Convolutional-Neural-Network-Based Handwritten Character Recognition: An Approach with Massive Multisource Data

Maritime ship recognition based on convolutional neural network and linear weighted decision fusion for multimodal images

Ship images are easily affected by light, weather, sea state, and other factors, making maritime ship recognition a highly challenging task. To address the low accuracy of ship recognition in visible images, we propose a maritime ship recognition method based on the convolutional neural network (CNN) and linear weighted decision fusion for multimodal images. First, a dual CNN is proposed to learn the effective classification features of multimodal images (i.e., visible and infrared images) of the ship target. Then, the probability value of the input multimodal images is obtained using the softmax function at the output layer. Finally, the probability value is processed by linear weighted decision fusion method to perform maritime ship recognition. Experimental results on publicly available visible and infrared spectrum dataset and RGB-NIR dataset show that the recognition accuracy of the proposed method reaches 0.936 and 0.818, respectively, and it achieves a promising recognition effect compared with the single-source sensor image recognition method and other existing recognition methods.