Application of Full Connection Network in Submarine Formation Recognition
Liye Tian, Zuohao Shao, Jinping Wu · 2020
Recognizing enemy formation automatically is difficult for submarine's current software. Once implemented, the combat parameters can be calculated automatically and quickly. Full connection network in deep learning is introduced to recognize enemy's formation. Experiment with simulation data is carried out by Google's Tensorflow. The experimental results show that the double layer full connection network can realize the recognition function. Three kinds of formation are recognized. Experimental process and data are recorded in detail. The scale of training set suitable for research is about 200. When the nodes number in the first full connection layer is set to close to 64, the recognition rate is high.