Navigation is a broad topic that has been receiving considerable attention from the mobile robotic community over the years. In order to execute autonomous driving in outdoor urban environments it is necessary to identify parts of the terrain that can be traversed and parts that should be avoided. This paper describes an analyses of terrain identification based on different visual information using a MLP artificial neural network and combining responses of many classifiers. Experimental tests using a vehicle and a video camera have been conducted in real scenarios to evaluate the proposed approach.

A Road Following Approach Using Artificial Neural Networks Combinations

Autonomous mobile robot navigation is a very relevant problem in robotics research. This paper proposes a vision-based autonomous navigation system using artificial neural networks (ANN) and finite state machines (FSM). In the first step, ANNs are used to process the image frames taken from the robot´s camera, classifying the space, resulting in navigable or non-navigable areas (image road segmentation). Then, the ANN output is processed and used by a FSM, which identifies the robot´s current state, and define which action the robot should take according to the processed image frame. Different experiments were performed in order to validate and evaluate this approach, using a small mobile robot with integrated camera, in a structured indoor environment. The integration of ANN vision-based algorithms and robot´s action control based on a FSM, as proposed in this paper, demonstrated to be a promising approach to autonomous mobile robot navigation.

Vision-Based Autonomous Navigation System Using ANN and FSM Control

Navigation is a broad topic that has been receiving considerable attention from the mobile robotic community. In order to execute a safe navigation on outdoors it is necessary to identify parts of the terrain that can be traversed by the robot and parts that should be avoided. This paper describes an analyses of an image-based terrain identification based on different visual information using a multi-layer perceptron neural network. Experimental tests using an outdoor robot and a video camera have been conducted in real scenarios to evaluate the proposed methods.

Path recognition for outdoor navigation using artificial neural networks: Case study

A context-sensitive framework for mobile terminals for assisting Type 2 diabetes patients

Using Red-Green-Blue triples to describe colors is common and often quite convenient. However, other color models may be more convenient or appropriate for many types of image analysis and processing. We will discuss conversion between RGB and YUV, YIQ, HSV, HLS, HSI color spaces. We create histograms of the distribution of the color components in those spaces and give examples of how to correct defects and make improvements in various color spaces.

With J: image processing 2: color spaces

F AST FOultIElt TII.ANSFOILMS make it possible to convert back and forth between image space and frequency space. It is amazing that it is possible to remove of some of the blur caused by motion or improper focus [5]. J offers a powerful add-on packageffl'w based on work of Frigo and Johnson [2], that makes it easy to combine fast Fourier transforms with other array processing. This column closely follows the discussion of motion blur in [4]. We will not discuss the mathematics of computing the Fourier transform. However, we remark that the Fourier transform of an argument array results in another array. While the input array is ordinarily a real array representing an image, any complex valued array is allowed. The result, typically a complex valued array, can be thought of as giving a real and imaginary part for each entry. Or, more important for applications, we can think of those complex entries as a magnitude and phase. Images of the magnitude essentially give diffraction patterns [1,3]. Thefl~o add-on package can be downloaded from wzow.jsoftware.com. We load the ffkw add-on package, create a small example and compute the. Fourier transform of that array as follows: require 'system\\packages\\fftw\\fftw' ]a=:</-i.4 0 i 1 1 0 0 1 1 0 0 0 1 0 0 0 0 fftw a 6 _2j2 _2 _2j_2 2j_2 0 0 _2j2 2 0 _2 0 2j2 _2j_2 0 0 The real and imaginary parts may be computed: <\"2 ] 9 ii o./ fftw a 6 2 2 _2 2 0 0 _2 2 0 _2 0 2 _2 0 0 0 2 0 _2 _2 0 0 2 0 0 0 0 2 _2 0 0 The magnitude and phase may be computed:

With J: Fast Fourier transforms and removing motion blur

Last summer I asked the Jforum for advice regarding the conversion of J text to a format suitable for placing J expressions and session logs into TeX word processing documents. There were a couple dozen follow-up posts to the Jforum and I received many private e-mail messages. There have been independent threads regarding TeX in comp.lang.apl too. I was quite surprised by the high level of interest and impressed by the number of sophisticated comments. I also enjoyed a few good laughs. This note offers some remarks on the Jforum discussion, what I eventually did, what the alternatives were, and what I might have done if I could have foreseen the future. While the core computation was simple escape sequence substitution, there also were subtle contexts for modifications when lines end and when box drawing characters appear. J's adverb facility helped with those situations and the use of J-defined function keys made the J to TeX conversion much easier.

With J for TeX

A T APL971 RAN A WORKSHOP on creating images withJ. The first exercise involved creating a Sierpinski Triangle. Ken ~ Iverson commented that I seem to start everything I do with the Sierpinski Triangle as an example. I don't think that is true, but it does seem like a good idea. Thus, my first Quote Quad column on usingJ focuses on ways to create Sierpinski Triangles. Of course, this is really an excuse to discuss a bunch of J in a way that results in some interesting images. Many of these ideas may be familiar, but readers are encouraged to look for new points of view. You might note the usefulness of inverse functions, gerunds and extended integers and may even discover new uses of base-Cliff Reiter's APL97 workshop three representations. Did you know you don't have to \"build\" Sierpinski's Triangle? You can decide the value of each pixel from the pixel coordinates. at the APL97 \"Chaos with Symmetry\" exhibit Future columns planned include dot-by-dot creation of one of the fi'actal types from the APL97 exhibit \"Chaos with Symme-try\" [1] and another takes a look at creating \"music\" with J. While I often create fractals in my work with J, my goal for the \"With]\" column is the illustration of J usage with fun examples from a variety of topics. 8 I will define what I mean by the Sierpinski Triangle by examples that approximate it. Purists will want to take some sort of limit of the approximation process we will explore. Of course, they won't see their result. We will be able to see our finite approximations. Sometimes these are discrete matrices. Other times, lists of lit pixels or line segments. We won't worry about whether the Sierpinski Triangle will \"open\" toward the upper right or some other direction; nor will we worry about whether it is equilateral or not. This list of 101 ways is not meant to be complete. There are at least two rather different ways appearing in [8]. Remarkable connections with the tower of Hanoi are described in [13] but won't be considered here. Sierpinski's original paper [12] described a curve that, roughly speaking, intersects itself at every point. Our last \"Way\" constructs approximations to that curve. We look at matrix-and pixel-based constructions first. We can readily create a matrix and juxtapose it above or beside itself. We say m, m gives m adjoined to …

101 ways to build a Sierpinski Triangle

Image processing includes techniques that are used to correct defects in images and to enhance the visibility of features of interest. In this note we will look at several different methods for smoothing images in order to remove specks of dust and artifacts from scanning. In future notes we plan to discuss other techniques for enhancing images in the spatial domain including the removal of orientation and lighting defects. Previous discussions [6,7] considered the removal of motion blur in frequency domain using Fast Fourier Transforms.

https://webbox.lafayette.edu/~reiterc/j/withj/withj12_im_proc01_pp.pdf

Cliff Reiter

Papers

With J: image processing 2: color spaces

With J: Fast Fourier transforms and removing motion blur

With J for TeX

101 ways to build a Sierpinski Triangle

With J: image processing 1: smoothing filters