Bijo Sebastian Department of Computer Applications, SNGIST Group of InstitutionsThekkethazham, Mannam P.O, N-Paravur- 683520

OVER VIEW1) Introduction.

2) Digital Image Processing.

3) Machine Learning (ML).

4) OpenCV.

5) How to train a machine.

6) Applications of Computer Vision

7) Conclusion.

Computer vision is the science and technology of machines that see, where see in this case means that the machine is able to extract information from an image(from a video frame) that is necessary to solve some task.

RELATED FIELDS Content-based image retrieval Machine vision Machine Learning Opencv Gesture recognition Robotic vision Face –recognition Camera based game Bionic Vision.

What is vision? Recognize objects

– people we know

– things we own Locate objects in space

– to pick them up Track objects in motion

– catching a baseball

– avoiding collisions with cars on the road Recognize actions

– walking, running, pushing

Human vision is passive

It relies on external energy sources

(sunlight, light bulbs, fires) providing light that reflects off of objects to our eyes

Vision systems can be “active” - carry their own energy sources Radars Bat acoustic imaging systems

Fig1: The components of a computer vision system

Digital Image Processing

Digital image processing is the use of computer algorithms to perform image processing on digital images .

Application of Digital Image Processing

Image processingImage processing is a physical process used to convert an image signal into a physical image. The image signal can be either digital or analog. The actual output itself can be an actual physical image or the characteristics of an image. The most common type of image processing is photography.

Why do we need Image Processing?It Is motivated by three major applicationsImprovement of pictorial information for human perception.

Image processing for autonomous machine application.

Efficient storage and transmission.

Digital representation of image.

Edge DetectionEdge detection is a fundamental tool in image processing and computer vision, particularly in the areas of feature detection and feature extraction, which aim at identifying points in a digital image at which the image brightness changes sharply or more formally, has discontinuities.

Edge detection Canny edge detection applied to a photograph

Application of Sobel operator on a colored image.

Machine Learning

Field of study that gives computers the ability to learn without being explicitly programmed. A major focus of machine learning research is to automatically learn to recognize complex patterns and make intelligent decisions based on data

Definition

A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.

Learning Models.Memory-Based LearningIn machine learning, instance-based learning or memory-based learning is a family of learning algorithms that, instead of performing explicit generalization, compare new problem instances with instances seen in training, which have been stored in memory. Instance-based learning is a kind of lazy learning.

Case-Based ReasoningA sub field of AI, called “Case-Based Reasoning" (CBR), can be viewed as a generalized kind of memory-based learning. In CBR a stored library of “cases" is used to help in the analysis, interpretation, and solution of new cases.

Neural NetworksDuring the 1960s, neural net researchers employed various methods for changing a network's adjustable weights so that the entire network made appropriate output responses to a set of “training" inputs. For example, Frank Rosenblatt at Cornell adjusted weight values in the final layer of what he called the three-layer alpha-perceptron.

1.The Backprop AlgorithmThat problem was solved in the mid-1980s by the invention of a technique called “back propagation" (backprop for short) introduced by DavidRumelhart,Georey E. Hinton, and Ronald J. Williams.33 The basic idea behind backprop is simple, but the mathematics (which I'll skip) is rather complicated. In response to an error in the network's output, backprop makes small adjustments in all of the weights so as to reduce that error.

2.NETtalk One very interesting application of the backprop learning method was developed by Terrence J. Sejnowski (1947 ) and Charles Rosenberg (1961).They taught a neural network to talk!. In one of their experiments, their system, called NETtalk, learned to “read" text that was transcribed from informal, continuous speech of a six-year-old child and produced acoustic output.

3.ALVINNAnother neural network application, this one for steering a van, was developed by Dean Pomerleau, a Ph.D. student at Carnegie Mellon University. The system, which included the van, a TV camera for looking at the road ahead, and interface apparatus, was called ALVINN, an acronym for Autonomous

Reinforcement Learning

In its simplest setting, reinforcement learning is about learning how to traverse a collection of states, going from one state to another and so on, to reach a state in which a reward is obtained. The rat's problem is to go from its starting position to the cheese at the goal position. The gray dots in the figure are meant to depict situations that the rat might find itself in and recognize. In reinforcement learning terminology, these situations are called “states." At each state, the rat can select from among, say, four actions, namely, turn left, turn right, go forward, or go back. Depending on the state, only some of the actions are possible one cannot go forward when up against a dead end for example. Each possible action takes the rat from one state to an adjacent one in the maze. The collection of states and the actions that link them can be thought of as a graph.

OPEN-CV OpenCV [OpenCV] is an open source computer vision library available from The library is written in C and C++ and runs under Linux, Windows and Mac OS X.

OpenCV was designed for computational efficiency and with a strong focus on real time applications

One of OpenCV’s goals is to provide a simple-to-use computer vision infrastructure that helps people build fairly sophisticated vision applications quickly.

The OpenCV library contains over 500 functions that span many areas in vision, including factory product inspection, medical imaging, security, user interface, camera calibration, stereo vision, and robotics. Because computer vision and machine learning oft en go hand-in hand.

OpenCV also contains a full, general-purpose Machine Learning Library (MLL). The MLL is highly useful for the vision tasks that are at the core of OpenCV’s mission, but it is general enough to be used for any machine learning problem.

HAAR TRAININGHowever, face detection is such a common need that it is worth having a baseline technique that works fairly well; also, the technique is built on the well-known and often used field of statistical boosting and thus is of more general use as well. In fact, several companies have engineered the “face” detector in OpenCV to detect “mostly rigid” objects (faces, cars, bikes, human body) by training new detectors on many thousands of selected training images for each view of the object. This technique has been used to create state-of-the-art detectors, although with a different detector trained for each view or pose of the object. Thus, the Haar classifier is a valuable tool to keep in mind for such recognition tasks.

an example of OpenCV face Detector in action

A set of basic Haar-like features.

A set of extended Haar-like features

Computer Vision ApplicationsHere the interest is on procedures for extraction of image information for computer processing .

Typical ApplicationsIndustrial Machine vision for product assembly and inspection.

Automated target detection and tracking. Finger print recognition. Machine processing of areal and satellite

imagery for weather prediction and crop assesment etc.

Another important feature of artificial vision is that it is tireless and that all products can be scrutinized and measured.

Other application areas are surveillance, automated inspection, robot assembly, vehicle guidance, traffic monitoring and control, biometric measurement, and analysis of remotely sensed images. By way of example, fingerprint analysis and recognition have long been important applications of computer vision, as have the counting of red blood cells, signature verification and character recognition, and aero plane identification (both from aerial silhouettes and from ground surveillance pictures taken from satellites). Face recognition and even iris recognition have become practical possibilities, and vehicle guidance by vision will in principle soon be sufficiently reliable for urban use.

Conclusion. Computer vision is a subfield of computer Science. It’s a general term for the whole field, including image

processing, medical image analysis, image understanding. It’s the same as Machine Vision. Computer Vision technology improve human life. Computer Vision is an active research field with many

research groups in countries all over the world. There exists a large body of research results to build on.