Showing papers on "Feature extraction published in 1988"

Techniques for calibration of the scale factor and image center for high accuracy 3-D machine vision metrology

Abstract: Techniques are described for calibrating certain intrinsic camera parameters for machine vision. The parameters to be calibrated are the horizontal scale factor, and the image center. The scale factor calibration uses a one-dimensional fast Fourier transform and is accurate and efficient. It also permits the use of only one coplanar set of calibration points for general camera calibration. Three groups of techniques for center calibration are presented: Group I requires using a laser and a four-degree-of-freedom adjustment of its orientation, but is simplest in concept and is accurate and reproducible; Group II is simple to perform, but is less accurate than the other two; and the most general, Group II, is accurate and efficient, but requires a good calibration plate and accurate image feature extraction of calibration points. Group II is recommended most highly for machine vision applications. Results of experiments are presented and compared with theoretical predictions. Accuracy and reproducibility of the calibrated parameters are reported, as well as the improvement in actual 3-D measurement due to center calibration. >

Dynamic monocular machine vision

Abstract: A new approach to real-time machine vision in dynamic scenes is presented based on special hardware and methods for feature extraction and information processing. Using integral spatio-temporal models, it by-passes the nonunique inversion of the perspective projection by applying recursive least squares filtering. By prediction error feedback methods similar to those used in modern control theory, all spatial state variables including the velocity components are estimated. Only the last image of the sequence needs to be evaluated, thereby alleviating the real-time image sequence processing task.

A New Scheme for Practical Flexible and Intelligent Vision Systems

Abstract: We propose a new scheme for practical vision systems which are simple in structure, directly and adaptively trainable for various purposes. The feature extraction consists of two stages: At first, general and primitive features which are shift-invariant and additive are extracted over the retinal image plane. Then those features are linearly combined on the bases of multivariate analysis methods so as to provide new effective features for each task. Such a system can adaptively and automatically learn the task from the given supervised training examples and show an intelligent performance. Some examples of practical applications to object recognition and measurements are experimentally shown. which is simple in structure and directly and adap tively trainable for various purposes. The basic idea is similar to Perceptron [I] or neural networks (for example, see 12)) but more practical being based on closed-form solutions for statistical performance criteria without employing such slow iterative learning processes.

Cooperative methods for road tracking in aerial imagery

Abstract: A description is given of research in digital mapping and image understanding in the area of automated feature extraction from aerial imagery. The authors discuss a system for road tracking, ARF (A Road Follower), that uses multiple cooperative methods for extracting information about road location and structure from complex aerial imagery. This system is a multilevel architecture for image analysis that follows for cooperation among low-level processes and aggregation of information by high-level analysis components. Two low-level road tracking methods have been implemented: road-surface texture correlation and road-edge following. Each works independently to establish a model of the centerline of the road, its width, and other local properties. >

A rule based approach for visual pattern inspection

Abstract: The authors investigate the use of a priori knowledge about a scene to coordinate and control bilevel image segmentation, interpretation, and shape inspection of different objects in the scene. The approach is composed of two main steps. The first step consists of proper segmentation and labeling of individual regions in the image for subsequent ease in interpretation. General as well as scene-specific knowledge is used to improve the segmentation and interpretation processes. Once every region in the image has been identified, the second step proceeds by testing different regions to ensure they meet the design requirements, which are formalized by a set of rules. Morphological techniques are used to extract certain features from the previously processed image for rule verification purposes. As a specific example, results for detecting defects in printed circuit boards are presented. >

Pattern recognition device

Abstract: PURPOSE:To identify a similar character group with high accuracy without increasing dictionary capacity by utilizing the property that pieces of pattern information by directions which are found by a recognition part represent the fine structure of character segments CONSTITUTION:A normalization pattern which is preprocessed is applied to a directional feature extraction part 2 to generate four feature patterns by directions Further, feature surfaces by the directions are fogged and resampled to compares information The extracted directional features are applied to a matching part 25 to calculate the similarity with standard patterns stored in a feature dictionary 26 A candidate character kind storage part 27 select similar character couples successively from a similar character group decided by the recognition part 12 and a difference information extraction part 28 stores directional features of an input patter in an input feature pattern Difference patterns of similar character couples stored in a difference patterns dictionary 29 are read out to cut a difference part according to directional features corresponding to the input character pattern and a standard pattern, so that the similarity between the directional feature of the input character pattern and standard pattern is calculated by a character kind couple identification part 30

