Multi-Sensor Data Fusion Algorithm Based On Intrusion Detection System

Invasion detection systems are mostly based on a single type of sensor system. In this system, we use sensors to monitor a variety of targets, and miscarriage of justice or missing issue often happens. Therefore we upgrade the intrusion detection system to multi- types of sensor based systems. How can a variety of sensors collect data for effective use, in order to achieve multi-sensor system functions similar to a single sensor system to avoid the occurrence of false positives or missing is the significance question of data fusion technology. The use of data fusion technology can collect multiple types of sensors for efficient use of target data obtained detailed feature information from each target, which can produce more than a single type of sensor system completely accurate judgments.Existing data fusion algorithms include estimation, statistics, information theory method, and so on. These methods are too cumbersome; require a lot of math operation, not suitable for the present sensor system. Therefore, this thesis presents a neural network data fusion, neural network method does not require a lot of complicated mathematical derivation, and like a similar human brain for data processing, the final results obtained on the target of discrimination. Through the neural network data fusion methods can be very convenient to reach multi-sensor data fusion for the purpose and neural network multi-sensor data fusion target than a single sensor to determine the rate has greatly improved.In this paper, complete the following tasks: the completion of the Ash River basin for water quality monitoring system monitoring stations of the multi-sensor intrusion detection system hardware, the use of the hardware system to collect the body, motor vehicles and dogs sensor information; completed collecting sensor data analysis and processing, extraction of feature information, after normalization, the feature to get a new neural network input vector as a sample; completed the training of BP neural network, and the inadequacies of proposed improvements, for unimproved and improved neural network after the simulation training, and through the target with a single sensor data to determine the results were compared.Verified by actual testing, improved neural network data fusion can be a convenient and effective method for data integration, and for a single type of sensor systems for target identification of uncertainty has been significantly improved, determine the success rate has been significantly improved.