Marian S. Stachowicz

Image segmentation and classification using color features

Color provides a wealth of information for interpretation of image content. The increased availability of affordable digital color cameras has created the opportunity to explore the degree to which color is useful in computer vision. This paper shows that a system for image segmentation and classification can be created using color as the primary feature. This system is comprised of two phases: segmentation and classification. In the first step, an image is searched with a blob detection algorithm to determine the location of any possible foreground elements. These areas are extracted from the image to be used in the next step. Classification is done using a set of eight color features that are optimally selected for each database. The appropriate feature vector is created for each foreground area removed from the original image. The vector is then compared to a preconstructed database to be identified. For this paper USA postage stamps on envelopes were used as the test cases.

Architecture for event-driven intelligent fuzzy controller

The authors describe a model for synchronous finite state machines based on fuzzy logic. Such finite state machines can be used to build both event-driven time-varying rule-based systems and the control unit section of a fuzzy logic computer. The architecture of a pipelined intelligent fuzzy controller is presented, and the linguistic model is represented by an overall fuzzy relation stored in a single rule memory. A VLSI integrated circuit implementation of the fuzzy controller is suggested. At a clock rate of 30 MHz, the controller can perform 3 M fuzzy logical inferences per second (FLIPS) on multidimensional fuzzy data.<<ETX>>

Fuzzy-neural control of film forming processes

This paper presents a new technique for thickness control in film forming processes. The technique utilizes fuzzy logic and neural network methods to achieve a MIMO controller that accounts for strong interactions between the thickness in different lanes and for the large time delay between the actuators and the sensor. Experimental results demonstrate the performance of the controller.

Parallel architecture for a multiple-input fuzzy logic controller

A hardware accelerator that performs fuzzy learning, fuzzy inference, and defuzzification strategy computations is presented. The hardware is based on two-valued logic. A universal space of 25 elements with five levels each is supported. To achieve a high processing rate for real-time applications, the basic units of the accelerator are connected in a four-level pipeline. The accelerator can receive two parallel fuzzy data as inputs. A flag will be set if the fuzzy model R(u,w), constructed in a learning process, exhibits the property as follows: for all (u,w) belonging to the set UXW, R(u,w) equals 1. At a clock rate of 20 MHz, the accelerator can perform more than 1,400,000 fuzzy logic inferences per second on multi- dimensional fuzzy data.

A Hybrid Framework for Servo-Actuated Systems Fault Diagnosis

A hybrid fault diagnosis method is proposed in this paper which is based on analytical and fuzzy logic theory. Analytical redundancy is employed by using statistical analysis. Fuzzy logic is then used to maximize the signal- to-threshold ratio of the residual and to detect different faults. The method was successfully demonstrated experimentally on hydraulic actuated system test rig. Real data and simulation results have shown that the sensitivity of the residual to the faults is maximized, while that to the unknown input is minimized. The decision of whether ‘a fault has occurred or not?’ is upgraded to ‘what is the severity of that fault?’ at the output. Simulation results show that fuzzy logic is more sensitive and informative regarding the fault condition, and less sensitive to uncertainties and disturbances.

Intelligent Diagnosis System

Auscultation of heart murmurs can detect various types of heart problems; however, this process is prone to human error because it involves a clinician evaluating and categorizing the heart sound via stethoscope. By implementing Soft Computing techniques to analyze the sound received from the stethoscope and classify the heart pathology, this research project introduces the intelligent diagnosis system that is not based on subjective evaluations. The sound signal from the stethoscope is transformed into a spectrogram (an image that shows the time-based analysis of the frequency components) formed by taking the Fourier transform over a small sliding window in time. The magnitudes of the resulting Fourier transforms are mapped to a color function in a density plot. Color Reduction and Color Feature extraction methods are applied to convert the density plot to a manageable image characteristic and reduce the colors of the density plot from sixteen million to eight: red, green, blue, cyan, magenta, yellow, white, and black. The density plot represented by the eight- color vector is then matched with a database to identify a possible pathology.

Hardware accelerator for linguistic data processing

A hardware accelerator that performs fuzzy learning, fuzzy inference, and defuzzification strategy computations is presented in this paper. The hardware is based on two-valued logic. A universal space of 25 elements with five levels each is supported. To achieve a high processing rate for real-time applications, the basic units of the accelerator are connected in a four-level pipeline. The accelerator can receive two parallel fuzzy data as inputs. At a clock rate of 20 MHz, the accelerator can perform 800,000 fuzzy logic inferences per second on multidimensional fuzzy data.