Research and development in agricultural robots are continuously increasing. However, dynamically changing agricultural environments provide adverse conditions to robotics operability. In order to perform the agricultural tasks safely and accurately, reliable landmarks from the surrounding environment need to be identified. In this work, deep learning is employed for accurate and fast detection of high-level features of vineyards, the vine trunks. More specifically, Faster regions-convolutional neural network (Faster R-CNN), You Only Look Once version 3 (YOLOv3) and YOLOv5 are tested for real-time vine trunk detection. The models are trained with an in-house dataset designed for the needs of this study, containing 1927 annotated vine trunks in 899 different images. Comparative results indicate YOLOv5 as the configuration that allows the faster and most accurate vine trunk detection, achieving an overall Average Precision of 73.2% in 29.6 ms. The high precision combined with the fast runtime performance prove that the YOLOv5 detector is suitable for real-time vine trunk detection executed by an autonomous harvesting robot.

The automation of agricultural production calls for accurate prediction of the harvest time. Our interest in particular here is in grape harvest time. Nevertheless, the latter prediction is not trivial also due to the scale of data involved. We propose a novel neural network architecture that processes whole histograms induced from digital images. A histogram is represented by an Intervals' Number (IN); hence, all-order data statistics are represented. In conclusion, the proposed "IN Neural Network", or INNN for short, emerges with the capacity of predicting an IN from past INs. We demonstrate a "proof-of-concept", preliminary application on a time series of digital images of grapes taken during their growth to maturity. Compared to a conventional Back Propagation Neural Network (BPNN), the results by INNN are superior not only in terms of prediction accuracy but also because the BPNN predicts only first-order data statistics, whereas the INNN predicts all-order data statistics.

Cyber-Physical Systems (CPSs) account for technical devices with both sensing and reasoning abilities. Our interest here is in classifying time series of sensory data by fuzzy logic in CPS applications. More specifically, a time series is represented by the distribution of its data samples using a novel Intervals' Number (IN) induction technique. We demonstrate the effectiveness of our techniques in a benchmark classification problem regarding ElectroEncephaloGraphy (EEG) signals. Mathematically sound extensions of the c-means as well as the kNN algorithms result in good classification performance comparatively to alternative methods from the literature. Potential future enhancements of the proposed techniques are discussed.

Watermarking regards the insertion of an "invisible" pattern to an image. It is known that watermarking can be pursued by modulating the coefficients of image moments, the latter are computationally extracted features; moreover, previous work has employed a Fuzzy Inference System (FIS) with triangular membership functions in order to compute each moment coefficient. This work proposes using irregular membership functions of multi-parametric Intervals' Numbers (INs) instead. Practical advantages of the proposed method are demonstrated comparatively in preliminary human-machine interaction applications of a Social Robot. The proposed techniques are presented as a case-specific study of modeling in Cyber-Physical Systems (CPSs) using Lattice Computing (LC) techniques.

This paper introduces a preliminary extension of the Fuzzy Cognitive Map (FCM) architecture based on Lattice Computing (LC) techniques namely Linguistic Fuzzy Cognitive Maps (LFCM). The proposed LFCM is able to handle large scale data in pattern classification applications. This enhancement is achieved by applying a novel data meta-representation, defined in a mathematical lattice, including several advantages. Based on this mechanism a new FCM classifier model is constructed and its performance is studied herein. Preliminary experimental results are both promising and competitive. Future work extensions are discussed.

Recent work has proposed an enhancement of Formal Concept Analysis (FCA) in a tunable, hybrid formal context including both numerical and nominal data [1]. This work introduces FCknn, that is a granular knn classifier based on hybrid concepts, whose effectiveness is demonstrated on benchmark datasets from the literature including both numerical and nominal data. Preliminary experimental results compare well with the results by alternative classifiers from the literature. Formal concepts are interpreted as descriptive decision-making knowledge (rules) induced from the data.

