The application of neural networks in photogrammetry of dynamic targets

The photogrammetry system can provide the coordinate, speed and acceleration of a flying target in space. Its background is disorderly and interference sources are many in number, so that the target recognition and signal processing are difficult. This paper adopts multiple front freeback neural networks to replace traditional mode recognition methods. It has the features of high parallel arithmetic capacity, distributive information storation and parallel proceeding. According to the features of flying targets, useful signals are extracted from various information,therefore, the sampling and recognition can be realized. The combination of artificial neural network techniques, photo recognition and artificial intelligence techniques, improves the speed of photo processing and recognition, and increases the exactness.