Showing papers on "Feature extraction published in 1985"

Classification of textures using Gaussian Markov random fields

Abstract: The problem of texture classification arises in several disciplines such as remote sensing, computer vision, and image analysis. In this paper we present two feature extraction methods for the classification of textures using two-dimensional (2-D) Markov random field (MRF) models. It is assumed that the given M × M texture is generated by a Gaussian MRF model. In the first method, the least square (LS) estimates of model parameters are used as features. In the second method, using the notion of sufficient statistics, it is shown that the sample correlations over a symmetric window including the origin are optimal features for classification. Simple minimum distance classifiers using these two feature sets yield good classification accuracies for a seven class problem.

Rule-Based Interpretation of Aerial Imagery

Abstract: In this paper, we describe the organization of a rule-based system, SPAM, that uses map and domain-specific knowledge to interpret airport scenes. This research investigates the use of a rule-based system for the control of image processing and interpretation of results with respect to a world model, as well as the representation of the world model within an image/map database. We present results on the interpretation of a high-resolution airport scene wvhere the image segmentation has been performed by a human, and by a region-based image segmentation program. The results of the system's analysis is characterized by the labeling of individual regions in the image and the collection of these regions into consistent interpretations of the major components of an airport model. These interpretations are ranked on the basis of their overall spatial and structural consistency. Some evaluations based on the results from three evolutionary versions of SPAM are presented.

Applications of Tensor Theory to Object Recognition and Orientation Determination

Abstract: A method is developed by which images resulting from orthogonal projection of rigid planar-patch objects arbitrarily oriented in three-dimensional (3-D) space may be used to form systems of linear equations which are solved for the affine transform relating the images. The technique is applicable to complete images and to unlabeled feature sets derived from images, and with small modification may be used to transform images of unknown objects such that they represent images of those objects from a known orientation, for use in object identification. No knowledge of point correspondence between images is required. Theoretical development of the method and experimental results are presented. The method is shown to be computationally efficient, requiring O(N) multiplications and additions where, depending on the computation algorithm, N may equal the number of object or edge picture elements.

Experiments in Intensity Guided Range Sensing Recognition of Three-Dimensional Objects

Abstract: With the advent of devices that can directly sense and determine the coordinates of points in space, the goal of constructing and recognizing descriptors of three-dimensional (3-D) objects is attracting the attention of many researchers in the image processing community. Unfortunately, the time required to fully sense a range image is large relative to the time required to sense an intensity image. Conversely, a single intensity image lacks the depth information required to construct 3-D object descriptors. This paper presents a method of combining the two sensory sources, intensity and range, such that the time required for range sensing is considerably reduced. The approach is to extract potential points of interest from the intensity image and then selectively sense range at these feature points. After the range information is known at these points, a graph structure representing the object in the scene is constructed. This structure is compared to the stored graph models using an algorithm for partial matching. The results of applying the method to both synthetic data and real intensity/range images are presented.

Feature extraction techniques for recognizing solid objects with an ultrasonic range sensor

Abstract: Ultrasonic range sensing has attracted attention in the robotics community because of it simplicity in construction and low cost. However, determining range direction rather than just range magnitude is made difficult by the expanding signal beam of the sensor. This direction ambiguity can be reduced to some extent by increasing the operating frequency or diameter of the sensor, but some ambiguity will still remain. A technique is described for obtaining the true direction to a planar surface using three sensors or three positions of one sensor. A direct solution of the vector equation is discussed to illustrate the solution complexity in direct form. A simplifying transformation applied to the direct form and a further simplifying sensor configuration are described which greatly reduces the solution complexity. An alternate formulation of the problem is described from which a geometrical insight can be obtained. This alternate solution also lends itself to the generation of tangential constraint surfaces for bounding curved object surfaces. These local feature extraction techniques are extended to perform surface tracking for the extraction of global surface features. The connectivity of local features provides additional information. From this information classification of solid objects is possible.

Feature Extraction Techniques for Sorting Tomatoes by Computer Vision

Abstract: SPECIAL algorithms were developed using digital image analysis and pattern recognition techniques for orientation of fresh market tomatoes and classification based on size, shape, color and surface defects. Performance testing of the classification algorithms demonstrated the feasibility of applying computerized visual inspection to tomatoes.

