CNN Universal Machine Research Articles

A real-time multitarget tracking system with robust multichannel CNN-UM algorithms

This paper introduces a tightly coupled topographic sensor-processor and digital signal processor (DSP) architecture for real-time visual multitarget tracking (MTT) applications. We define real-time visual MTT as the task of tracking targets contained in an input image flow at a sampling-rate that is higher than the speed of the fastest maneuvers that the targets make. We utilize a sensor-processor based on the cellular neural network universal machine architecture that permits the offloading of the main image processing tasks from the DSP and introduces opportunities for sensor adaptation based on the tracking performance feedback from the DSP. To achieve robustness, the image processing algorithms running on the sensor borrow ideas from biological systems: the input is processed in different parallel channels (spatial, spatio-temporal and temporal) and the interaction of these channels generates the measurements for the digital tracking algorithms. These algorithms (running on the DSP) are responsible for distance calculation, state estimation, data association and track maintenance. The performance of the proposed system is studied using actual hardware for different video flows containing rapidly moving maneuvering targets.

Real time monitoring of radiation instabilities in TOKAMAK machines via CNNs

In this paper a new strategy for real time detection of plasma instabilities, called MARFEs, is proposed through real-time image processing on plasma video sequences. These sequences are recorded by a vision system based on a charge coupled device (CCD) camera installed at Frascati Tokamak Upgrade (FTU). The strategy used to perform the task is based on a new family of nonlinear analog processors, digitally programmable, implemented into the so-called Cellular Neural Network Universal Machine (CNN-UM). The basic idea deals with a system capable to carry out safer nuclear fusion experiments, preventing the plant from excessive mechanical and thermal stress which occurs during plasma instability phenomena (i.e., disruptions). Experimental results, obtained on the FTU machine, are reckoned fully satisfactory.

Pixel‐level snakes on the CNNUM: algorithm design, on‐chip implementation and applications

AbstractIn this paper, a new algorithm for the cellular active contour technique called pixel‐level snakes is proposed. The motivation is twofold: on the one hand, a higher efficiency and flexibility in the contour evolution towards the boundaries of interest are pursued. On the other hand, a higher performance and suitability for its hardware implementation onto a cellular neural network (CNN) chip‐set architecture are also required. Based on the analysis of previous schemes the contour evolution is improved and a new approach to manage the topological transformations is incorporated. Furthermore, new capabilities in the contour guiding are introduced by the incorporation of inflating/deflating terms based on the balloon forces for the parametric active contours. The entire algorithm has been implemented on a CNN universal machine (CNNUM) chip set architecture for which the results of the time performance measurements are also given. To illustrate the validity and efficiency of the new scheme several examples are discussed including real applications from medical imaging. Copyright © 2005 John Wiley & Sons, Ltd.

Toward CNN Chip-Specific Robustness

The promising potential of cellular neural networks (CNN) has resulted in the development of several template design methods. The CNN universal machine (CNN-UM), a programmable CNN, has made it possible to create image-processing algorithms that run on this platform. However, very large-scale integration implementations of CNN-UMs presented parameter deviations that do not occur on ideal CNN structures. Consequently, new design methods were developed aiming at more robust templates. Although these new templates were indeed more robust, erroneous behavior can still be observed. An alternative for chip-independent robustness is chip-specific optimization, where the template is targeted to an individual chip. This paper describes a solution proposal in this sense to automatically tune templates in order to make the chip react as an ideal CNN structure. The approach uses measurements of actual CNN-UM chips as part of the cost function for a global optimization method to find an optimal template given an initial approximation. Further improvements are achieved by generating chip-specific robust templates by doing a search for the best template among the optimal ones. The tuned templates are therefore customized versions that are expected to be much less sensitive to imperfections on the operation of CNN-UM chips. Results are presented for the binary and grayscale cases, including the case of grayscale output. It is expected that as this technique matures, it will give CNN-UM chips enough reliability to compete with digital systems in terms of robustness in addition to advantages of speed.

Implementation of a Pixel-Level Snake Algorithm on a CNNUM-Based Chip Set Architecture

In this paper, an on-chip implementation of the active contour technique called pixel-level snakes is proposed. This is based on an optimized cellular neural network (CNN) algorithm with capabilities to support changes in the contour topology. The entire algorithm has been implemented on a 64/spl times/64 CNN universal machine chip-set architecture for which the results of the time performance measurements are given. To illustrate the validity and capabilities of the proposed implementation some on-chip experiments are also included.

