Accession Number : ADA273753

Title :   A Feasibility Study on Bird Classification with Neural Network.

Descriptive Note : Final rept.,

Corporate Author : FYSISCH EN ELEKTRONISCH LAB TNO THE HAGUE (NETHERLANDS)

Personal Author(s) : Meiler, P. P.

Report Date : JUN 1992

Pagination or Media Count : 61

Abstract : This study shows that it is feasible to classify flying birds using radar and neural network technology. The Royal Dutch Airforce is interested in the capability to classify birds because this capability can be used to avoid collisions between birds and airplanes. The Automatic Gain Control (AGC) signal which is generated by the Flycatcher tracking radar has a relationship with the wing stroke pattern of a bird. An automatic system to classify birds using the AGC signal could be used in a bird collision warning system. Such a system does not yet exist. A prototype of a bird classification system has been implemented and evaluated. Test results based on simulated AGC data show that the prototype is able to classify birds. The prototype uses simulated AGC data because there is not yet enough real AGC data available to use neural network technology. Acquisition of real AGC data is too expensive to be done in the framework of a feasibility study. According to the test results it is recommended to acquire real AGC data.

Descriptors :   *BIRDS, *CLASSIFICATION, *NEURAL NETS, FEASIBILITY STUDIES, PROTOTYPES, BIRD STRIKES, COLLISION AVOIDANCE, NETHERLANDS, DUTCH LANGUAGE, DATA ACQUISITION, EXPERIMENTAL DATA, RADAR TRACKING.

Subject Categories : Biology
      Cybernetics
      Military Aircraft Operations

Distribution Statement : APPROVED FOR PUBLIC RELEASE