1.1 Introduction At present, the design of multi-sensor information fusion system lacks a unified specification and can not find an effective method. The information fusion model mainly includes the fusion functional model, structural model and fusion mathematical model. The functional model is based on the process of fusion, describing the functions of the information fusion, the database and the process of the star lake between the components of the system when the information is merged. The structural model starts from the fusion composition and explains the structure of the information fusion system; the mathematical model It is an information fusion algorithm and integrated logic. These three models must be solved by any fusion system, so they constitute the core problem of the fusion system, which is the key to the mathematical model, and is the most part of the current research.
1.2 Functional model of information fusion Functional model of information fusion At present, many scholars have proposed the general functional model of information fusion system from different angles. The most authoritative is the technical committee under the laboratory of the United States Armed Forces Government. Functional model. The model divides data fusion into three levels. The first level is single-source or multi-source processing, mainly digital processing, tracking correlation and correlation; the second level is to evaluate the set of target estimates, and their relationship with each other to assess the whole situation; the third pole uses a system A priori set of targets to test the assessment.
1.2.1 Distributed multi-sensor information fusion Distributed multi-sensor information fusion is the observation of the parameter model to be estimated or the phenomenon to be determined by each local sensor. By the way, the estimation or decision is given and their results are transmitted. At the Fusion Center, the Fusion Center combines the results of all sensors to get the final estimate or decision.
1.2.2 Central multi-sensor information fusion Central multi-sensor information fusion is that the sensor can completely transmit the observation to the fusion center, which is equivalent to all the observations directly obtained by the fusion center. This fusion method is called central multi-sensor information fusion.
1.3 Hierarchical description of information fusion In the actual environment, the information received by various types of sensors may be real-time information or non-real-time information; it may be fast-changing or slow-changing; it may be fuzzy or possible. It is certain; it may be mutual support or complementarity, or it may be contradictory or competitive. The basic principle or starting point of multi-sensor information fusion is to make full use of multiple sensor resources. Through reasonable control and use, the redundant or complementary information of multiple sensors in space or time is fused according to certain criteria to obtain the measured. The consistent description or interpretation of the object results in the system obtaining superior performance over systems consisting of a subset of the other components. Complementary, or it may be contradictory or competitive. The basic principle or starting point of multi-sensor information fusion is to make full use of multiple sensor resources. Through reasonable control and use, the redundant or complementary information of multiple sensors in space or time is fused according to certain criteria to obtain the measured. The consistent description or interpretation of the object results in the system obtaining superior performance over systems consisting of a subset of the other components. The essential difference between multi-sensor information fusion and classical signal processing methods is that the multi-sensor information processed by information fusion has more complex forms and can be embodied at different information levels. The main information representation levels are data layer, feature layer and decision layer. Multi-sensor information fusion adopts different implementation forms in different problem areas, so it is difficult to classify and describe a large number of emerging information problems. In general, most of the fusion problems are researched on the same level of information, so we simply describe the method of information fusion based on the information level of the fusion system.
1.3.1 Data layer fusion Data layer fusion is characterized by fusion directly on the detection decision layer or signal layer in the multi-sensor distributed detection system. It belongs to the underlying data fusion, and the advantage is that the amount of information is large! The information is accurate, and the defect is that it is difficult to achieve real-time requirements, data communication is large, and anti-interference ability is poor. At the same time, each sensor information is required to have homogeneity, otherwise scale calibration is required. Data layer fusion is commonly used for multi-source image synthesis, image analysis and understanding.
1.3.2 Feature Layer Fusion Feature layer fusion is the fusion of the information obtained by extracting the original data of each sensor. These feature information includes edge, direction, speed, shape, and the like. In general, the process of forming features is a relatively large process of information compression, which provides a prerequisite for real-time processing. Feature layer fusion can be divided into two categories: target state information fusion and target feature fusion. (1) Target state information Fusion target state information fusion is mainly applied to the target tracking field of multi-sensors. A large number of methods in the target tracking field can be modified to be multi-sensor target tracking methods. The fusion system first preprocesses the sensor data to complete the data registration, that is, transforms the input data of each sensor into a unified data expression form through coordinate transformation and unit conversion. After data registration, the fusion processing mainly implements parameter correlation and State estimate. Commonly used are sequential estimation techniques, including Kalman filtering and extended Kalman filtering. (2) Target feature information Fusion target feature information fusion is the feature layer joint recognition, which is essentially the pattern recognition problem. The multi-sensor system provides more feature information about the target than the single sensor for recognition, increasing the feature space dimension. The specific fusion method is still the corresponding technology of pattern recognition, but the feature must be correlated before the fusion, and then the feature vector is classified into a meaningful combination. The fusion recognition of the target is based on the joint feature vector after association, and the specific implementation technology includes the parameter template method! Feature compression and clustering algorithms, K-order nearest neighbors, neural networks, etc. In addition, knowledge-based reasoning techniques can also be applied to feature fusion recognition, but because of the difficulty in extracting prior knowledge of environment and target features, this aspect The research only began. Feature layer fusion has gradually matured in theory and application, forming a set of specific solutions to problems. In the three levels of integration, the fusion on the feature layer can be said to be the most developed, and since a set of effective feature association techniques have been established in the feature layer, the consistency of the fusion information can be guaranteed. Layer integration has good application and development prospects.
