Xinsheng IntelligentCompany NewsThe relationship between computer vision and image processing, pattern recognition, machine learning disciplines

In my understanding, the realization of computer vision must have the help of image processing, and image processing relies on the effective use of pattern recognition, and pattern recognition is an important branch in the field of artificial intelligence, and artificial intelligence and machine learning are inseparable. Looking at all relationships, it is found that applications of computer vision serve machine learning. All aspects are indispensable and complement each other.

Computer vision uses computers to simulate the ability of human visual mechanism to acquire and process information. It refers to the use of cameras and computers instead of human eyes to identify, track and measure the target and other machine vision, and further do graphics processing, and use computer processing to become images that are more suitable for human eyes to observe or transmit to instruments for detection. Theories and technologies related to computer vision research, trying to build artificial intelligence systems that can obtain 'information' from images or multi-dimensional data. The challenge of computer vision is to develop human-level visual abilities for computers and robots. Machine vision requires image signals, texture and color modeling, geometric processing and reasoning, and object modeling. A capable vision system should have all of these processing tightly integrated.

Image processing, the technique of using a computer to analyze images to achieve a desired result. Also called image processing. Basic content image processing generally refers to digital image processing. Digital image refers to a large two-dimensional array obtained by sampling and digitization with digital cameras, scanners and other equipment. The elements of the array are called pixels, and its value is an integer, called gray value. The main content of image processing technology includes image compression, enhancement and restoration, matching, description and recognition. Common processes include image digitization, image coding, image enhancement, image restoration, image segmentation, and image analysis. Image processing generally refers to digital image processing.

Pattern recognition refers to the process of processing and analyzing various forms of (numerical, textual and logical relationship) information that characterize things or phenomena to describe, identify, classify and explain things or phenomena. , is an important part of information science and artificial intelligence. Pattern recognition is also often referred to as pattern classification. From the perspective of dealing with the nature of the problem and the method of solving the problem, pattern recognition is divided into two types: supervised classification (Supervised Classification) and unsupervised classification (Unsupervised Classification). Patterns can also be divided into abstract and concrete forms. The former, such as consciousness, thought, discussion, etc., belongs to the category of concept recognition research and is another research branch of artificial intelligence. The pattern recognition we mean is mainly to identify and classify the specific patterns of objects such as speech waveforms, seismic waves, electrocardiograms, electroencephalograms, pictures, photos, texts, symbols, and biosensors. Pattern recognition research mainly focuses on two aspects, one is to study how objects (including people) perceive objects, which belongs to the category of cognitive science, and the other is to use computers to realize the theory and method of pattern recognition under a given task. A computer is used to identify and classify a set of events or processes. The identified events or processes can be specific objects such as words, sounds, and images, or abstract objects such as states and degrees. These objects are distinguished from information in digital form and are called pattern information. Pattern recognition is related to statistics, psychology, linguistics, computer science, biology, cybernetics, etc. It has a cross relationship with the research of artificial intelligence and image processing.

Machine learning is the study of how computers simulate or implement human learning behaviors to acquire new knowledge or skills, and to reorganize existing knowledge structures to continuously improve their performance. It is the core of artificial intelligence and the fundamental way to make computers intelligent. It is applied in all fields of artificial intelligence. It mainly uses induction and synthesis instead of deduction. Machine learning plays a very important role in the research of artificial intelligence. An intelligent system without learning ability can hardly be called a real intelligent system, but the previous intelligent systems generally lacked the ability to learn. These limitations are becoming more and more prominent with the in-depth development of artificial intelligence. It is in this context that machine learning has gradually become one of the cores of artificial intelligence research. Its applications have spread to various branches of artificial intelligence, such as expert systems, automatic reasoning, natural language understanding, pattern recognition, computer vision, intelligent robots and other fields. The research of machine learning is based on the understanding of human learning mechanism, such as physiology and cognitive science, to establish a computational model or cognitive model of the human learning process, to develop various learning theories and learning methods, to study general learning algorithms and to carry out theoretical analysis. , to build task-oriented learning systems with specific applications. These research goals influence each other and promote each other.

The purpose of human research on computers is to improve the level of social productivity, improve the quality of life, and rescue people from monotonous, complex and even dangerous work. Today's computers have far surpassed humans in computing speed. However, in many aspects, especially in aspects related to human intelligent activities such as visual function, auditory function, olfactory function, natural language comprehension function, etc., it is not as good as people.