This paper compares two alternative feature data meta-representations using Intervals' Numbers (INs) in the context of the Minimum Distance Classifier (MDC) model. The first IN meta-representation employs one IN per feature vector, whereas the second IN meta-representation employs one IN per feature per class. Comparative classification experiments with the standard minimum distance classifier (MDC) on two benchmark classification problems, regarding face/facial expression recognition, demonstrate the superiority of the aforementioned second IN meta-representation. This superiority is attributed to an IN's capacity to represent discriminative, all-order data statistics in a population of features.

A fuzzy inference system (FIS) typically implements a function f : ℝN→ I, where the domain set R denotes the totally ordered set of real numbers, whereas the range set I may be either I = RM (i.e., FIS regressor) or T may be a set of labels (i.e., FIS classifier), etc. This study considers the complete lattice (F, ≤) of Type-1 Intervals' Numbers (INs), where an IN F can be interpreted as either a possibility distribution or a probability distribution. In particular, this study concerns the matching degree (or satisfaction degree, or firing degree) part of an FIS. Based on an inclusion measure function σ : F × F → [0, 1] we extend the traditional FIS design toward implementing a function f : FN→ I with the following advantages: 1) accommodation of granular inputs; 2) employment of sparse rules; and 3) introduction of tunable (global, rather than solely local) nonlinearities as explained in the manuscript. New theorems establish that an inclusion measure σ is widely (though implicitly) used by traditional FISs typically with trivial (i.e., point) input vectors. A preliminary industrial application demonstrates the advantages of our proposed schemes. Far-reaching extensions of FISs are also discussed.

An Intervals' Number (IN) is a mathematical object known to represent either a probability distribution or a possibility distribution. The space of INs has been studied during the last years. After summarizing some instrumental mathematical results, this work demonstrates comparatively novel schemes for tunable fuzzy rule interpolation and extrapolation. Extensions to Type-2 fuzzy sets are straightforward. Finally, this work demonstrates a preliminary application, regarding the reconstruction of partially occluded human facial expressions, based on a neural network that may predict a data distribution from other ones. Far reaching extensions of the proposed techniques are discussed.

This work introduces a novel methodology for human face recognition based on lattice computing kNN classification techniques applied on thermal infrared images. Novel feature extraction and knowledge-representation engage populations of orthogonal moments represented by intervals' numbers, or INs for short. Preliminary experimental results compare well with the results by alternative classifiers as well as with alternative feature extraction techniques from the literature. We point out the far-reaching potential of the proposed techniques to big data applications.

This paper proposes a fundamentally novel extension, namely, flrFAM, of the fuzzy ARTMAP (FAM) neural classifier for incremental real-time learning and generalization based on fuzzy lattice reasoning techniques. FAM is enhanced first by a parameter optimization training (sub)phase, and then by a capacity to process partially ordered (non)numeric data including information granules. The interest here focuses on intervals' numbers (INs) data, where an IN represents a distribution of data samples. We describe the proposed flrFAM classifier as a fuzzy neural network that can induce descriptive as well as flexible (i.e., tunable) decision-making knowledge (rules) from the data. We demonstrate the capacity of the flrFAM classifier for human facial expression recognition on benchmark datasets. The novel feature extraction as well as knowledge-representation is based on orthogonal moments. The reported experimental results compare well with the results by alternative classifiers from the literature. The far-reaching potential of fuzzy lattice reasoning in human-machine interaction applications is discussed.

This work presents a novel Lattice Computing (LC) approach for person identification based on biometric data including fingerprints. In particular, this work, first, engages conventional techniques for fingerprint image preprocessing towards inducing distributions of fingerprint minutiae, moreover an induced distribution is represented by an Intervals’ Number (IN). Second, it employs a kNN classifier in a metric lattice of INs. The effectiveness of the proposed approach is demonstrated comparatively by computational experiments using a software developed for the needs of this work. The far reaching potential of the proposed approach is discussed.