Pattern recognition device

Abstract: A pattern recognition device is arranged to have learning of a reference pattern vector carried out in a recognition unit by making use of the pause periods in the recognition processing, without particularly providing a learning section for learning the reference pattern vector. Namely, a part or the entirety of the arithmetic processing unit where the recognition result is obtained in the recognition unit by collating the input pattern with the recognition dictionary, can be utilized as the learning portion of the reference pattern vector. In concrete terms, the operation of multiplication-accumulation (inner product) which represents the main operation in the recognition processing and the learning processing, can be carried out by means of similar processes of handling. Therefore, by utilizing the processing section for sum of the products operation, of the recognition unit, which is in the idle state for pattern recognition, it becomes possible to carry out learning the reference pattern vector efficiently in time, without forcibly interrupting the pattern recognition processing by formally setting a learning condition.

Pattern recognition system

Abstract: PURPOSE:To attain ease of forming of a dictionary by comparing and operating plural extracted feature parameters and a dictionary, storing the result and using a dictionary vector selected from the storage as the result of recognition so as to realize an excellent pattern recognition. CONSTITUTION:A comparison operation circuit 3 uses a new picture element and a dictionary vector being the result of preceding operation in a storage device 4 for the purpose of operation every time a new specific element of a feature parameter Pi is given from a feature extraction circuit 1 with the operation of a control circuit 9 and the dictionary vector is stored in a storage device 5. Then the dictionary vector stored in the storage devices 4, 5 is inputted to a maximum similarity decision circuit 8 and when the selected dictionary is one, its dictionary vector is outputted as the result of recognition. Moreover, when plural dictionary vectors are obtained, a vector having the maximum similarity is selected and used as the result of recognition.

Parallel Algorithms for Syllable Recognition in Continuous Speech

Abstract: A distributed rule-based system for automatic speech recognition is described. Acoustic property extraction and feature hypothesization are performed by the application of sequences of operators. These sequences, called plans, are executed by cooperative expert programs. Experimental results on the automatic segmentation and recognition of phrases, made of connected letters and digits, are described and discussed.

Rotation-invariant pattern recognition using optimum feature extraction

Abstract: In this paper we consider the problem of detecting a target regardless of its orientation when it is known that the target must be from one of two classes. We assume significant random intraclass variability, a complication which requires techniques from statistical pattern recognition for amelioration. The Foley-Sammon transformation for selecting optimum features from random training samples is used to solve the problem.

Feature extraction technique for use in a pattern recognition system

Abstract: A method is provided for forming feature vectors representing the pixels contained in a pattern desired to be recognized, and reference patterns. One part of the feature vector is representative of the pixels contained in the pattern itself, while not requiring a very large feature vector which exactly defines each pixel of the pattern. One embodiment of this invention provides that another part of the feature vector, consisting of one or more bytes of the feature vector, defines the aspect ratio of the pattern. In one embodiment, each byte of the feature vector representing the pixels contained in the character represents the relative ratio of black pixels to total pixels in a specific area of the character; other functions relating input matrix and output feature vector information can be used. In one embodiment of this invention, those areas of the character which are defined by the feature vector together cover the entire character, providing a feature vector describing what might loosely be thought as a "blurred" version of the pattern.

Automatic Feature Design for Optical Character Recognition Using an Evolutionary Search Procedure

Abstract: An automatic evolutionary search is applied to the problem of feature extraction in an OCR application. A performance measure based on feature independence is used to generate features which do not appear to suffer from peaking effects [17]. Features are extracted from a training set of 30 600 machine printed 34 class alphanumeric characters derived from British mail. Classification results on the training set and a test set of 10 200 characters are reported for an increasing number of features. A 1.01 percent forced decision error rate is obtained on the test data using 316 features. The hardware implementation should be cheap and fast to operate. The performance compares favorably with current low cost OCR page readers.

Feature Extraction And Pattern Classification In Space - Spatial Frequency Domain

Abstract: A novel feature extraction method, useful for 2-D shape description, is proposed. It is based on an optimal representation of a 1-D signal in space - spatial frequency domain, the Wigner distribution. For shape clasification, one of the many 1-D representations of the 2-D contours is employed. Boundary features, or shape descriptors, are obtained using sigular value decomposition of the Wigner distribution (WD). Properties of WD singular values are presented and shown to encode certain shape features such as the space-bandwidth product, the shape complexity in terms of number of components and their spacing, and the spatial frequency vs. the space dependence. The singular values of the boundary Wigner distri bution possess all the properties required of good shape descriptors. To illustrate the effectiveness of these descriptors in shape classification, a number of examples are presented. The proposed method is useful for robust classification of any 1-D patterns.

