远洋船舶目标检测中图像分类识别研究

Research on image classification and recognition in ocean going ship target detection

为准确划分远洋船舶类别,实现远洋船舶目标检测,保障海洋生态环境、海洋交通与国防安全,研究远洋船舶目标检测中图像分类识别方法。采用小波分析法,通过二维小波变换获取低频子图像、水平细节子图像、垂直细节子图像和对角细节子图像,将其作为远洋船舶图像特征;采用采用双向递归神经网络(BRNN)的深度学习方法构建分类器,将远洋船舶图像特征作为输入向量,通过递归神经网络将输入向量映射至输出向量,得到船舶目标图像分类结果。同时利用改进粒子群算法优化分类器中的权重与偏差参数,提升船舶目标图像分类精度。实验结果显示,所研究方法能够有效划分船舶目标图像类别,且划分精度高于97%。

In order to accurately classify ocean ships, realize the target detection of ocean ships, and ensure the protection of marine ecological environment, marine traffic and national defense security, the image classification and recognition methods in ocean ship target detection are studied. The low-frequency sub image, horizontal sub image, vertical sub image and diagonal sub image are obtained by two-dimensional wavelet transform, which are used as the image features of ocean going ships;The deep learning method of bidirectional recurrent neural network(brnn) is used to construct a classifier. The features of ocean going ship images are taken as input vectors, and the input vectors are mapped to the output vectors through the recurrent neural network to obtain the classification results of ship target images;At the same time, the improved particle swarm optimization algorithm is used to optimize the weight and deviation parameters in the classifier, so as to improve the classification accuracy of ship target images. Experimental results show that the proposed method can effectively classify ship target images, and the classification accuracy is higher than 97%.