A novel distance measure between two intuitionistic fuzzy sets (IFSs) is proposed in this paper. The introduced measure formulates the information of each set in matrix structure, where matrix norms in conjunction with fuzzy implications can be applied to measure the distance between the IFSs. The advantage of this novel distance measure is its flexibility, which permits different fuzzy implications to be incorporated by extending its applicability to several applications where the most appropriate implication is used. Moreover, the proposed distance might be expressed equivalently by using either intuitionistic fuzzy sets or interval-valued fuzzy sets. Appropriate experimental configurations have taken place to compare the proposed distance measure with similar distance measures from the literature, by applying them to several pattern recognition problems. The results are very promising because the performance of the new distance measure outperforms the corresponding performance of well-known IFSs measures, by recognizing the patterns correctly and with high degree of confidence. © 2012 Wiley Periodicals, Inc.

We introduce novel (set- and lattice-theoretic) perspectives and tools for the analysis and design of fuzzy inference systems (FISs). We present an FIS, including both fuzzification and defuzzification, as a device for implementing a function f: RNrarr RM. The family of FIS functions has cardinality aleph2=2aleph1, where aleph1is the cardinality of the set R of real numbers. Hence the FIS family is much larger than polynomials, neural networks, etc.; furthermore a FIS has a capacity for local generalization. A formulation in the context of lattice theory allows us to define the set F*of fuzzy interval numbers (FINs), which includes both (fuzzy) numbers and intervals. We present a metric dKon F*, which can introduce tunable nonlinearities. FIS design based on dKhas advantages such as: an alleviation of the curse of dimensionality problem and a potential for improved computer memory utilization. We present a new FIS classifier, namely granular self-organizing map (grSOM), which we apply to an industrial fertilizer modeling application

Accurate and on-line decision-making is required by decision support systems including those ones used for environmental information management. This paper focuses on air quality assessment and demonstrates the added value of applying data mining techniques in operational decision-making. More specifically, the application of fuzzy lattice reasoning (FLR) classifier is investigated. An enhanced FLR learning algorithm is presented that employs a sigmoid valuation function for introducing tunable non-linearities. The FLR classifier is applied here beyond the unit-hypercube. The FLR with a sigmoid positive valuation function demonstrates an improved performance on a dataset from the region of Valencia, Spain regarding an environmental problem. Descriptive decision making knowledge (i.e. rules) for classification is also induced.

In this paper we introduce a fuzzy implication. The proposed fuzzy implication does not belong in one of the well known three general classes of fuzzy implications (S-implications, R-implications and QL-implications). Also we give an extended model of this fuzzy implication in intuitionistic fuzzy set and/or interval-valued fuzzy sets.

Industrial products are often output in batches at discrete times. A batch gives rise to distributions of measurements, one distribution per variable of interest. There may be a need for modeling to predict a distribution from other distributions. This work represents a distribution by a fuzzy interval number (FIN) interpreted as an information granule. Based on vector lattice theory it is shown that the lattice F+of positive FINs is a cone in a non-linearly tunable, metric, linear space. In conclusion, a multivariate granular autoregressive moving average (grARMA) model is proposed for predicting a distribution from other distributions. A recursive neural network implementation is shown. We report preliminary results regarding two real-world applications including, first, industrial fertilizer production and, second, environmental pollution monitoring along seashore in northern Greece. The far-reaching potential of novel techniques is discussed.

This work presents a fuzzy inference system (FIS) as a look-up table for function approximation by interpolation involving fuzzy interval numbers or FINs for short. It is shown that the cardinality of the set F of FINs equals N/sub 1/, that is the cardinality of the totally ordered lattice R of real numbers. Hence a FIS can implement in principle all N/sub 2/=2/sup N//sub 1/>N/sub 1/ real functions, moreover a FIS is endowed with a capacity for local generalization. It follows a unification of Mamdani- with Sugeno-type FIS. Based on lattice theory novel interpretations are introduced and, in addition, a tunable metric distance d/sub K/ between FINs is shown. Several of the proposed advantages are demonstrated experimentally.