Techniques For Real-Time, 3D, Feature Extraction Using Range Information

Abstract: Recognition of three dimensional (3D) objects in low contrast scenes and in scenes where objects are partially occluded is a challenging problem for advanced 3D sensor based robotic systems. Laser ranging systems measure the surface depth directly and therefore avoid the computation required for construction of a depth map from multiple camera views. Since the physical level knowledge of the surfaces is directly available, the capability for real time object recognition in complex scenes is introduced. This paper discusses some elements of a real time 3D pattern recognition system for sensor based robot applications. The investigation includes a conceptual discussion of the sensor and techniques for feature extraction.

Connected Line Drawing Extraction from a Perspective View of a Polyhedron

Abstract: To extract line drawings with positional information from perspective veiws of three-dimensional objects is essential in image analysis and understanding. A new heuristic-search algorithm driven by a priori knowledge contained in a world model is presented which extracts a connected line drawing from a perspective view of a polyhedron. A main feature of our algorithm is that the search is concentrated on local areas centered at corners found with a corner finder. Therefore, the search time is significantly reduced and so are the positional errors in the extracted line drawing. An iterative process removes the false corners and lines and thus guarantees that our algorithm will work stably and reliably even in a noisy environment. Experimental results are presented.

Hybrid Optical/Digital Image Pattern Recognition: A Review

Abstract: The parallel processing, high-speed, compact system fabrication possibility, low power dissipation and size, plus weight advantages of optical processors have achieved great strides in recent years. The architectures, algorithms and system fabrication of hybrid pattern recognition processors are reviewed with attention and emphasis to recent results and to techniques appropriate for distortion-invariant multi-class pattern recognition applications.

Superiority of global features in classification and matching.

Abstract: Global superiority refers to the phenomenon that stimuli can be discriminated, classified, and matched faster using a global feature than a local feature. Global superiority has been shown by several authors with forms (global feature) composed of smaller forms (local features) as stimuli. In contrast, the local features used in the present study had about the same spatial extent as the global feature and each local feature was structurally relevant to the global feature. Nevertheless, global superiority was consistently observed in several classification and matching tasks. This result, together with the results of other researchers using different stimulus materials shows that global superiority is a pervasive phenomenon. Under brief exposure durations within the threshold range, local features could, however, be better discriminated than global features. This finding invalidates the perceptual explanation of global superiority that global features are faster extracted and, in this way, become available earlier than local features. Considering the magnitude of the observed differences between classification times and threshold durations of global and local features, it is argued that under normal viewing conditions global features can be better selected and utilized for the given task than local features, and that this advantage by far outweighs the faster feature extraction of local features.

Maximum likelihood classification of synthetic aperture radar imagery

Abstract: Classification of synthetic aperture radar (SAR) images has important applications in geology, agriculture, and the military. A statistical model for SAR images is reviewed and a maximum likelihood classification algorithm developed for the classification of agricultural fields based on the model. It is first assumed that the target feature information is known a priori. The performance of the algorithm is then evaluated in terms of the probability of incorrect classification. A technique is also presented to extract the needed feature information from a SAR image; then both the feature extraction and the maximum likelihood classification algorithms are tested on a SEASAT-A SAR image.

NASA image-based geological expert system development project for hyperspectral image analysis

Abstract: The NASA image-based geological expert system was applied to analyze remotely sensed hyperspectral image data. The major objective is for geologists to identify the earth surface mineral properties directly from the airborne and spaceborne imaging spectrometer data. With certain constraints, it is shown that the system can identify correctly different classes of mineral. It has the built-in learning paradigm to enhance the confidence factor of mineral identification. A very powerful natural language system was incorporated as the user-friendly front end, and the concurrent processing efficiency of the frame-based knowledge representation in the hypercube microsupercomputer simulation was tested.

