An Efficient Multisensor Hybrid Data Fusion Approach Based on Artificial Neural Networks and Particle Swarm Optimization Algorithms

Abstract

Data fusion plays a crucial part in performance evaluation processes in multisensor systems; thus, it is important to use an effective technique to cut down on errors. By improving the sensors' location and their capacity to adjust to the deployment geometry, the paper's technique for reducing data fusion errors is proposed. The preprocessing stage of data is also included in the suggested technique, which starts with the design of the data-collecting device and ends with a hybrid model algorithm. Particle swarm optimization and artificial neural network methods are combined in the hybrid algorithm. Preprocessing is based on the compensation and adjustment factors to ensure data gathering synergy. A true experimental apparatus was designed under three different deployment geometries, comprising five ultrasonic sensors constituting a multisensor system and a hardwood target (Triplochiton scleroxylon and Milicia excelsa). The results obtained show that the suggested deployment geometry has mean absolute percentage errors of 3.97, mean square errors of 0.182, and mean absolute errors of 0.045 which are lower than those associated with the deployment configurations found in the literature. The results of the proposed method show an overall detection accuracy in linear deployment geometry of 97.308%. Then, the circular deployment geometry obtained a detection accuracy of 97.005%. Finally, elliptical deployment geometry obtained a detection accuracy of 98.745%. These results show that the proposed method better predicts the range of a target with a lower error rate than conventional methods.