Wyniki wyszukiwania - BazTech

1

The decision tree seabed type classifier operating on acoustic data

Dung T. V., Moszyński M., Stepnowski A.

Hydroacoustics

|

2001

|

Vol. 4

53--56

EN

A decision tree classifier was developed for sea bottom recognition from acoustic echoes. The acoustic data was acquired by DT4000 echosounder at 200 kHz frequency. The performance of the classifying system was investigated involving various backscattered echo parameters, in particular wavelet coefficients. The results of the decision tree classification were compared with those obtained from the adaptive neuro-fuzzy system (IFNN) involving reduced number of input parameters by the use of Principal Component Analysis (PCA).

2

Wieloczęstotliwościowe wieloetapowe klasyfikatory neuronowo-rozmyte do rozpoznawania typu dna morskiego

Dung T. V., Stepnowski A.

Hydroacoustics

|

2000

|

Vol. 3

39--44

EN

A hybrid multistage neuro-fuzzy classifiers were developedfor sea-bottom recognition from aeoustie eehoes. A multistage fuzry neural network was implemented and tested on the data eolleeted on two eehosounder's frequencies. Two struetures termed as incremental fuzz» neural network (IFNN) and aggregated fuzzy neural network (AFNN), were analysed. In IFNN, an approximate decision is undertaken firstly, based only on the one set of input variables. The decision is then fine-tuned by eonsidering more faetors in following stages until the final decision, assigning the output class, is undertaken. In AFNN, the input variables are divided into M subsets, where eaeh of them isfed to one sub-stage. The fina l output is derived by the reasoning with alt intermediate variables, which work as the outputs of substages in the preeeding stage. The proposed structures improve the generalisation ability of the system and reduees requirements on computation power and memory.

3

Incremental neuro-fuzzy classifiers of seabed using wavelet decomposition

Dung T. V., Stepnowski A., Moszyński M.

Hydroacoustics

|

2000

|

Vol. 3

19--24

EN

The neuro-fuzzy classifier of seabed type from acoustic echoes was investigated in the context of possible reducing the number of input parameters. The incremental architecture of fuzzy neural network classifier (IFNN) was used in the experiment, utilising dual-frequency echo collection. In particular, the wavelet decomposition of these bottom echoes was used to generate input parameters of IFNN. The Principal Component Analysis (PCA) was subsequently applied for redundant parameters reduction.

4

Sea boftom typing using neuro-fuzzy classifier operating on multi-frequency data

Dung T. V., Maciołowska J., Stepnowski A.

Hydroacoustics

|

1999

|

Vol. 2

79--84

EN

A hybrid neuro-fuzzy classifier was development for sea-boftom identification from acoustic echoes. A multistage ANFIS structure was constructed and tested on data collected on 38kHz and 120kHz echosounder's frequencies. In multistage systems available data is processed in stages. The decisions about assigning a boftom echo, represented by digitised echo envelope's parameters. to one of the classes is made hierarchically. Firstly, an approximate decision is made based only on one set of input variables. The decision is then fine-tuned by considering more and more factors, it is in following stages next parameters are taken under account until the final decision, corresponding to the output class. is made. The proposed approach nof only gives better classification results, as compared to paralleI ANFIS system, but also it demands less computation power.