Automatic Description of the Venation of Mosquito Wings from Digitized Images

Abstract: -Methods are described for the automatic measurement of morphological features of mosquito wings (wing outline and venation) from TV images of specimens mounted on microscope slides. The digitized images were preprocessed to locate the "ridge points" in the gray-scale image surface and the digitized images were thinned so that veins could be represented by curves only a single pixel wide. The image was then segmented into the wing outline and the 10 longitudinal wing veins. Polynomial functions were fitted to each vein. The coordinates of the end points of each vein and the coefficients of the polynomials could then be used as descriptors of each wing for subsequent multivariate analyses. Comparisons are made with other algorithms and the results are shown. [Feature extraction; orthogonal polynomials; image analysis; tracing.] Now that inexpensive hardware is available for the attachment of TV cameras and other imaging devices to computers (even microcomputers), there is an immediate interest in practical applications of image analysis. Most applications in biology have been related to the field of medicine, but there are many potential applications in systematics and areas of population biology. While the methods described in the present paper were developed specifically for the veins and outlines of mosquito wings, the general approach has applicability to many other types of images in which the features of interest consist of a set of lines. Mosquito wings are especially challenging because the veins vary in density and are partly obscured by scales, and are seen against a cluttered background. In this paper we describe a method for locating the veins of mosquito wings (using image-segmentation techniques) and fitting the coefficients of polynomials to each vein. The coefficients may be difficult to interpret biologically, but they are convenient descriptors for subsequent multivariate analyses. They also permit one to reconstruct the image of the veins from the coefficients (the importance of which was discussed by Rohlf and Archie, 1984). Our procedures are: (1) digitizing an image of a mosquito wing (mounted on a microscope slide) using a TV camera and digitizer; (2) enhancing the original image to reduce the effects of imperfections in the image; (3) tracing the lines corresponding to the wing outline and the veins in order to obtain sets of coordinates; and (4) computing the coefficients of polynomial functions fitted to the coordinates to serve as the final descriptive features of the wings. The present report describes in detail steps 2 and 3 (preprocessing of the image) and step 4 (feature extraction). Many methods exist to describe the features of objects in an image (Pratt, 1978; Rohlf and Ferson, 1983). Methods include Fourier descriptors (Kuhl, 1982; Chellappa and Bajdajian, 1984; Rohlf and Archie, 1984), moments (Gonzalez and Wintz, 1977), and topological descriptors (Gonzalez and Wintz, 1977). This study is an extension of that of Rohlf and Archie (1984), who considered only outline shape of mosquito wings and used coordinates produced by tracing the outline of a wing manually with a coordinate digitizer. Ferson et al. (1985) reported on methods for automatically obtaining the outline trace of mussel shells (which have images with distinct edges).

Image Processing System To Analyze Droplet Distributions In Sprays

Abstract: The General Motors Research Laboratories has developed an image processing system that automatically analyzes the size distributions in fuel spray video images. Images are generated by using pulsed laser light to freeze droplet motion in the spray sample volume under study. This coherent illumina-tion source produces images that contain droplet diffraction patterns representing the droplet's degree of focus. Thousands of images are recorded per sample volume to get an ensemble average of the distribution at that spray location. After image acquisition the recorded video frames are replayed and analyzed under computer control. The analysis is performed by extracting feature data describing droplet diffraction patterns in the images. This allows the system to select droplets from image anomalies and measure only those droplets con-sidered in focus. The system was designed to analyze sprays from a variety of environments. Currently these are an ambient spray chamber, a high pressure, high temperature spray facility, and a running engine. Unique features of the system are the totally automated analysis and droplet feature measurement from the gray scale image. Also, it can distinguish nonspherical anomalies from droplets, which allows sizing of droplets near the spray nozzle. This paper describes the feature extraction and image restoration algorithms used in the system. Preliminary performance data are also given for two experiments. One experiment gives a comparison between manual and automatic measurements of a synthesized distribution. The second experiment compares measurements of a real spray distribution using current methods and using the automatic system.

Image Segmentation Algorithms

Abstract: The paper describes a sequence of algorithms used to perform segmentation of aerial images of natural terrain for the purpose of extracting features pertinent to cartographic applications. Topics include image filtering, labeling, automated editing and refinement of the segmentation within a resolution pyramid. These techniques are considered to be preprocessing activities which will, in general, require some editing by trained cartographers. The objective of this work is to minimize the tedium of feature extraction using algorithms that do not require excessive computational overhead.

