors

A NEW GENERATION IMAGE INTERPRETATION TECHNOLOGY BASED ON OBJECT-ORIENTED SEGMENTATION AND CLASSIFICATION

A new method of parallel sar image classification

Optimized Algorithm Module for Large Volume Remote Sensing Image Processing System

High Performance Hyperspectral Image Classification using Graphics Processing Units

A spatial Gaussian process method for hyperspectral remote sensing imagery classification

Self-adapted Genetic Hyperplane Classifier Algorithm for Multi-dimensional Remote Sensing Image

Adaptable Parallel Strategy To Extract Polygons From Massive Classified Images On Multi-Core Clusters

Object-oriented Classification of High Resolution Imagery Combining Support Vector Machine with Granular Computing

An automatic region-based image segmentation algorithm for remote sensing applications

Real-Time Semantic Segmentation using Hyperspectral Images for Mapping Unstructured and Unknown Environments

Target Classification and Recognition for High-Resolution Remote Sensing Images: Using the Parallel Cross-Model Neural Cognitive Computing Algorithm

Classifying geospatial objects from multiview aerial imagery using semantic meshes

AI-supported Framework of Semi-Automatic Monoplotting for Monocular Oblique Visual Data Analysis

Improved Mean-Shift Segmentation Approach for Remote Sensing Images

Using heterogeneous computing and edge computing to accelerate anomaly detection in remotely sensed multispectral images

An N-dimensional approach towards object based classification of remotely sensed imagery