Search

A practical hyperspectral target characterization task estimates a target signature from imprecisely labeled training data. The imprecisions arise from the characteristics of the real-world tasks. First, accurate pixel-level labels on training data are often unavailable. Second, the subpixel targets and occluded targets cause the training samples to contain mixed data and multiple target types. To address these imprecisions, this paper proposes a new hyperspectral target characterization method to produce diverse multiple hyperspectral target signatures under a multiple instance learning (MIL) framework. The proposed method uses only bag-level training samples and labels, which solves the problems arising from the mixed data and lack of pixel-level labels. Moreover, by formulating a multiple characterization MIL and including a diversity-promoting term, the proposed method can learn a set of diverse target signatures, which solves the problems arising from multiple target types in training samples. The experiments on hyperspectral target detections using the learned multiple target signatures over synthetic and real-world data show the effectiveness of the proposed method.

Recently, multiple instance learning (MIL) tends to be a popular method for hyperspectral target representation since it does not require the detailed information about the target. However, the data imbalance problem occurred in MIL usually has negative effects on the representation of hyperspectral target. To solve the problem, this paper propose a bag-level data balance-inducing multiple instance learning method for hyperspectral target representation. The method aims to balance the positive and negative bags by synthesizing new positive bags since the number of positive bags in the training data set is much smaller than that of negative bags. Experimental results show that the proposed method can improve the target representation ability and enhance the target detection performance.

Recently, deep models have shown their superiority on the representation of the hyperspectral image. However, the limited number of training samples in hyperspectral image tasks makes it difficult to obtain well-trained deep model. Generally, deep belief network (DBN) which makes use of unsupervised learning can be introduced to partly solve the problem. To further improve the above-mentioned problem, this work feds DBN into deep ensemble. However, traditional deep ensemble usually produces models that tend to be very similar to the MAP solution and each other. To overcome this problem, this work introduces a novel strategy which divides the training samples into several subsets and thus the obtained multiple models would be diversified. Then a special information fusion method is proposed to obtain the final inference. Experiments are conducted over Pavia Unversity dataset to evaluate the effectiveness of the proposed method.

Recently, deep models have shown powerful ability for remote sensing scene representation. However, the training process of these deep methods requires large amount of labelled samples while usual remote sensing image datasets cannot provide enough training samples. Therefore, the learned model is usually suboptimal. To solve the problem, this work focuses on obtaining multiple choices by training multiple models simultaneously, and then the human oracle can choose a proper one from these choices. However, training several models separately usually makes the obtained results similar. This paper tries to diversify the obtained choices by encouraging the obtained choices to repulse from each other. Experiments are conducted on Ucmerced Land Use dataset to validate the effectiveness of the proposed method to provide multiple diversified choices.

Due to the high spectral resolution and the similarity of some spectrums between different classes, hyperspectral image classification turns out to be an important but challenging task. Researches show the powerful ability of deep learning for hyperspectral image classification. However, the lack of training samples makes it difficult to extract discriminative features and achieve performance as expected. To solve the problem, a multi-scale CNN which can extract multi-scale features is designed for hyperspectral image classification. Furthermore, D-DSML, a diversified metric, is proposed to further improve the representational ability of deep methods. In this paper, a D-DSML-MSCNN method, which jointly learns deep multi-scale features and diversified metrics for hyperspectral image classification, is proposed to take both advantages of D-DSML and MSCNN. Experiments are conducted on Pavia University data to show the effectiveness of our method for hyperspectral image classification. The results show the advantage of our method when compared with other recent results.

With the rapid growth in quantity and quality of remote sensing images, extracting the useful information in them effectively and efficiently becomes feasible but also challenging. Convolutional neural network (CNN) is a suitable method to deal with such challenge since it can effectively represent and extract the information. However, the CNN can release their potentials only when enough labelled data provided for the learning procedure. This is a very time-consuming task and even infeasible for the applications with non-cooperative objects or scenes. Unsupervised CNN learning methods, which relieve the need for the labels in the training data, is a feasible solution for the problem. In this work, we investigate a real-world motivated sparsity based unsupervised deep CNN learning method. At first, the method formulates a balanced data driven population and lifetime sparsity prior and thus construct the unsupervised learning method through a layerwise mean. Then we further perform the method on the deep model with multiple CNN layers. Finally, the method is used for the remote sensing image representation and scenes classification. The experimental results over the public UC-Merced Land-use dataset demonstrate that the developed algorithm obtained satisfactory results compared with the recent methods.