This status quo cannot meet the requirements of some advanced applications. For example, we hope that computers can detect suspicious situations on the road early and alert car drivers to avoid accidents, and we hope that computers can help us drive autonomously. The current technology is not enough to meet the requirements of such advanced applications, and more needs to be done. A lot of artificial intelligence research results and experience in system implementation.

What is artificial intelligence? Artificial intelligence is a technology designed by humans and implemented in a computer environment to simulate or reproduce the intelligent behavior of some people. It is generally believed that human intelligence activities can be divided into two categories: perceptual behavior and thinking activities. Some examples of artificial intelligence research that simulate perceptual behavior include "computer hearing" related to human auditory functions such as speech recognition, speaker recognition, etc., and "computer vision" related to human vision, such as shape knowledge, distance, and speed perception of three-dimensional representations of objects. ",and many more. Examples of artificial intelligence research that simulate thinking activities include symbolic reasoning, fuzzy reasoning, theorem proving, etc. "computer thinking" related to human thinking, and so on.

One of the research objects of computer vision developed from image processing and pattern recognition is how to use two-dimensional projection images to restore the three-dimensional scene world. The theoretical methods used in computer vision are mainly based on geometric, probabilistic and kinematic calculations and visual computing theory of three-dimensional reconstruction. Its foundations include projective geometry, rigid body kinematics, probability theory and random processes, image processing, artificial intelligence and other theories . The basic goals to be achieved by computer vision are as follows:

(1) Calculate the distance from the observation point to the target object according to one or more two-dimensional projection images;

(2) Calculate the motion parameters of the target object according to one or more two-dimensional projection images;

(3) Calculate the surface physical properties of the target object according to one or more two-dimensional projection images;

(4) According to the multiple two-dimensional projection images, the projection image of the larger space area is recovered.

The ultimate goal of computer vision is to use computers to understand the three-dimensional scene world, that is, to realize some functions of the human visual system.

In the field of computer vision, the requirements for pattern recognition in medical image analysis and optical character recognition need to be raised to a certain height. Another example is the preprocessing and feature extraction in pattern recognition using image processing technology; image analysis in image processing also applies pattern recognition technology. In most practical applications of computer vision, computers are preset to solve specific tasks. However, machine learning-based methods are becoming more and more popular. Once machine learning research is further developed, future "generic" computer vision applications may be able to come true.

One of the main problems of artificial intelligence research is: how to make the system have "planning" and "decision-making ability"? So that it can perform a specific technical action (eg: moving a robot through a specific environment). This problem is closely related to computer vision problems. Here, the computer vision system acts as a perceptron, providing information for decision-making. Other research directions include pattern recognition and machine learning (which is also part of the field of artificial intelligence, but has an important connection with computer vision), and because of this, computer vision is often regarded as a branch of artificial intelligence and computer science.

Machine learning is the study of how computers simulate or implement human learning behaviors to acquire new knowledge or skills, and to reorganize existing knowledge structures to continuously improve their performance. It is the core of artificial intelligence and the fundamental way to make computers intelligent. It is applied in all fields of artificial intelligence. It mainly uses induction and synthesis instead of interpretation.

To achieve the purpose of computer vision, there are two technical approaches to consider. The first is the bionic method, which starts from analyzing the process of human vision, and uses the best reference system that nature provides us, the human visual system, to establish a computational model of the visual process, and then implement it with a computer system. The second is the engineering method, which is to break away from the constraints of the human visual system and use all feasible and practical technical means to realize the visual function. The general approach of this method is to treat the human visual system as a black box, and only care about what output the visual system will give for a certain input. Both methods are theoretically usable, but the difficulty is that the output of the human visual system corresponding to a certain input is not directly measurable. Moreover, since human intelligent activity is the result of the integrated action of a multifunctional system, even if an input-output pair is obtained, it is difficult to be sure that it is only a response generated by the current input visual stimulus, rather than a synthesis with the historical state. result of action.

It is not difficult to understand that the research of computer vision has a double meaning. One is to meet the needs of artificial intelligence applications, that is, to use computers to realize the needs of artificial vision systems. These achievements can be installed on computers and various machines, enabling computers and robots to have the ability to "see". Second, the research results of the visual computing model, in turn, also have considerable reference significance for us to further understand and study the mechanism of the human visual system itself, and even the mechanism of the human brain.