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

Radiometric distortions along the scan line, caused mainly by the atmosphere, reflection properties of the imaged surface objects, and topography are observed frequently in images acquired by optical remote-sensing sensors, especially airborne sensors with a wide field of view. In practical applications, such as classification or mosaicking of image data, a relative correction or normalization of these radiometric distortions is needed. Therefore, the goal of this paper is to derive a relative correction algorithm for multispectral line scanner images, which allows the spectral radiance values for different scan angles to be converted to radiance values of a defined reference scan angle. Under the simplifying assumption of complete diffuse reflectors, these radiometric distortions are modelled using a linear equation between the radiances of a pair of scan angles for each class of surface objects separately. A linear regression, applied to all these equations, results in a linear model for the radiance correction, a new empirical radiometric correction method, which is a generalization of earlier proposed strictly additive or multiplicative models. The proposed method is based solely on the digital image data. So, no additional information or user interaction is required, and the method is fully automatic. Applications for data acquired with the airborne multispectral scanner DAEDALUS AADS 1268 ATM show the effectiveness and potential of the proposed method especially for mosaicking and classification purposes.

Radiometric normalization of sensor scan angle effects in optical remote sensing imagery

Recent sensors give valuable data for remote sensing applications Among these, building and change detection are important problems Therefore, researchers worked on these problems using both 2D and 3D data Some previous studies used only 2D data due to their availability Yet others used either 3D data alone or 2D and 3D data in a joint manner Besides, some studies only focused on building detection Yet others used detected building information in change detection In this study, we focus on 3D change detection based on building information Therefore, we first detect buildings At this step, we benefit from both 2D and 3D data Then, we locate changes based on these detected buildings We detect building pixels using panchromatic, multispectral, and Digital Surface Model (DSM) data using a decision tree classifier Then, we refine the detected building pixels using morphological and shape based operations Finally, we apply an object based hierarchical change detection method on the refined pixels We tested our method on 780 buildings and obtained promising results

/pdf/a-systematic-approach-for-building-change-detection-using-4dzyi2du18.pdf

A systematic approach for building change detection using multi-source data

With Time Delay and Integration (TDI) detector arrays, high resolution space camera systems can be developed. The main advantage of TDI-detectors in comparison to conventional CCD-detectors is their very high responsivity which allows extremely short exposure times. This short exposure times on the other hand correspond to very fast data read out rates, e.g. at a ground sampling distance of 0.25 m from orbit height, a line read out frequency of approximately 27 kHz is required. To test the image quality of TDI cameras at these very high data read out frequencies laboratory tests with a "Dalsa TDI-Line Scan Camera CT-E2" were carried out. The TDI-detector array of 2048 elements with a single detector size of 13 x 13 mm², has 96 integrating steps and a radiometric resolution of 8 bit. A moving test pattern simulated the relative motion between space camera and earth surface. The processed images of the test pattern could be used to determine three bar resolving power in across and along flight direction, to derive the MTF curves and to obtain a value for the noise level. Furthermore the influence of slight deviations from the required read out frequency on the resolving power was investigated. The paper describes the laboratory experiment and presents the resolving power results.

Resolution Tests with a TDI-Camera

The Earth Observation Center (EOC) of DLR realizes processors for ALOS and EnMAP high-resolution optical remote sensing satellite missions. The functional and developmental similarities and differences of the DLR processors are analyzed. It turns out that despite or precisely because of pan-chromatic and multispectral versus hyperspectral imaging as well as the post- versus pre-launch establishment of the DLR processing chains, both activities strongly benefit from each other.

/pdf/image-processing-chains-for-alos-and-enmap-data-similarities-3mi1hyrpwz.pdf

Image Processing Chains for ALOS and EnMAP Data: Similarities and Differences

. Small satellites play an increasing role in earth observation. This article evaluates different possibilities of utilizing data from Planet’s SkySat and PlanetScope satellites constellations for derivation of digital elevation models. While SkySat provides high resolution image data with a ground sampling distance of up to 50 cm, the PlanetScope constellation consisting of Dove 3U cubesats provide images with a resolution of around 4 m. The PlanetScope acquisition strategy was not designed for stereo acquisitions, but for daily acquisition of nadir viewing imagery. Multiple different products can be acquired by the SkySat satellites: Collects covering an area of usually 12 by 6 km, tri-stereo collects and video products with a framerate of 30 Hz. This study evaluates DSM generation using a Semi-Global Matching from multi date stereo pairs for SkySat and PlanetScope, and the dedicated Video and tri-stereo SkySat acquisitions. DSMs obtained by merging many PlanetScope across track stereo pairs show an normalized median deviation against LiDaR first pulse data of 5.2 meter over diverse landcover at the test sites around the city of Terrassa in Catalonia, Spain. SkySat tri-stereo products with 80 cm resolution reach an NMAD of 1.3 m over Terrassa.

/pdf/digital-elevation-models-from-stereo-video-and-multi-view-1ua0yu7oma.pdf

Peter Reinartz

Papers

Radiometric normalization of sensor scan angle effects in optical remote sensing imagery

A systematic approach for building change detection using multi-source data

Resolution Tests with a TDI-Camera

Image Processing Chains for ALOS and EnMAP Data: Similarities and Differences

Digital elevation models from stereo, video and multi-view imagery captured by small satellites