【摘要】 根据遥感图像的特点,针对海面目标难以准确识别的问题,提出了一种基于深度卷积神经网络的船只识别方法。首先利用分类网络进行图像的预分类,然后在分类结果的基础上,构成双通道的识别体制,识别网络采用Faster R-CNN。针对受云雾遮挡的船只识别问题,利用改进的深度卷积神经网络结构开展网络训练与调优,处理结果的F1-Score最高可达0.725 3。训练的网络模型表现出很好的船只目标识别能力,处理结果证明了该方法的有效性与准确性。更多还原

【Abstract】 In view of the characteristics of remote sensing images, according to the problem that the sea surface targets are difficult to accurately identify, we presented a method for ship recognition based on the deep convolution neural network in this paper. We used the classification network to classify the images firstly. And then, based on the classification results, we formed a recognition system with two channels, and the recognition network adopted Faster R-CNN. We used the improved deep convolution neural network structure to carry out the network training and optimization for the ship recognition mission with cloud occlusion. The maximum F1-Score of the processing results is up to 0.725 3. The trained network model has the capacity of ship target recognition, and the experimental results can prove the validity and accuracy of this method.更多还原