RECEPTIVE FIELD ATLAS AND RELATED CNN MODELS

In this paper we demonstrate the potential of the cellular nonlinear/neural network paradigm (CNN) that of the analogic cellular computer architecture (called CNN Universal Machine — CNN-UM) in modeling different parts and aspects of the nervous system. The structure of the living sensory systems and the CNN share a lot of features in common: local interconnections ("receptive field architecture"), nonlinear and delayed synapses for the processing tasks, the potentiality of feedback and using the advantages of both the analog and logic signal-processing mode. The results of more than ten years of cooperative work of many engineers and neurobiologists have been collected in an atlas: what we present here is a kind of selection from these studies emphasizing the flexibility of the CNN computing: visual, tactile and auditory modalities are concerned.

CNN universal machine as classificaton platform: an art-like clustering algorithm.

Fast and robust classification of feature vectors is a crucial task in a number of real-time systems. A cellular neural/nonlinear network universal machine (CNN-UM) can be very efficient as a feature detector. The next step is to post-process the results for object recognition. This paper shows how a robust classification scheme based on adaptive resonance theory (ART) can be mapped to the CNN-UM. Moreover, this mapping is general enough to include different types of feed-forward neural networks. The designed analogic CNN algorithm is capable of classifying the extracted feature vectors keeping the advantages of the ART networks, such as robust, plastic and fault-tolerant behaviors. An analogic algorithm is presented for unsupervised classification with tunable sensitivity and automatic new class creation. The algorithm is extended for supervised classification. The presented binary feature vector classification is implemented on the existing standard CNN-UM chips for fast classification. The experimental evaluation shows promising performance after 100% accuracy on the training set.

SUDDEN GLOBAL SPATIAL-TEMPORAL CHANGE DETECTION AND ITS APPLICATIONS

We are watching the news on TV: the change of the background tells us when a new story begins. A glance at the clock and we can clearly see what time it is. These are special spatial-temporal episodes caused by ballistic eye movements and sudden optical changes. In this paper we give a useful definition for generalized sudden global change events, present its main properties and give an algorithm to recognize them in any video-flow. The proposed algorithm is implemented on a standard Cellular Nonlinear Network Universal Machine (CNN-UM). The processing time of the detection is roughly one millisecond on the ACE4k CNN-UM chip in the Aladdin environment. Therefore it can serve as a common function in any online, real-time spatial-temporal algorithm. The sudden global change event can be used to define the instant when the time-dependent part of that algorithm, e.g., homotopy, has to be initialized. The first steps of the multimedia processing, such as searching for key frames — for compression, browsing or indexing — can be helped by the proposed algorithm as a robust video shot change detection method.

Analysis of multidimensional neural activity via CNN-UM.

In this paper we show that the Cellular Nonlinear Network Universal Machine (CNN-UM) is an excellent tool for analyzing time series of multidimensional binary signals. The developed algorithm is dedicated to process electrophysiological multi-neuron recordings: our aim is to find specific multidimensional activity patterns, which may reflect higher order functional cell-assemblies. The analysis consists of two parts: first, the occurrences of different patterns are counted, then the statistical significance of each occurrence frequency is calculated separately.

PDE-BASED HISTOGRAM MODIFICATION WITH EMBEDDED MORPHOLOGICAL PROCESSING OF THE LEVEL-SETS

This paper describes parallel histogram modification techniques with embedded morphological preprocessing methods within the CNN-UM framework. The procedure is formulated in terms of nonlinear partial differential equations (PDE) and approximated through finite differences in space, resulting in coupled nonlinear ordinary differential equations (ODE). The I/O mapping of the system (containing both local and global couplings) can be calculated by a complex analogic (analog and logic) algorithm executed on a stored program nonlinear array processor, called the cellular nonlinear network universal machine (CNN-UM3). We describe and illustrate how the implementation of the algorithm results in an adaptive multi-thresholding scheme when histogram modification is combined with embedded morphological processing at a finite (low) number of gray-scale levels. This has obvious advantages if the further processing steps are segmentation and/or recognition. Experimental results processing real-life and echocardiography images are measured on different hardware/software platforms, including a 64 × 64 CNN-UM chip (ACE4k6,17).