An extension of the self-organizing map (SOM) is presented, namely granular SOM or grSOM for short, applicable beyond R/sup N/ to F/sup N/, where F denotes the set of fuzzy interval numbers (FINs). Rigorous analysis establishes that F is a metric mathematical lattice. A FIN is interpreted as a linguistic granule, which corresponds to a local probability distribution function. The grSOM can be used for structure identification in linguistic system modeling applications. Experimental results using the greedy grSOM algorithm compare favorably with the corresponding results by alternative algorithms from the literature in two benchmark classification problems; in addition, descriptive decision-making knowledge (fuzzy rules) is induced from the data.

This paper presents a prototype software platform, namely Platform for Adaptive and Reliable Evaluation of Students (PARES), for student testing and evaluation. The motivation has been to provide educators in the Greek higher education system a useful tool for both carrying out tests using minimal resources and for keeping the students actively involved during a semester. PARES has been developed in Java for launching it in the WWW. A pilot study using PARES at the Technological Educational Institution of Kavala is presented. A statistical evaluation of PARES is also described. A future extension, based on a synergy of machine-learning with state feedback-control techniques, for adaptive student learning is outlined.

This paper presents novel mathematical tools developed during a study of an industrial-yield prediction problem. The set F of fuzzy interval numbers, or FINs for short, is studied in the framework of lattice theory. A FIN is defined as a mapping to a metric lattice of generalized intervals, moreover it is shown analytically that the set F of FINs is a metric lattice. A FIN can be interpreted as a convex fuzzy set, moreover a statistical interpretation is proposed here. Algorithm CALFIN is presented for constructing a FIN from a population of samples. An underlying positive valuation function implies both a metric distance and an inclusion measure function in the set F of FINs. Substantial advantages, both theoretical and practical, are shown. Several examples illustrate geometrically on the plane both the utility and the effectiveness of novel tools. It is outlined comparatively how some of the proposed tools have been employed for improving prediction of sugar production from populations of measurements for Hellenic Sugar Industry, Greece.

Three novel fuzzy lattice neurocomputing (FLN) classifiers, namely FLN first fit (FLNff), FLN ordered tightest fit (FLNotf), and FLN selective fit (FLNsf), are introduced in this work. Learning is incremental, memory-based, data order dependent, and polynomial O(n/sup 3/) where n is the number of the training data. Convenient geometric interpretations on the plane illustrate the mechanics of the aforementioned FLN classifiers whose capacity is demonstrated in three benchmark classification problems. The classification results compare favorably with the results by alternative classification methods from the literature. In addition, an FLN classifier can both induce rules from the data and it can deal with numeric and/or nominal data including missing attribute values. An important experimental outcome of this work is that the computation of "smaller than maximal" lattice intervals can increase considerably the capacity for generalization.

This work, first, shows the inherent capacity of neural net /spl sigma/-FLNMAP for classification based on semantics and, second, it demonstrates the capacity of an ensemble of /spl sigma/-FLNMAP voters to improve classification accuracy. The /spl sigma/-FLNMAP neural network is presented here as a tool for function approximation. New definitions and useful properties extend coherently the applicability of /spl sigma/-FLNMAP. An ensemble of /spl sigma/-FLNMAP voters is treated as a statistical model whose parameters can be estimated from the training data. Noise canceling effects are discussed. Experimental results in four classification problems compare favorably with results by alternative classification methods from the literature.