Three-dimensional vision structure for robot applications

Abstract: Focuses on the structure of robot sensing systems and the techniques for measuring and preprocessing 3-D data. To get the information required for controlling a given robot function, the sensing of 3-D objects is divided into four basic steps: transduction of relevant object properties (primarily geometric and photometric) into a signal; preprocessing the signal to improve it; extracting 3-D object features; and interpreting them. Each of these steps usually may be executed by several alternative techniques (tools). Tools for the transduction of 3-D data and data preprocessing are surveyed. The performance of each tool depends on the specific vision task and its environmental conditions, both of which are variable. Such a system includes so-called tool-boxes, one box for each sensing step, and a supervisor, which controls iterative sensing feedback loops and consists of a rule-based program generator and a program execution controller. Sensing step sequences and tools are illustrated for two 3-D vision applications at SRI International Company: visually guided robot arc welding and locating identical parts in a bin. >

An automatic circuit diagram reader with loop-structure-based symbol recognition

Abstract: A high-performance logic circuit diagram reader was developed for VLSI-CAD data input. Almost all logic circuit symbols include one or more loop structures. A description is given of an efficient method for recognition of these loop-structured symbols. The proposed method consists of two processes: symbol segmentation and symbol identification. Symbol identification is achieved by a powerful hybrid method which uses heuristics to mediate between template matching and feature extraction. The entire symbol recognition process is carried out under a decision-tree control strategy. The entire recognition system for circuit diagrams is briefly explained, including character string recognition and connecting line analysis. >

Automated X-ray inspection of aluminum castings

Abstract: The experience gained with several approaches to automatic flaw detection in X-ray images of cast aluminum wheels is described. Basic problems are mentioned, and the applicability of segmentation methods to actual inspection tasks is demonstrated. The discussion focuses on the definition, extraction, and combination of local features for pixel classification. Results of pilot tests are described briefly. Further investigations are suggested, aiming at more generality of the methods and greater stability of the segmentation. >

Finite prolate spheroidal sequences and their applications. II. Image feature description and segmentation

Abstract: For pt.I see ibid., vol.9, no.6, p.787 (1987). The problem of uncertainty in image feature description is discussed, and it is shown how finite prolate spheroidal sequences can be used in the construction of feature descriptions that combine spatial and frequency-domain locality in an optimal way. Methods of constructing such optimal feature sets, which are suitable for graphical implementation, are described, and some generalizations of the quadtree concept are presented. These methods are illustrated by examples from image processing applications, including feature extraction and texture description. The problem of image segmentation is discussed, and the importance of scale invariance in overcoming the limitations imposed by uncertainty is demonstrated. A novel texture segmentation algorithm that is based on a combination of the new feature description and multiresolution techniques is described and shown to give accurate segmentations on a range of synthetic and natural textures. >

Object-oriented feature extraction method for image data compaction

Abstract: An online unsupervised feature-extraction method for high-dimensional remotely sensed image data compaction is proposed. This method is directed at the reduction of data redundancy in the scene representation of satellite-borne, high-resolution multispectral sensor data. The algorithm partitions the observation space into an exhaustive set of disjoint objects, and pixels belonging to each object are characterized by an object feature. The set of object features, rather than the pixel features, is used for data transmission and classification. Illustrative examples of high-dimensional image data compaction are presented, and the feature representation performance is investigated. Example results show an average compaction coefficient of more than 25 to 1 when this method is used; the classification performance is improved slightly by using object features rather than the original data, and the CPU time required for classification is reduced by a factor of more than 25 as well. The feature extraction CPU time is less than 15% of CPU time for original data classification. >

Validation of magnetic resonance imaging (MRI) multispectral tissue classification

Abstract: NASA multispectral image processing technology has been adapted to the analysis of magnetic resonance imaging (MRI) scans. A statistical evaluation of multispectral analysis applied to tissue classification in MRI brain scans has been performed. The radiometric and geometric distortions introduced by the MRI scanner can be removed before feature extraction and classification. Signatures may be derived from one set of multispectral MR images at a single anatomic level and be applied in the same subject at later times, to other anatomic levels containing the same tissues, or to other subjects. >

Cartographic feature extraction with integrated SIR-B and Landsat TM images

Abstract: A digital cartographic multisensor image database of excellent geometry and improved resolution was created by registering SIR-B images to a rectified Landsat TM reference image and applying intensity-hue-saturation enhancement techniques. When evaluated against geodetic control, RMSE(XY) values of approximately + or - 20 m were noted for the composite SIR-B/TM images. The completeness of cartographic features extracted from the composite images exceeded those obtained from separate SIR-B and TM image data sets by approximately 10 and 25 percent, respectively, indicating that the composite images may prove suitable for planimetric mapping at a scale of 1:100,000 or smaller. At present, the most effective method for extracting cartographic information involves digitizing features directly from the image processing display screen.

