Neuro-fuzzy model for multi-channel underwater imaging
V.M. Contarino, Pavlo A. Molchanov, I.M. Petrosyuk, Yuliya Podobna, O.V. Asmolova · 2006
Multispectral imaging system usually consist 2-15 different color channels, hyperspectral system - 100-200 channels. The image processing in each channel includes the complicated calculations and the final results have quite a large error. As is well known that there are large number of input parameters and some their uncertainty in the case of airborne and underwater LIDAR systems modeling. The using of statistical and determined models give the result having quite a large error of optical information processing and the given calculations take a lot of time to compute. The fundamentally different mathematical algorithms - the neural networks and the fuzzy logic is offered to use. It is realized with specially developed algorithms for multi-channel image processing. The new neuro-fuzzy model of foam coverage for four color channels has been developed to determine the interval of the minimal reflections and to obtain the images of non-foam covered areas.