Indoor Navigation Aid Systems for the Blind and Visually Impaired Based on Depth Sensors

This chapter focuses on the contributions made in the development of assistive technologies for the navigation of blind and visually impaired (BVI) individuals. A special interest is placed on vision-based systems that make use of image (RGB) and depth (D) information to assist their indoor navigation. Many commercial RGB-D cameras exist on the market, but for many years the Microsoft Kinect has been used as a tool for research in this field. Therefore, first-hand experience and advances on the use of Kinect for the development of an indoor navigation aid system for BVI individuals is presented. Limitations that can be encountered in building such a system are addressed at length. Finally, an overview of novel avenues of research in indoor navigation for BVI individuals such as integration of computer vision algorithms, deep learning for the classification of objects, and recent developments with stereo depth vision are discussed.

Technical Vision Model of the Visual Systems for Industry Application

The problems of highlighting the main informational aspects of images and creating their adequate models are discussed in the chapter. Vision systems can receive information about an object in different frequency ranges and in a form that is not accessible to the human visual system. Vision systems distort the information contained in the image. Therefore, to create effective image processing and transmission systems, it is necessary to formulate mathematical models of signals and interference. The chapter discusses the features of perception by the human visual system and the issues of harmonizing the technical characteristics of industrial systems for receiving and transmitting images. Methods and algorithms of pattern recognition are discussed. The problem of conjugation of the characteristics of the technical vision system with the consumer of information is considered.

Recognition System by Using Machine Vision Tools and Machine Learning Techniques for Mobile Robots

The systems based on image recognition play an important role in many cases where inspection methods are critical for industrial processes. Machine vision is required in industry for monitoring and detecting objects in real-time applications. Industrial robots are increasingly used in a variety of industries and applications such as manipulators and mobile robots. These devices are necessary for dangerous work conditions and tasks that humans cannot do; however, these industrial robots can also do some human activities such as transport material, select, and identify objects. In order to substitute human capabilities is no easy task. Nevertheless, the use of artificial intelligence has been replacing some human activities for increased productivity. By supporting machine learning technologies, this chapter presents a k nearest neighbors' algorithm for image classification of mobile robots to detect and recognize objects.

Optoelectronic Devices Fusion in Machine Vision Applications

This chapter presents the application of optoelectronic devices fusion as the base for those systems with non-linear behavior supported by artificial intelligence techniques, which require the use of information from various sensors for pattern recognition to produce an enhanced output. It also included a deep survey to define the state of the art in industrial applications following this tendency to identify and recognize the most used optoelectronic sensors, interconnectivity, raw data collection, data processing and interpretation, data fusion, intelligent decision algorithms, software and hardware instrumentation and control. Finally, it exemplifies how these technologies implemented in the industry can also be useful for other kinds of sector applications.

Statistical Characteristics of Optical Signals and Images in Machine Vision Systems

The chapter is devoted to the creation of a comprehensive approach to the physical and mathematical description of signals in optoelectronics in machine vision, taking into account the phenomena of interaction of optical radiation with system elements. A new methodology for the study of the statistical properties of input and output signals in optoelectronic systems is proposed, taking into account the availability of grouped statistical properties that do not obey the Poisson statistics. The basis is the joint use of wave and corpuscular description of signals in systems, stochastic flow theories, and elements of statistical detection theory. Information and energetic technology have been developed that integrates the theoretical justification of signal description under various observation conditions and decision-making methods.

Advances in Laser Scanners

One focus of present chapter is defined by the further development of the technical vision system (TVS). The TVS mainly contains two principal parts, the positioning laser (PL) and the scanning aperture (SA), which implement the optomechanical function of the dynamic triangulation. Previous versions of the TVS uses stepping motors to position the laser beam, which leads to a discrete field of view (FOV). Using stepping motors, inevitable dead zones arise, where 3D coordinates cannot be detected. One advance of this TVS is defined by the substitution of these discrete actuators by DC motors to eliminate dead zones and to perform a continuous laser scan in the TVS FOV. Previous versions of this TVS also uses a constant step response as closed-loop input. Thereby the chapter describes a new approach to position the TVS laser ray in the FOV, using a trapezoidal velocity profile as trajectory.

Detection of Landmarks by Mobile Autonomous Robots Based on Estimating the Color Parameters of the Surrounding Area

Remote detection of landmarks for navigation of mobile autonomous robots in the absence of GPS is carried out by low-power radars, ultrasonic and laser rangefinders, night vision devices, and also by video cameras. The aim of the chapter is to develop the method for landmarks detection using the color parameters of images. For this purpose, the optimal system of stochastic differential equations was synthesized according to the criterion of the generalized variance minimum, which allows to estimate the color intensity (red, green, blue) using a priori information and current measurements. The analysis of classical and nonparametric methods of landmark detection, as well as the method of optimal estimation of color parameters jumps is carried out. It is shown that high efficiency of landmark detection is achieved by nonparametric estimating the first Hilbert-Huang modes of decomposition of the color parameters distribution.

Informational Model of Optical Signals and Images in Machine Vision Systems

The rapid development and use of optical systems in measurement, navigation, and space technology to obtain accurate and detailed information about an object of observation is accompanied by the problems of transmitting high quality information through the optical system and processing of the obtained data. Integration of artificial intelligence systems in industry requires the creation and improvement of objective assessment and self-assessment systems. This is especially designed for automated recognition and classification systems. The problem of the object movement registration also contains some peculiarities such as background and main signal separation, noise influence and main objects selecting. Information about data quality is a set of properties that reflects the degree of suitability of specific information. It contains the data about objects and their relationship to achieve the goals of user requirements.

Reducing the Optical Noise of Machine Vision Optical Scanners for Landslide Monitoring

An application of landslide monitoring using optical scanner as vision system is presented. The method involves finding the position of non-coherent light sources located at strategic points susceptible to landslides. The position of the light source is monitored by measuring its coordinates using a scanner based on a 45° sloping surface cylindrical mirror. This chapter shows experiments of position light source monitoring in laboratory environment. This work also provides improvements for the optical scanner by using digital filter to smooth the opto-electronic signal captured from a real environment. The results of these experiments were satisfactory by implementing the moving average filter and median filter.

Methods of Reception and Signal Processing in Machine Vision Systems

The chapter covers development of mathematical model of signals in optoelectronic systems. The mathematical model can be used as a base for detection algorithm development for optical signal from objects. Analytical expressions for mean values and signal and noise components dispersion are cited. These expressions can be used for estimating efficiency of the offered algorithm by the criterion of detection probabilistic characteristics and criterion of signal/noise relation value. The possibility of signal detection characteristics improvement with low signal-to-noise ratio is shown. The method is proposed for detection of moving objects and combines correlation and threshold methods, as well as optimization of the interframe processing of the sequence of analyzed frames. This method allows estimating the statistical characteristics of the signal and noise components and calculating the correlation integral when detecting moving low-contrast objects. The proposed algorithm for detecting moving objects in low illuminance conditions allows preventing the manifestation of the blur effect.