Application of the convolutional neural network to design an algorithm for recognition of tower lighthouses
2017 24th Saint Petersburg International Conference on Integrated Navigation Systems (ICINS), 2017
The paper considers the main features of the convolutional neural network architecture and their ... more The paper considers the main features of the convolutional neural network architecture and their application to design the algorithm for recognition of tower lighthouse images. Tower lighthouses are supposed to have red and white horizontal bands. Their images can be received from some electronic video devices. A brief review of the definitions used in this subject area, such as layers and layer types: convolution layer, pooling layer, and fully-connected layer, is given. The process of training the neural network and the most popular technologies for developing convolution neural networks are described. A recognition algorithm has been designed, which proved to be correct in about 69–73% of the test cases.
Uploads
Papers by Mohamed Salama