The prerequisites for applying classical inference, Bayesian inference, Dempster-Shafer evidential theory, artificial neural networks, voting logic, and fuzzy logic data fusion algorithms to target detection, classification, and identification have been discussed. These data fusion techniques require expert knowledge, probabilities, or other information from the designer to define either:
â¢ Acceptable Type 1 and Type 2 errors;
â¢ a priori probabilities and likelihood functions;
â¢ probability mass;
â¢ neural network type, numbers of hidden layers and weights, and training data sets;
â¢ confidence levels and conditional probabilities; or
â¢ membership functions and production rules.
The information required to execute these algorithms is summarized in Table 11.1. Implementation of the data fusion algorithms is thus dependent on the expertise and knowledge of the designer (e.g., to develop production rules or define the artificial neural network type and parameters), analysis of the operational situation (e.g., to establish values for the Type 1 and Type 2 errors), applicable information stored in databases (e.g., to calculate the required probabilities or confidence levels), and the types of information provided by the sensor data (e.g., is the information adequate to calculate probability masses or differentiate among confidence levels?).