THE CNN UNIVERSAL MACHINE: 10 YEARS LATER

In 1992, a new spatial temporal computing idea had been proposed, the CNN Universal Machine. It turned out that a new paradigm in computing on image flows, a Universal Machine on Flows, has ignited the intellect of hundreds of researchers. Today, visual microprocessors based on this idea can perform about TeraOPS computing power and 10 000 frames per second. In this paper, after a brief description of the history of the invention, architectural advances, physical implementation, algorithmic developments, as well as the biology relevance, theoretical aspects, mission critical applications, and new directions are reviewed.

COMPUTATIONAL AND COMPUTER COMPLEXITY OF ANALOGIC CELLULAR WAVE COMPUTERS

The CNN Universal Machine is generalized as the latest step in computational architectures: a Universal Machine on Flows. Computational complexity and computer complexity issues are studied in different architectural settings. Three mathematical machines are considered: the universal machine on integers (UMZ), the universal machine on reals (UMR) and the universal machine on flows (UMF). The three machines induce different kinds of computational difficulties: combinatorial, algebraic, and dynamic, respectively. After a broader overview on computational complexity issues, it is shown, following the reasoning related the UMR, that in many cases the size is not the most important parameter related to computational complexity. Emerging new computing and computer architectures as well as their physical implementation suggest a new look on computational and computer complexities. The new analog-and-logic (analogic) cellular array computer paradigm, based on the CNN Universal Machine, and its physical implementation in CMOS and optical technologies, proves experimentally the relevance of the role of accuracy and problem parameter in computational complexity. We introduce also the rigorous definition of computational complexity for UMF and prove an NP class of problems. It is also shown that choosing the spatial temporal elementary instructions, as well as taking into account the area and power dissipation, these choices inherently influence computational complexity and computer complexity, respectively. Comments related to relevance to biology of the UMF are presented in relation to complexity theory. It is shown that algorithms using spatial-temporal continuous elementary instructions (α-recursive functions) represent not only a new world in computing, but also, a more general type of logic inference.

Automatic detection and tracking of moving image target with CNN‐UM via target probability fusion of multiple features

AbstractA high speed target detection and tracking algorithm for a CNN‐UM chip is presented in this paper. The target confidence value is computed based on the fusion of target existence probabilities of features using products of weighted sums. The target decision is done with such a confidence value and target initiation is done through the temporal accumulation of the confidence. The probability of the target existence for each feature is created in the region of influence depending on the reliability and the strength of the feature. By virtue of the analogic parallel processing structure of the CNN‐UM (Roska T, Chua LO. The CNN universal machine: an analogic array computer. IEEE Trans. Circuits Systems II 1993; CAS‐40: 163–173), real time tracking can be achieved with presently available technologies with the speed of several kilo‐frames per second. Due to the utilization of multiple features of target, robust target detection is possible via the proposed algorithm. On‐chip experiments of the proposed target‐tracking algorithm have been done and properties of the proposed approach are disclosed through the various experiments. Copyright © 2003 John Wiley & Sons, Ltd.

Configurable multilayer CNN-UM emulator on FPGA

A new emulated digital multilayer cellular neural network (universal machine (CNN-UM) chip architecture called Falcon has been developed. In this brief, the main steps of the field-programmable gate array (FPGA) implementation are introduced. The main results are as follows. The CNN-UM architecture emulated on Xilinx Virtex series FPGA, three-dimensional nonlinear spatio-temporal dynamics can be implemented on this architecture. The critical parameters of the implementation in a single-layer configuration are 55 million cell update/s/processor core, or, equivalently 1 giga-operation per second (GOPS) computing performance. In the face of the high performance, the power requirements of the architecture are relatively low only /spl sim/3 W per processor core. Using reconfigurable devices to implement emulated digital architectures provides more flexibility compared to the custom very large-scale integration designs because different Falcon architectures can be used on the same FPGA device.

Object-oriented image analysis using the cnn universal machine: new analogic cnn algorithms for motion compensation, image synthesis, and consistency observation

Image-analysis algorithms are of great interest in the context of object-oriented coding schemes. With reference to the utilization of the cellular neural network (CNN) universal machine for object-oriented image analysis, this paper presents new analogic CNN algorithms for obtaining motion compensation, image synthesis, and consistency observation. Along with the already developed segmentation and object labeling technique, the proposed method represents a framework for implementing CNN-based real-time image analysis. Simulation results, carried out for different video sequences, confirm the validity of the approach developed herein.