An advantage of the /spl sigma/-FLNMAP neural network for classification is known to be the capacity to employ an underlying positive valuation function different than v(x) = x. This work demonstrates the effectiveness of a non-parametric probability distribution function used as an underlying positive valuation. Moreover a novel positive valuation function is introduced here analytically in a lattice of intervals. An ensemble of /spl sigma/-FLNMAP classifiers is employed as a majority voting model whose parameters 1) a weight w/sub i/, i = 1, ..., N for each constituent lattice, and 2) the number n/sub v/ of /spl sigma/-FLNMAP voters, are estimated optimally from the training data using a genetic algorithm. Similarities and differences are delineated with various ensemble methods from the literature. The genetic-fuzzy-neural-computing techniques presented in this work, despite their computational complexity, imply significant comparative improvements in three benchmark classification problems. It is explained how both the adaptive resonance theory (ART) and the min-max neural networks can benefit from the tools presented here.

A connectionist scheme, namely, /spl sigma/-Fuzzy Lattice Neurocomputing scheme or /spl sigma/-FLN for short, which has been introduced in the literature lately for clustering in a lattice data domain, is employed for computing clusters of directed graphs in a master-graph. New tools are presented and used, including a convenient inclusion measure function for clustering graphs. A directed graph is treated by /spl sigma/-FLN as a single datum in the mathematical lattice of subgraphs stemming from a master-graph. A series of experiments is detailed where the master-graph emanates from a thesaurus of spoken language synonyms. The words of the thesaurus are fed to /spl sigma/-FLN in order to compute clusters of semantically related words, namely hyperwords. The arithmetic parameters of /spl sigma/-FLN can be adjusted so as to calibrate the total number of hyperwords computed in a specific application. It is demonstrated how the employment of hyperwords implies a reduction, based on the a priori knowledge of semantics contained in the thesaurus, in the number of features to be used for document classification. In a series of comparative experiments for document classification, it appears that the proposed method favorably improves classification accuracy in problems involving longer documents, whereas performance deteriorates in problems involving short documents.

A basis for rigorous versatile learning is introduced theoretically, that is the framework of fuzzy lattices or FL-framework for short, which proposes a synergetic combination of fuzzy set theory and lattice theory. A fuzzy lattice emanates from a conventional mathematical lattice by fuzzifying the inclusion order relation. Learning in the FL-framework can be effected by handling families of intervals, where an interval is treated as a single entity/block the way explained here. Illustrations are provided in a lattice defined on the unit-hypercube where a lattice interval corresponds to a conventional hyperbox. A specific scheme for learning by clustering is presented, namely /spl sigma/-fuzzy lattice learning scheme or /spl sigma/-FLL (scheme) for short, inspired from adaptive resonance theory (ART). Learning by the /spl sigma/-FLL is driven by an inclusion measure /spl sigma/ of the corresponding Cartesian product to be introduced here. We delineate a comparison of the /spl sigma/-FLL scheme with various neural-fuzzy and other models. Applications are shown to one medical data set and two benchmark data sets, where /spl sigma/-FLL's capacity for treating efficiently real numbers as well as lattice-ordered symbols separately or jointly is demonstrated. Due to its efficiency and wide scope of applicability the /spl sigma/-FLL scheme emerges as a promising learning scheme.

Stapedotomy is a surgical procedure aimed at the treatment of hearing impairment due to otosclerosis. The treatment consists of drilling a hole through the stapes bone in the inner ear in order to insert a prosthesis. Safety precautions require knowledge of the nonmeasurable stapes thickness. The technical goal herein has been the design of high-level controls for an intelligent mechatronics drilling tool in order to enable the estimation of stapes thickness from measurable drilling data. The goal has been met by learning a map between drilling features, hence no model of the physical system has been necessary. Learning has been achieved as explained in this paper by a scheme, namely the d-sigma Fuzzy Lattice Neurocomputing (d sigma-FLN) scheme for classification, within the framework of fuzzy lattices. The successful application of the d sigma-FLN scheme is demonstrated in estimating the thickness of a stapes bone "on-line" using drilling data obtained experimentally in the laboratory.

