Fraunhofer-Institut für Optronik, Systemtechnik und Bildauswertung IOSB

Filters

13
Reset filters

Settings
Now showing 1 - 10 of 13
  • Publication
    Few-parameter learning for a hierarchical perceptual grouping system
    ( 2023)
    Michaelsen, Eckart
    Perceptual grouping along well-established Gestalt laws provides one set of traditional methods that provide a tiny set of meaningful parameters to be adjusted for each application field. More complex and challenging tasks require a hierarchical setting, where the results aggregated by a first grouping process are later subject to further processing on a larger scale and with more abstract objects. This can be several steps deep. An example from the domain of forestry provides insight into the search for suitable parameter settings providing sufficient performance for the machine-vision module to be of practical use within a larger robotic control setting in this application domain. This sets a stark contrast in comparison to the state-of-the-art deep-learning neural nets, where many millions of obscure parameters must be adjusted properly before the performance suffices. It is the opinion of the author that the huge freedom for possible settings in such a high-dimensional inscrutable parameter space poses an unnecessary risk. Moreover, few-parameter learning is getting along with less training material. Whereas the state-of-the-art networks require millions of images with expert labels, a single image can already provide good insight into the nature of the parameter domain of the Gestalt laws, and a domain expert labeling just a handful of salient contours in said image yields already a proper goal function, so that a well working sweet spot in the parameter domain can be found in a few steps. As compared to the state-of-the-art neural nets, a reduction of six orders of magnitude in the number of parameters results. Almost parameter-free statistical test methods can reduce the number of parameters to be trained further by one order of magnitude, but they are less flexible and currently lack the advantages of hierarchical feature processing.
  • Publication
    Explorations on the Depth of Gestalt Hierarchies in Social Imagery
    ( 2021)
    Michaelsen, Eckart
    Apart from machine learning and knowledge engineering, there is a third way of challenging machine vision - the Gestalt-law school. In an interdisciplinary effort between psychology and cybernetics, compositionality in perception has been studied for at least a century along these lines. Hierarchical compositions of parts and aggregates are possible in this approach. This is particularly required for high-quality high-resolution imagery becoming more and more common, because tiny details may be important as well as large-scale interdependency over several thousand pixels distance. The contribution at hand studies the depth of Gestalt-hierarchies in typical social image genres-portraits and group pictures-exemplarily, and outlines technical means for their automatic extraction. The practical part applies bottom-up hierarchical Gestalt grouping as well as topdown search focusing and constraint enforcement, listing as well success as failure. In doing so, the paper discusses exemplarily the depth and nature of such compositions in imagery relevant to human beings.
  • Publication
    Estimating Efforts and Success of Symmetry-Seeing Machines by Use of Synthetic Data
    ( 2019)
    Michaelsen, Eckart
    ;
    Vujasinovic, Stéphane
    Representative input data are a necessary requirement for the assessment of machine-vision systems. For symmetry-seeing machines in particular, such imagery should provide symmetries as well as asymmetric clutter. Moreover, there must be reliable ground truth with the data. It should be possible to estimate the recognition performance and the computational efforts by providing different grades of difficulty and complexity. Recent competitions used real imagery labeled by human subjects with appropriate ground truth. The paper at hand proposes to use synthetic data instead. Such data contain symmetry, clutter, and nothing else. This is preferable because interference with other perceptive capabilities, such as object recognition, or prior knowledge, can be avoided. The data are given sparsely, i.e., as sets of primitive objects. However, images can be generated from them, so that the same data can also be fed into machines requiring dense input, such as multilayered perceptrons. Sparse representations are preferred, because the author's own system requires such data, and in this way, any influence of the primitive extraction method is excluded. The presented format allows hierarchies of symmetries. This is important because hierarchy constitutes a natural and dominant part in symmetry-seeing. The paper reports some experiments using the author's Gestalt algebra system as symmetry-seeing machine. Additionally included is a comparative test run with the state-of-the-art symmetry-seeing deep learning convolutional perceptron of the PSU. The computational efforts and recognition performance are assessed.
  • Publication
    Foreword to the Special Issue on Advances in Pattern Recognition in Remote Sensing
    ( 2018)
    Michaelsen, Eckart
    ;
    Shan, J.
    ;
    Stilla, Uwe
    ;
    Li, W.
    The papers in this special section were presented at the Pattern Recognition in Remote Sensing (PRRS) Workshop. The technical committee for pattern recognition in remote sensing (PRRS) and mapping (TC7) of the international association for pattern recognition (IAPR) organizes a biennial workshop on PRRS. This workshop series has been a popular forum for experts of both communities, pattern recognition and remote sensing, and accordingly there is co-sponsoring from IAPR, IEEE-GRSS, and ISPRS. In particular, the intercommission working group for pattern analysis in remote sensing (ICWG II/III) of the ISPRS is closely cooperating. The latest workshops have been very successfully held in December 2016 in Cancun, Mexico, and in August 2018 in Beijing, China. Additionally, TC7 compiles special issues on PRRS in journals such as IEEE-JSTARS or Pattern Recognition Letters [items 1)-6) in Related Works]. The issue at hand fits in this series. Contained are 29 papers, including some work of authors that have participated and published preceding work in the 2016 PRRS in Cancun.