Texture inspection with self-adaptive convolution filters

Abstract: A resolution-independent method for detection of imperfections in quasi-periodic textures is described. After image standardization, the period is estimated in the horizontal and vertical directions. This determines the size of a sparse convolution mask. Mask coefficients are determined by the well-known technique of eigenfilter extraction. The method thus offers a completely automated generation of a bank of suitable filters, the form and the coefficients of which are made dependent on the texture type to be inspected. After feature extraction in the filtered images, a Mahalanobis classifier is applied. >

Morphological skeleton representation and shape recognition

Abstract: The nature of the morphological skeleton representation of a binary shape is related to the composition of structuring elements through the distance function defined by morphological set transforms in digital space. Two digital metrics, uniform-step distance and periodically-uniform-step distance, are introduced to provide useful spatial measures for morphological transforms. A natural shape representation by ribbonlike components is accomplished by extraction of skeletal feature primitives from the morphological skeleton of a shape. The hierarchical structure of the representation makes it stable and insensitive to noise disturbance. The matching is a simple top-down process in which the inverses of the skeletal feature primitives at each level are compared. The recognition is based on the similarity measure provided by the matching process. >

Comparison of statistical pattern-recognition algorithms for hybrid processing. I. Linear-mapping algorithms

Abstract: Two groups of pattern-recognition algorithms for hybrid optical–digital computer processing are theoretically and experimentally compared. The first group is based on linear mapping, while the second group is based on feature extraction and eigenvector analysis. We study the relations among various linear-mapping-based algorithms by formulating a more general unified pseudoinverse algorithm. We show that the least-squares linear-mapping technique, the simplified least-squares linear-mapping technique, the synthetic discriminant function, the equal-correlation-peak method, and the Caulfield–Maloney filter are in fact all special cases of the unified pseudoinverse algorithm. When the total number of the training images (KM, where K is the number of classes and M is the number of training images in each class) is larger than the dimension of the images (N), the overdetermined case of the unified pseudoinverse algorithm is the same as the least-squares linear-mapping technique, because both algorithms are based on optimization processes of minimization of the mean-square error. When KM < N, the underdetermined case of the unified pseudoinverse algorithm is the same as the least-squares linear-mapping technique and the synthetic discriminant function. Furthermore, when KM < N, the synthetic discriminant function method can be considered a degenerate case of the least-squares linear-mapping technique. Among the algorithms studied, the simplified least-squares linear-mapping technique requires the least computation time for filter synthesis. Experimental results on classification with linear-mapping-based algorithms are provided and show good agreement with the theoretical analysis.

Sensing and recognition of rigid objects using structured light

Abstract: Word directed toward the development of a vision system for bin picking of rigid 3D objects is reported. Any such system must have components for sensing, feature extraction, modeling, and matching. A structured light system which attempts to deliver a rich 2/sup 1///sub 2/D representation of the scene is described. Surface patches are evident as connected sets of stripes whose 3D coordinates are computed by means of triangulation and constraint propagation. Object edges are detected by the intersection of surface patches or by backprojecting image edges to intersect with the patches. Two matching paradigms are given for drawing correspondence between structures in the scene representation and structures in models. Three major contributions are reported: a method for sensing object surface patches without having to solve uniquely for stripe labels; the use of both an intensity image and a striped image, allowing scenes to be represented by detected edges along with 3D surface patches; and a pose-clustering algorithm, a uniform technique to accumulate matching evidence for recognition while averaging out substantial errors of pose. >

FRAFES, a frame-based feature extraction system

Abstract: A prototype feature extraction system based on frame representation of part data and features is presented. A feature description language (FDL), based on semantic nets, is described. The objective of this approach is to provide an intelligent and flexible link between design and manufacturing. Finally, the system implementation and sample results are discussed. >

Image compression using a neural network

Abstract: Data compression of speckled images poses a non-trivial model identification problem. We train an unsupervised neural network on a set of archetype images in order to form an internal representation (or model) of the image features. We find that a multi-layer topographic mapping network has the necessary properties successfully to compress and reconstruct imagery. We show how to extend and improve upon existing learning algorithms for this type of network, and we express the network learning dynamics as a diffusion equation. We then present some examples of the application of this technique to synthetic aperture radar images.