Analog addition/subtraction on the cnn-um chip with short-time superimposition of input signals

The cellular-neural-network universal machine (CNN-UM) technique which performs analog addition/subtraction between image frames has been developed. The equivalent circuit of the uncoupled CNN without self feedback is reduced to a simple RC circuit. If two inputs are presented to the circuit one after another during a very short time period, the voltages that are proportional to their input signals are superimposed on the state capacitor. The output of such superimposition is a reduced version of the addition/subtraction between the two signals. Simple amplification of the output can recover the actual output. The characteristics of analog addition/subtraction with the proposed algorithm are shown via on-chip experiment. Application of the proposed algorithm to moving target detection is also presented.

Toward visual microprocessors

This paper outlines motivations and models underlying the design of visual microprocessors based on the cellular neural network universal machine. We also overview the state of the art regarding the realization of these microprocessors in the form of very large-scale integration chips. Examples corresponding to measurements realized on these chips are enclosed for illustration purposes.

Adaptive image sensing and enhancement using the cellular neural network universal machine

AbstractAs an attempt to introduce interactive, content‐dependent adaptive (ICDA) image processing, a simple but powerful active image sensing and two image enhancement methods are introduced via adaptive CNN‐UM sensor‐computers. Thus, the method ICDA can be used for adaptive control of image sensing and for subsequent on‐line or off‐line image enhancement as well. The algorithms use both intensity and contrast content. The image sensing technology can be realized with the current CNN‐UM chip. Our first image enhancement method is also executable on this chip, but it is more suitable for the adaptive cellular neural network universal machine (ACNN‐UM) architecture. Some results of simulator and chip experiments and an adaptive extended cell are presented. Our second, dynamical image enhancement method is planned to be executable on a multi‐layer, complex cell CNN architecture. In (Proceedings of the 6th IEEE International Workshop on Cellular Neural Networks and their Applications (CNNA‐2000) Catania, 2000; 213–217) 3‐layer architecture is described which is capable of realizing the main part of the second enhancement method.The main issues of our paper are as follows: the novel outlook of the ICDA framework, three new methods for two key application areas of CNN‐UM, the notion of ‘regional’ adaptive computing, the novelty of application of equilibrium‐computing in the third method. However, the key novelty of our work is not just a new method and a new realization: by combining sensing and computing, dynamically and pixelwise, a new quality becomes practical. Copyright © 2002 John Wiley & Sons, Ltd.

Cellular neural networks for real-time DNA microarray analysis.

A new methodology for real-time processing of DNA chip images is proposed. The idea developed here is to use the cellular neural network (CNN) array to analyze the DNA microarray. A CNN is an analog dynamic processor array that reflects this property: the processing elements interact directly within a finite local neighborhood. Due to its architecture, a two-dimensional CNN array is widely used to solve image processing and pattern recognition problems; moreover, the parallelism characteristic of this structure allows one to perform the most computationally expensive image analysis tasks three orders of magnitude faster than a classical CPU-based computer. This approach, thanks to the supercomputing capabilities of the CNN architecture, makes the whole DNA chip methodology fully parallel and also makes the processing phase, until now very time consuming, a real-time step. We discuss the results of testing an algorithm based on the CNN universal machine (CNN-UM) that has been designed to classify the image data. The algorithm is implemented in an analogic (analog and logic) microprocessor.

A CNN framework for modeling parallel processing in a mammalian retina

AbstractWe present here a simple multi‐layer cellular neural/non‐linear network (CNN) model of the mammalian retina, capable of implementation on CNN Universal Machine (CNN‐UM) chips. The basis of the model is a simple multi‐layer cellular neural/non‐linear Network (IEEE Trans. Circuits Systems 1988; 35:1257; IEEE Trans. Circuits Systems 1993; 40:147). The characterization of the elements in the CNN model is based on anatomical and physiological studies performed in the rabbit retina. The living mammalian retina represents the visual world in a set of about a dozen different ‘feature detecting’ parallel representations (Nature 2001; 410:583–587). Our CNN model is capable of reproducing qualitatively the same full set of space–time patterns as the living retina in response to a flashed square. The modelling framework can then be used to predict the set of retinal responses to more complex patterns and is also applicable to studies of the other biological sensory systems. The work represents a major step forward in the complexity and programmability of retinal models. Copyright © 2002 John Wiley & Sons, Ltd.