Two-dimensional outlines

Abstract: An image-analytic technique that automatically determines the outlines for simple silhouettes is described. These outlines are numerically characterized using elliptic Fourier decomposition. Since the resulting Fourier coefficients can be normalized so that they are in- variant to changes in magnification and rotation as well as other such transformations of the original silhouette, they quantify shape per se, and can be used as variables in multivariate analysis of form. In an application, discriminant analysis distinguishes between electrophoret- ically distinct populations of the mussel Mytilus edulis. While the technique was able to dem- onstrate an association between genotype and morphdlogy, it was not able to identify reliably the population from which each specimen was collected. (Automatic identification; pattern recognition; feature extraction; elliptic Fourier description; Mytilus edulis; shape analysis; species discrimination.) For many years there has been interest in the possibility of using the computer to view directly, via a television camera and a digitizer, a series of biological speci- mens, determine the most informative characters, and then construct a rule that could be applied by the computer to sep- arate automatically a collection of speci- mens into two or more taxonomic groups. Once a suite of characters has been record- ed, there are many methods for determin- ing rules for identifying specimens to minimize the expected number of misclas- sifications. Many statistical techniques are described, for example, in the general text by Hand (1981). The main difficulties for automatic recognition by computer are those of image segmentation (isolating the image of the desired object from its back- ground) and feature extraction (the defi- nition and actual measurement of a suite of characters from the image). Rohlf and Ferson (1983) gave a general review of these problems. In the present paper, we report the successful practical application of these techniques to a set of biological images. Shells of the mussel Mytilus edulis are easily distinguished from image back- ground, and their form is simple enough to be described by fitting mathematical functions with few parameters.

Smart visual sensors for real-time image processing and pattern recognition based upon human visual system characteristics (matrix inversion, subtractive inhibition, multiple spatial filtering)

Abstract: To design lightweight smart sensor systems which are capable of outputting motion-invariant features useful for automatic pattern recognition systems, we must turn to the simultaneous image processing and feature extraction capability of the human visual system (HVS) to enable operation in real time, on a mobile platform, and in a "natural environment." This dissertation studies the smart sensor problem, develops new concepts, supported by simulation, and discusses research areas for further exploration. An n('2) parallel data throughput architecture implemented through a hardwired algotecture which accomplishes, without computation, an equivalent logarithmic coordinate mapping is presented. The algotecture mapping provides, at the sensor level, the ability to change scales and rotations in the input plane to shifts in the algotecture mapped space. The resulting invariant leading edge is shown to possess an intensity preserving property for arbitrary variations of image size. The sensitivity of the algotecture to center mismatch is discussed in terms of the difference between coordinate and functional transformation methods. A mathematical link between the lateral subtractive inhibition (SI) and multiple spatial filtering (MSF) mechanisms of the HVS coexist and function simultaneously. The feature extraction filter in visual neurophysiology, known as the novelty filter, is identified to be the first feedback term of an iterative expansion of the sensory mapping point spread function. The use of the algotecture space combined with the image plane MSF approach is explored for detection and classification using template crosscorrelation methods of recognition. The concept of using a three spatial frequency band model, based upon HVS physiological and psychophysical data, for an intra-class, and inter-class, and a membership identification classification scheme is introduced. This concept is extended to represent the feature vector entries for, not only each spatial frequency band, but for each image view angle in the recognition library. A new algorithm for implementing an intelligent enormous matrix inversion is proposed. Such an inverse problem exists in the solution of the negative feedback equation for SI and has a form that lends itself easily to parallel processing. The solution provides for a means to solve the inversion even though one of the partitioned submatrices is singular. Procedures are given to construct a partition tree which is analogous to quadtree partitioning methods in image processing. Finally a simple matrix inversion example is worked out to demonstrate how the algorithm works.

Automated Rain-Rate Classification of Satellite Images Using Statistical Pattern Recognition

Abstract: This paper describes an automated procedure to determine rain rates in visible and infrared satellite images by means of statistical pattern recognition. Using brightness and textural features extracted from the images, the procedure classifies 8 km X 8 km windows of data into one of three classes of rain rate: none, light, and heavy. The training process utilizes both weather radar and cloud-development information derived from image sequences. Images from three different days were tested and classification accuracies of 70 percent or better were obtained. An automated scheme of this type has the potential to greatly speed the process of producing an estimate of rainfall from satellite imagery with little compromise in overall accuracy.