The two-level fuzzy-lattice (2L-FL) learning scheme is introduced for application on an intelligent surgical (mechatronic) drill in the stapedotomy surgical procedure in the ear. The 2L-FL scheme learned from past cases to evaluate pointedly the thickness of a stapes bone using a force/torque pair of drilling profiles. Hence it is possible, in principle, to drill safely a hole through supple stapes by retracting automatically the drill upon bone breakthrough. The 2L-FL scheme was applied on two different partly ordered sets defined on a set of square matrices mapping features of two different data profiles to one another. Results are presented comparatively on experimental drilling data. The authors also discuss extension of their techniques to other surgical and industrial applications especially ones involving fusion of disparate sensory data where effective learning could compensate for the lack of accurate analytical models.

This paper proposes two hierarchical schemes for learning, one for clustering and the other for classification problems. Both schemes can be implemented on a fuzzy lattice neural network (FLNN) architecture, to be introduced herein. The corresponding two learning models draw on adaptive resonance theory (ART) and min-max neurocomputing principles but their application domain is a mathematical lattice. Therefore they can handle more general types of data in addition to N-dimensional vectors. The FLNN neural model stems from a cross-fertilization of lattice theory and fuzzy set theory. Hence a novel theoretical foundation is introduced in this paper, that is the framework of fuzzy lattices or FL-framework, based on the concepts fuzzy lattice and inclusion measure. Sufficient conditions for the existence of an inclusion measure in a mathematical lattice are shown. The performance of the two FLNN schemes, that is for clustering and for classification, compares quite well with other methods and it is demonstrated by examples on various data sets including several benchmark data sets.

The generalization capabilities, for learned subtle image pattern categories, of neural network and algorithmic classification techniques are described. Several neural network and algorithmic techniques have been applied to a set of feature vectors extracted from thermal infrared images which characterize the extent of whiplash injury. Thermography recently has been reported to have clinical utility in a multitude of neuromusculoskeletal disorders, particularly with soft tissue injuries such as whiplash in which there are few widely agreed upon diagnostic standards. The results of this research indicate that the backpropagation neural network produces the best classification results and provides significantly better generalization from a set of training patterns. Results of unsupervised classification of the data using clustering algorithms and the Adaptive Resonance Theory (ART2) neural network demonstrate the difficulties of learning and of generalization of patterns from such data.>

A report is presented on a comparison between neural network and algorithmic classification techniques applied to a specific thermography program: the analysis of image patterns which characterize the extent of whiplash injury. Thermography recently has been reported to have clinical utility in a multitude of neuromusculoskeletal disorders. Of particular import is the application of thermography to soft-tissue injuries in which there are few diagnostic gold standards. Likewise, neural networks have proven to be powerful pattern separators and classifiers in a variety of real-world problems. This is primarily due to their capabilities of pattern completion and nonlinear separability in feature space. Research results show that backpropagation neural networks can accurately classify up to 90% of the whiplash thermal images, while conventional algorithmic classifiers accurately classify only up to 75%. Slight differences were found in the classification accuracy of two commercially available backpropagation implementations

The authors report on the application of neurocomputing techniques to a specific thermography program: the analysis of image patterns which characterize the extent of whiplash injury. A specific sequence of preprocessing steps designed to optimize neurocomputer classification of biomedical image patterns with symmetry is described. Data acquisition methods, preprocessing steps, and experimental results are discussed. The results presented are derived from incorporating neurocomputing analysis as part of an overall image characterization process. It is demonstrated that the features do represent certain characteristics of the pathology present in the images.>

A family of neural networks and learning algorithms is introduced: the plastic self-adaptive learning machines (PSALM), together with a new interpretation of these neural networks as hyperpolyhedra in the N-dimensional Euclidean space. These networks self-adapt to a continually changing environment by properly changing the orientation of the faces of a hyperpolyhedron as well as its volume. The current structure of the hyperpolyhedron reflects the structure of the current outside world. The network optimally classifies its noise-distorted excitations into categories, after a competition between all possible categories. New categories can be created, and the old ones can be changed, or be forgotten if they are not used for a long time.>