Feature extraction by rank-order filtering for image segmentation

Abstract: The aim of this paper is to outline a unified approach to feature extraction for segmentation purposes by means of the rank-order filtering of grey values in a neighbourhood of each pixel of a digitized image. In the first section an overview of rank-order filtering for image processing is given, and a fast histogram algorithm is proposed. Section 2 deals with the extraction of a “locally most representative grey value”, defined as the maximum of the local histogram density function. In Section 3 several textural features are described, which can be extracted from the local histogram by means of rank-order filtering, and their properties are discussed. Section 4 formulates some general requirements to be met by the process of image segmentation, and describes a method based upon the features introduced in the former sections. In the last section some experimental results applied to aerial views obtained with the segmentation method of Sect. 4 are reported. These test images have been analyzed within the scope of an investigation centered on terrain recognition for agricultural and ecological purposes.

An optical Fourier/electronic neurocomputer automated inspection system

Abstract: An optical Fourier/electronic neurocomputer automated inspection system prototype is described. The system is composed to two modules: (1) a video-input optical/electronic Fourier feature extraction module, and (2) a PC/AT-based neurocomputer for feature signature (i.e., image) classification. Global shape and texture analysis, capable of discriminating relatively small and unpredictable image differences, is performed at speeds up to 15 images/s. The system performs 2-D image data compression by utilizing the attractive properties of coherent optical Fourier transform generation and optical feature sampling. Neural network multiclass pattern classifier algorithms (i.e., backpropagation and counterpropagation) are used to ensure system robustness in the presence of noisy, degraded, partial, or distorted images. It is expected that discrimination results can be used to track image-change trends for adaptive process control. Preliminary experimental results are presented. >

On The Congruence Of Noisy Images To Line Segment Models

Abstract: An algorithm is described for matching two-dimensional images that does not depend on the extraction of features, a step which is often difficult or impractical. It combines computational efficiency with robustness against noisy data. It has been tested with two kinds of data. The algorithm matches a two-dimensional image, in the form of a set of points, to a line segment model. It estimates a positive congruence (i.e., a translation plus a rotation) that moves the image onto the model. The algorithm is most appropriate for matching when the required congruence is small, and it seems very well suited to motion-tracking. Experimental results have been obtained for two kinds of images: (1) points obtained from an optical range finder, and (2) edge points extracted from an intensity image. Only the former results are presented.

Pixel-Level Sensor Fusion For Improved Object Recognition

Abstract: A method is proposed to exploit simultaneous, co-registered FLIR and TV images of isolated objects against relatively bland backgrounds to improve recognition of those objects The method uses edges extracted from the TV imagery to segment objects in the FLIR imagery A binary tree classifier is shown to perform significantly better with objects defined in this manner than with objects extracted separately from the FLIR or TV images, or with a feature level fusion scheme which combines features of separately extracted objects The structure of the tree indicates that the cross-segmented objects are simply ordered in feature space An argument is presented that this sensor fusion scheme is natural in terms of the known or-ganization of neural vision systems Generalizations to other sensor types and fusion schemes should be considered, since it has been shown that co-registered imagery can be exploited to improve recognition at no additional computational cost

Shape Recognition Using A CMAC Based Learning System

Abstract: This paper discusses pattern recognition using a learning system which can learn an arbitrary function of the input and which has built-in generalization with the characteristic that similar inputs lead to similar outputs even for untrained inputs. The amount of similarity is controlled by a parameter of the program at compile time. Inputs and/or outputs may be vectors. The system is trained in a way similar to other pattern recognition systems using an LMS rule. Patterns in the input space are not separated by hyperplanes in the way they normally are using adaptive linear elements. As a result, linear separability is not the problem it is when using Perceptron or Adaline type elements. In fact, almost any shape category region is possible, and a region need not be simply connected nor convex. An example is given of geometric shape recognition using as features autoregressive model parameters representing the shape boundaries. These features are approximately independent of translation, rotation, and size of the shape. Results in the form of percent correct on test sets are given for eight different combinations of training and test sets derived from two groups of shapes.

Automated fast recognition and location of arbitrarily shaped objects by image morphology

Abstract: Morphological operations are used for segmentation, feature generation and location extraction. A recursive adaptive thresholding algorithm transforms a gray-level image into a set of multiple level regions of objects. A distance transformation algorithm then is used to transform a binary image into the minimum distance from each object point to the object's boundary. This algorithm uses a morphological erosion with a large structuring element which may correspond to Euclidean, city-block, or chessboard distance measures. A shape library database with hierarchical features is automatically generated. The features extracted are the shape number and the skeletal local-maximum points radii and coordinates. Object recognition is achieved by comparing the shape number and the hierarchical radii. Object location is detected by a hierarchical morphological bandpass filter. >