  • Publication
    On the automation of gestalt perception in remotely sensed data
    ( 2018)
    Michaelsen, Eckart
    Gestalt perception, the laws of seeing, and perceptual grouping is rarely addressed in the context of remotely sensed imagery. The paper at hand reviews the corresponding state as well in machine vision as in remote sensing, in particular concerning urban areas. Automatic methods can be separated into three types: 1) knowledge-based inference, which needs machine-readable knowledge, 2) automatic learning methods, which require labeled or un-labeled example images, and 3) perceptual grouping along the lines of the laws of seeing, which should be pre-coded and should work on any kind of imagery, but in particular on urban aerial or satellite data. Perceptual grouping of parts into aggregates is a combinatorial problem. Exhaustive enumeration of all combinations is intractable. The paper at hand presents a constant-false-alarm-rate search rationale. An open problem is the choice of the extraction method for the primitive objects to start with. Here super-pixel-segmentation is used.
  • Publication
  • Publication
    Self-organizing maps and Gestalt organization as components of an advanced system for remotely sensed data: An example with thermal hyper-spectra
    ( 2016)
    Michaelsen, Eckart
    The thermal hyper-spectral data provided for research by IEEE-GRS/Telops in 2014 give an interesting example for combining such spectral information with perceptual grouping according to Gestalt laws in the geographic plane. Self-organizing maps are used for unsupervised learning. By watershed segmentation and subsequent merging on the map certain classes are found automatically. Back-projection of these regions to the geographic plane reveals considerable coincidence with classes of human interest such as vegetation, building roofs, and roads, respectively. The partial ground-truth provided with the data by IEEE-GRS/Telops allows the estimation of quantitative recognition accuracies. Human observers assign such meaning to the back-projected segments relying mainly on their perceptual grouping capabilities - roads appear as elongated stripes organized in a net, buildings come as blobs in organized patterns of repetitive rows and mirror-symmetry, and subsequently the rest is inferred as probably being vegetated. The automatic Gestalt grouping presented in this work follows the rules of Gestalt algebra. Gestalt hierarchies of depth three can be instantiated on the building class in accordance with human perception. Interesting feed-back possibilities are proposed from the perceptual grouping to the interpretation of the segments on the self-organizing map and further on to the assignment of meaning to the spectra. Again the ground-truth is used to estimate the gain quantitatively.
  • Publication
    A single performance characteristic for the evaluation of seeker tracking algorithms
    ( 2014)
    Doktorski, Leo
    ;
    Michaelsen, Eckart
    ;
    Repasi, Endre
    This paper presents a single numerical performance characteristic for the evaluation of seeker tracking algorithms. It concentrates on ship IR seeker tracking algorithms. Assessing the threat from guided missiles needs a sound evaluation of their performance. The main goal is to introduce a characteristic which is able to assess the threat for ships depending on various scenario parameters. It is shown that for these applications such a single characteristic is sufficient. In order to achieve this five popular tracking algorithms are used. Synthetic IR image sequences are generated to simulate a large set of attack approaches and assemble sufficient statistics on the behavior of the algorithms. The introduced characteristic can also be used for investigations on algorithms themselves, e.g. for sensitivity analyses and parameter optimization of a single algorithm, and for comparison of different algorithms.
  • Publication
    Simple gestalt algebra
    ( 2014)
    Michaelsen, Eckart
    ;
    Yashina, V.V.
    The laws of Gestalt perception rule how parts are assembled into a perceived aggregate. This contribution defines them in an algebraic setting. Operations are defined for mirror symmetry, repetition in rows, and arrangement in rotational symmetry patterns respectively. While the mirror operation is a classical binary operation, the other two operations are of arity n > 1. Thus the Gestalt domain with its three operations forms a general algebra. Deviations from the perfect mutual positioning are handled using positive and differentiable assessment functions achieving maximal value for the case of perfect symmetry and approaching zero if the parts mutually violate the Gestalt laws. Theorems of closure are proven, stating that any of the operations on any Gestalten will produce again a well-defined new Gestalt. It is also proven that no neutral elements and no inverse Gestalten exist for the three operations. Practically, these definitions and calculations can be used in two ways: 1. Images with Gestalts can be rendered by using random decisions with the assessment functions as densities; 2. given an image (in which Gestalts are supposed) Gestalt-terms are constructed successively, and the ones with high assessment values are accepted as plausible, and thus recognized.