A. Nabout

Redundant fuzzy rules exclusion by genetic algorithms

Abstract A genetic-algorithm-based method for exclusion of the potential redundant if-then fuzzy rules that have been extracted from numerical input-output data is proposed. The main idea is the input-space separation into activation rectangles, corresponding to certain output intervals. The generation of fuzzy rules and the membership functions are based on these activation rectangles and appropriate fuzzy rules inference mechanism is proposed. As the method usually produces too many rules, it is necessary to exclude the potential redundant if-then rules. The concept for varying the family of sensitivity parameters, defining the overlapping of the fuzzy regions is proposed. The genetic algorithms are used to resolve the following combinatorial optimization problem: the generation of families of sensitivity parameters. In this way the potential redundant if-then fuzzy rules are excluded. The method formalizes the synthesis of the fuzzy system and could be used for function approximation, classification and control purposes. An illustrative example for implementation of the method for traffic fuzzy control is given.

The topological contour closure requirement for object extraction from 2D-digital images

Object contour extraction from given 2D-digital images is an essential processing step for object extraction, recognition and analysis. The exact execution of this contour extraction, which is known in the literature as a segmentation problem is extremely important as subsequent steps or algorithms applied are highly dependent on this object contour extraction step. 2D-objects have always closed contours. Scanning the exhaustive number of publications on object contour extraction will disclose that this basic topological property has not been taken into consideration. It has not even been explicitly required from segmentation algorithms. The presently published image processing methods for object contour extraction possess the severe disadvantage that the extracted contours are not necessarily closed. They exhibit discontinuous contours or bifurcated segments. This is also the reason for the obligation to use consecutive contour restoration algorithms requiring intensive computation time. The paper sets the central condition on object contour extraction methods to be topologically preserving. Thus, extracted contours of 2D-objects should necessarily be closed without discontinuous contours or bifurcation. By applying graph theoretical methods, it can be shown that preserving the object contour closure requirements implies that the resulting extracted contour graph is a directional Euler line. Contour segmentation algorithms or inadequate reduction of image resolution that lead to contour graphs other than the directed Euler line, should therefore not be acceptable. In the paper, the necessary and sufficient conditions that should be satisfied by an algorithm to produce only closed contours are given. The global requirement of contour closure is reduced to a set of local conditions for contour extractions. With these conditions one can examine whether a segmentation algorithm will produce closed contours. Further, an algorithm that satisfies these conditions is introduced and practical segmentation results with guaranteed object contour closure from 2 D-object images are shown.<<ETX>>

Method for fuzzy rules extraction from numerical data

A method for extraction of fuzzy rules from numerical input-output data is proposed. The main idea consists in separation of the input space into activation rectangles, corresponding to different output intervals. The generation of fuzzy rules and membership functions is based on these activation rectangles, whereas an appropriate fuzzy rules inference mechanism is proposed. The method formalises the synthesis of the fuzzy system and could be used for function approximation, classification and control purposes. An illustrative example for implementation of the method for synthesis of traffic fuzzy control is given.

A novel closed contour extractor, principle and algorithm

The contour extraction is one of the most important steps in digital image processing. But the extracted contours with conventional algorithms are not always closed and continuous. They have also bifurcated segments. The needed computing time is intensive. In this paper, a novel algorithm called OCE (object-oriented contour extraction) has been proposed, which ensures a closure of all extracted contours, enables a parallel implementation in real-time and suppresses noise.

Plant species identification using fuzzy set theory

The automatic identification of plant species is a great challenge because their patterns are complex and uncertain. In this paper, the fuzzy set theory was applied to identify weed species. A membership function was established. The experiment has shown, that the average rate of correct identification has improved from 67% to greater than 82%.<<ETX>>

Object Shape Recognition Using Mexican Hat Wavelet Descriptors

In this paper we present the results of object recognition using Mexican Hat wavelet descriptors. These descriptors are derived from the continuous wavelet transformation using the Mexican Hat function as mother wavelet. To describe an object shape we use an angle function derived from the extracted contour polygon. The angle function is periodical and independent from the size of the object, its position or orientation. It depends only on the starting point of the contour. The contour extraction is based here on the object oriented contour extraction method (OCE). The polygon representation uses the curvature dependent contour approximation method (CDCA). The continuous wavelet transform (CWT) is used in order to derive a suitable number of wavelet descriptors (WD).

Measurement of ring-shaped and toothed mechanical parts through image processing

The automatic non-touch measurement of ring- shaped and toothed mechanical parts is an important task in the industry. This paper introduces two methods for image processing of such parts: the method using the dominant point detection and the method using the distance transform. The two methods are applied successfully to measure the characteristic quantities from practical scenes. The results are discussed and compared with each other.

Path Planning for Mobile Robots by Means of Approximate Routes

This paper describes a new method for the computation of paths for mobile robots in a known working area. The desired path connects a given starting position to a target position under avoidance of static obstacles. Path planning takes place off-line for the entire route and does not consider dynamic obstacles. This type of path planning can be used, for example, for fully automated storekeeping, in order to manoeuvre driverless vehicles flexibly and automatically through an entire storage depot. The method presented here uses auxiliary positions and computes several possible paths in the form of approximate partial routes. The computed paths vary in their overall length as well as the number and intensity of steering events. The results of this method are verified and discussed on the basis of a Matlab implementation.

Extended Method for Path Planning Using Auxiliary Corners

This paper presents an extended path planning method for obstacle avoidance for mobile robots using auxiliary positions. The desired path connects a given start position to a target position under avoidance of static obstacles. The corners of the rectangles that cover the obstacles serve as the auxiliary positions. The extended path planning uses the bounding rectangle with minimal area as a basis of simple bounding rectangle. Thus the accessible action area which is available for the path planning becomes larger and the path planning can be done more flexibly. The results of this method will also be discussed using MATLAB simulations and an implementation as part of the development of a system for fully automated storekeeping.

Wavelet descriptors for object recognition using Mexican hat function

Abstract In this paper, we propose an object recognition method for 2D objects using wavelet descriptors. The descriptors are derived from the continuous wavelet transform using the Mexican hat function as mother wavelet. In contrast to the other known methods we apply an angle function to describe object contours extracted as polygons. The contour extraction is based on the object oriented contour extraction method (OCE). The polygon representation is based on the curvature dependent contour approximation (CDCA). The continuous wavelet transform (CWT) is used in order to apply a suitable number of wavelet descriptors (WD), which are qualified to characterize the object shapes.