There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

On robust estimation of the location parameter

A scheme is developed for classifying the types of motion perceived by a humanlike robot. It is assumed that the robot receives visual images of the scene using a perspective system model. Equations, theorems, concepts, clues, etc., relating the objects, their positions, and their motion to their images on the focal plane are presented. >

http://ieeexplore.ieee.org/iel2/815/3148/00100651.pdf

Robot vision

For 11 days in February 2000, the Shuttle Radar Topography Mission (SRTM) successfully recorded by interferometric synthetic aperture radar (InSAR) data of the entire land mass of the earth between 60°N and 57°S. The data acquired in C- and X-bands are processed into the first global digital elevation models (DEMs) at 1 arc sec resolution, by NASA-JPL and German aerospace center (DLR), respectively. From the perspective of the SRTM-X system, we give in this paper an overview of the mission and the DEM production, as well as an evaluation of the DEM product quality. Special emphasis is on challenges and peculiarities of the processing that arose from the unique design of the SRTM system, which has been the first single-pass interferometer in space.

The shuttle radar topography mission—a new class of digital elevation models acquired by spaceborne radar

As a SAR image is often the only available data in crisis situations, e.g. after an earthquake, a change analysis of the SAR image with previously acquired data may enable a fast analysis of the damage caused by the disaster. This paper presents a method for change analysis of the simulated appearance of a digital surface model (DSM) and a SAR image. A simulated SAR image is generated using the DSM and is separated into four masks indicating double bounce reflection, layover, shadow areas and ground reflection. Temporal changes of the corresponding areas in the real SAR image are analysed using pixel-based methods. Finally, the change detection results for the four masks are combined in order to obtain a robust change analysis result for single buildings. In this regard, the application of the proposed concept is presented for the city centre of Munich using TerraSAR-X data.

Detecting changes between a DSM and a high resolution SAR image with the support of simulation based separation of urban scenes

. Near infrared bands (NIR) provide rich information for many remote sensing applications. In addition to deriving useful indices to delineate water and vegetation, near infrared channels could also be used to facilitate image pre-processing. However, synthesizing bands from RGB spectrum is not an easy task. The inter-correlations between bands are not clearly identified in physical models. Generative adversarial networks (GAN) have been used in many tasks such as generating photorealistic images, monocular depth estimation and Digital Surface Model (DSM) refinement etc. Conditional GAN is different in that it observes some data as a condition. In this paper, we explore a cGAN network structure to generate a NIR spectral band that is conditioned on the input RGB image. We test different discriminators and loss functions, and evaluate results using various metrics. The best simulated NIR channel has a mean absolute error of around 5 percent in Sentinel-2 dataset. In addition, the simulated NIR image can correctly distinguish between various classes of landcover.

/pdf/generating-artificial-near-infrared-spectral-band-from-rgb-1ckgrb7e06.pdf

Generating artificial near infrared spectral band from rgb image using conditional generative adversarial network

Two-dimensional building footprints are a basis for many applications: from cartography to three-dimensional
building models generation. Although, many methodologies have been proposed for building footprint extraction,
this topic remains an open research area. Neural networks are able to model the complex relationships between
the multivariate input vector and the target vector. Based on these abilities we propose a methodology using
neural networks and Markov Random Fields (MRF) for automatic building footprint extraction from normalized
Digital Surface Model (nDSM) and satellite images within urban areas. The proposed approach has mainly two
steps. In the first step, the unary terms are learned for the MRF energy function by a four-layer neural network.
The neural network is learned on a large set of patches consisting of both nDSM and Normalized Difference
Vegetation Index (NDVI). Then prediction is performed to calculate the unary terms that are used in the MRF.
In the second step, the energy function is minimized using a max
ow algorithm, which leads to a binary building
mask. The building extraction results are compared with available ground truth. The comparison illustrates the
efficiency of the proposed algorithm which can extract approximately 80% of buildings from nDSM with high
accuracy.

/pdf/building-footprint-extraction-from-digital-surface-models-1f290et8u1.pdf

Building footprint extraction from digital surface models using neural networks

For a detailed monitoring of cities only very high resolution (VHR) satellite images as well as airborne imagery are able to provide the required accuracy. Especially not well mapped cities and new build up areas and their infrastructure like buildings, streets, bridges and other objects are of interest. Just the 2D information delivered by single images is often not sufficient when dealing with changes. Therefore stereo imagery is of importance to derive the height component which is of high value in analyzing 3D objects. The paper describes 3D change detection methodology if several Digital Surface Models (DSM) from different dates and sensors are available and provides solutions by the joint use of height and spectral information. The subtraction of DSMs, computed from image pairs acquired at different dates, should give direct results about changes. However, when at least one of the DSMs presents artifacts, a simple DSM subtraction often results in the detection of a combination of real and virtual changes. Therefore morphological based post-processing steps are proposed and adapted to different DSM qualities in order to remove artificial changes. In addition to height information, vegetation and shadow masks are generated, which are of great help in reconstructing sharper and more correct building boundaries. The used data are acquired over the city of Munich, Germany. They consist of a time series of stereo images acquired within the years 2005 to 2010 from space borne (IKONOS and Worldview) and airborne (3K-Camera-System) sensors. Evaluation of the proposed approaches in terms of false alarm, missed alarm and Kappa Coefficient etc is performed, proving its efficiency and relatively high accuracy for different kinds and combinations of stereo images and consequently different DSM qualities.

/pdf/multitemporal-3d-change-detection-in-urban-areas-using-4i7vua1z15.pdf

Multitemporal 3D Change Detection in Urban Areas Using Stereo Information from Different Sensors

Recent remote sensing applications are moving from an interpretation of single image swaths to regional mosaics. Within-swath and between-swath radiometric variations and ortho-rectification errors, especially for airborne sensors with a wide field of view, cause most of the problems during the mosaicking process. The proposed procedure for mosaicking consists of the following three steps. Within-swath radiometric normalization is performed using an image-based empirical radiometric correction method, which accounts for sensor viewing angle effects. Individual swaths are ortho-rectified using a direct geo-referencing approach. To remove the between-swath radiometric variations we propose to use the radiometric correction method, which is based on the information contained in the overlapping region of the swaths. The experiments for data acquired by airborne multi-spectral scanner DAEDALUS AADS 1268 ATM show the effectiveness and potential of the proposed method especially for the thematic analysis applications.

Peter Reinartz

Papers

Detecting changes between a DSM and a high resolution SAR image with the support of simulation based separation of urban scenes

Generating artificial near infrared spectral band from rgb image using conditional generative adversarial network

Building footprint extraction from digital surface models using neural networks

Multitemporal 3D Change Detection in Urban Areas Using Stereo Information from Different Sensors

Mosaicking of optical remote sensing imagery