图像识别是数字图像处理的高级课题。统计法、句法法、模糊法和神经网络识别

Image recognition is an advanced topic of digital image processing. Statistical, syntactic, fuzzy and neural network recognition methods are currently commonly used image recognition methods. They have their own strengths and can effectively solve image recognition problems within a certain range. <br>Previously, the image recognition methods were classified, discussed and implemented. When solving practical problems, these methods are usually used in combination, and it is rare to use one method completely alone. For example, in order to improve the recognition rate, fuzzy vectors and fuzzy matching can be introduced in statistical recognition; fuzzy productions can be introduced in syntactic recognition; fuzzy learning mechanisms can be introduced in neural network recognition. The combined use of two or more methods can complement each other's strengths and complement each other's advantages, thereby improving the image recognition rate. <br>Today, fuzzy neural networks have not only been used in consumer electronics and industrial control, but also in many fields such as system identification, image processing, pattern recognition, data mining, and financial engineering. At the same time, the "dynamic" fuzzy neural network D-FNN algorithm that gradually formed fuzzy rules during the learning process was also developed. The author constructed a D-FNN classifier to classify and recognize benign and malignant tumors by extracting seven statistical features of the image texture, and through experiments with a single-layer BPN classifier and RBF neural network classifier structure The comparison shows that a better recognition effect is obtained. <br>In addition, with the development of pattern recognition research, the combination of multiple classifiers has also become a research hotspot in image recognition. Especially in the field of character recognition, there have been many systems that use different features and different classifiers to combine with each other, and their recognition performance has been greatly improved. The reported methods include: majority decision method, linear weighting method, Bayesian estimation, evidence reasoning method, fuzzy reasoning method, and neural network combination method that uses classification results as a new input feature, etc. There are many methods However, there is still a lack of a unified theory. There are also some problems in the field of face recognition, such as how to determine or recognize other posture images given a frontal image of a person. This involves the acquisition of three-dimensional information. There have been some reports in this regard in the world, and this is also one of the current hotspots in image recognition research. <br>In a word, the image processing toolbox, wavelet toolbox, fuzzy logic toolbox and neural network toolbox provided by Matlab can be used to solve the image recognition problem conveniently.

Image recognition is an advanced subject of digital image processing. Statistics, syntax, fuzzy and neural network recognition are commonly used image recognition methods, they each have their own strengths, each can effectively solve the problem of image recognition within a certain range.<br>The image recognition method is discussed and implemented. When solving real-world problems, these methods are often combined, and it is rare to use one method entirely alone. For example, in order to improve the recognition rate, fuzzy vectors and fuzzy matches can be introduced in statistical recognition. The introduction of fuzzy generation in syntactic recognition; Fuzzy learning mechanism can be introduced in neural network recognition. The combination of two or more methods can achieve the advantages and advantages of complementing each other, thereby improving the image recognition rate.<br>Today, fuzzy neural networks have been used not only in consumer electronics, industrial control, but also in many fields such as system identification, image processing, pattern recognition, data mining, financial engineering, etc. At the same time, the "dynamic" fuzzy neural network D-FNN algorithm, which gradually forms fuzzy rules in the course of learning, has also been developed. By extracting seven statistical features of image texture, the author constructed a D-FNN classifier to classify and identify benign and malignant tumors by mammograms, and compared the structure of single-layer BPN classifier and RBF neural network classifier with experiments to obtain better recognition effect.<br>In addition, with the development of pattern recognition research, the combination of multi-classifiers has become a hot research topic in image recognition. Especially in the field of character recognition, there are many systems that utilize different characteristics and combine different classifiers, and their recognition performance has been greatly improved. The methods reported in these reports are: multi-judgment method, linear weighting method, Bayesian estimation, evidence reasoning method, fuzzy reasoning method, as well as the classification results as a new input characteristics of the neural network combination method, the method is more, but there is still a lack of a unified theory. There are also some problems in the field of face recognition, such as how to identify or recognize images of a given person's frontal image and its other postures. This involves the acquisition of 3D information, there are some international reports in this regard, which is also one of the hot spots of image recognition research.<br>In short, the image recognition problem can be easily solved by combining Matlab's image processing toolbox, wavelet toolbox, fuzzy logic toolbox and neural network toolbox.

Image recognition is an advanced subject of digital image processing. Statistical method, syntax method, fuzzy method and neural network recognition method are commonly used image recognition methods at present. They have their own advantages and can effectively solve the problem of image recognition in a certain range.<br>In front of the image recognition methods are discussed and implemented. In solving practical problems, these methods are usually used together, and it is rare to use one method alone. For example, in order to improve the recognition rate, fuzzy vector and fuzzy matching can be introduced into statistical recognition; fuzzy production can be introduced into syntactic recognition; fuzzy learning mechanism can be introduced into neural network recognition. The combination of two or more methods can complement each other and improve the image recognition rate.<br>Up to now, fuzzy neural network has been applied not only in consumer electronics, industrial control, but also in many fields, such as system identification, image processing, pattern recognition, data mining, financial engineering and so on. At the same time, the "dynamic" fuzzy neural network D-FNN algorithm which gradually forms fuzzy rules in the learning process is also developed. The author constructs a D-FNN classifier by extracting seven statistical features of image texture to classify and recognize benign and malignant tumors in breast X-ray photos, and compares it with single-layer BPN classifier and RBF neural network classifier through experiments, and obtains better recognition effect.<br>In addition, with the development of pattern recognition research, the combination of multiple classifiers has become a research hotspot in image recognition. Especially in the field of character recognition, there are many systems which combine different features and classifiers, and their recognition performance has been greatly improved. The reported methods include: multiple decision, linear weighting, Bayesian estimation, evidential reasoning, fuzzy reasoning, and neural network combination method which takes the classification results as a new input feature. There are also some problems in the field of face recognition, such as how to determine or recognize other pose images given a person's frontal image. This involves the acquisition of three-dimensional information, which has been reported in the world, and it is also one of the hot spots of image recognition research.<br>In a word, the problem of image recognition can be easily solved by using the image processing toolbox, wavelet toolbox, fuzzy logic toolbox and neural network toolbox provided by MATLAB.<br>