1.3.3 Decision-making layer fusion Feature layer fusion gradually matures in theory and application, forming a set of specific solutions to the problem. In the three levels of integration, the fusion on the feature layer can be said to be the most developed, and since a set of effective feature association techniques have been established in the feature layer, the consistency of the fusion information can be guaranteed. Layer integration has good application and development prospects. Decision-making layer fusion output is a joint decision result. In theory, this joint decision should be more precise or clear than any single-sensor decision. The methods used in decision-making fusion are: Bayes theory, DS evidence theory, fuzzy set theory and expert system. Method, etc. Decision-making layer fusion has high flexibility in information processing. The system has low requirements for information transmission bandwidth. It can effectively integrate different types of information reflecting various aspects of the environment or target, and can process asynchronous information. Therefore, the current information fusion A large number of research results are obtained at the decision-making level and constitute a hot spot of information fusion. However, due to the time-varying dynamic characteristics of the environment and objectives, the difficulty in acquiring prior knowledge, the huge characteristics of the knowledge base, and the object-oriented system design requirements, the development of decision-making fusion theory and technology is still limited.
The existing mathematical models of information fusion can be divided into three categories: (1) embedded constraint perspectives (2) evidence combination perspectives (3) neural network methods, although some methods are not perfect, but these methods and specific methods based on these methods The algorithm does solve many practical problems and promotes the development of information fusion technology. Chapter 3 Prospects of Multi-sensor Information Fusion Although data fusion has been widely used, a complete theoretical system and effective fusion algorithms have not been formed so far. Most of them are researched for specific problems and specific fields. That is to say, the current research on data fusion is based on the types of problems, specific objects, specific levels to establish their own fusion models and inference rules, and some This forms a so-called best solution. The so-called best criteria, best judgments, etc. are only theoretically passed. There is still a large distance if applied. Even if it is applied in practice, there is no perfect evaluation system to make a reasonable evaluation. Therefore, the design of multi-sensor data fusion system has certain blindness, it is necessary to establish a complete methodology system to guide the design of data fusion system.
The specific deficiencies are: 1) no basic theoretical framework and effective generalized models and algorithms are formed;
2) Parallel ambiguity is the main obstacle to data fusion;
3) The fault tolerance or robustness of the fusion system is not well resolved;
4) The research on the specific methods of data fusion is still in its preliminary stage;
5) There are still many practical problems in the design of data fusion systems. With the development of related technologies such as sensor technology, data processing technology, computer technology, network technology, artificial intelligence technology, parallel computer software and hardware technology, multi-sensor data fusion will become the intelligent detection and data processing of complex T industry systems in the future. Important technology.
From the current domestic and international research data collected, the research directions of multi-sensor data fusion are summarized as follows:
1) Improve the fusion algorithm to further improve the performance of the fusion system. At present, the combination of fuzzy logic, neural network, evolutionary computation, rough set theory, support vector machine, wavelet transform and other computational intelligence technologies is an important development trend.
2) How to use the relevant prior data to improve the fusion performance of data is also a problem that needs to be studied carefully.
3) Develop parallel computing software and hardware to meet the requirements of computationally complex multi-sensor fusion of large amounts of data.
4) Research on integrated circuit chips that can handle multi-sensor integration and fusion, as well as sensor model and interface standardization are the main development directions of current system hardware.
5) Research on multi-sensor integration and fusion in unknown and dynamic environments.
6) Research on multi-sensor integration and fusion using parallel computer architecture.
7) Carry out research on virtual reality technology and provide an ideal simulation platform for multi-sensor data fusion research.
With the development of technology and advances in technology, multi-sensor information fusion technology is more and more widely used in various fields, especially military applications. Information fusion technology can synthesize different information of each sensor and preprocess, correlate, decide and integrate the target information. This paper mainly studies some key technologies in information fusion. The main work is: feature layer fusion is an important part of information fusion. Decision layer fusion is a key part of information fusion. Data decision making judgments from multiple targets. To determine the goal of the final strike "This paper analyzes the commonly used DS evidence synthesis theory and its improvement methods. In view of the shortcomings of DS evidence synthesis theory, the application of information theory to data decision-making, this method can well solve the emergence of DS evidence theory. The conflicts are too big, one vote is vetoed, etc., and reasonable integration effects are obtained. Multi-sensor information fusion still has a long way to go in the future, and it is necessary to constantly overcome various difficulties. On the basis of the present, it is better to be people. Use to facilitate people's lives.
Fiber Optical Distribution Box (Plastic)
Fiber Optic Distribution Box / Fiber Splitter Box is used as a termination point for the feeder cable to connect with drop cable. The fiber splicing, splitting, distribution can be done in this box, and meanwhile it provides solid protection and management for the FTTX network building.
Material: PC+ABS,wet-proof, water-proof, dust-proof, anti-aging, protection grade is IP65.
Small Volume, For Indoor/Outdoor application. Compact structure,convenient installation and maintenance.
Fiber Optical Distribution Box,Plastic Fiber Optical Distribution Box,Fiber Optic Distribution Box,Optical Distribution Boxes
Sijee Optical Communication Technology Co.,Ltd , https://www.sijee-optical.com