Due to the instantaneous field-of-view (IFOV) of the sensor and diversity of land cover types, some pixels, usually named mixed pixels, contain more than one land cover type. Soft classification can predict the portion of each land cover type in mixed pixels in the absence of spatial distribution. The spatial distribution information in mixed pixels can be solved by super resolution mapping (SRM). Typically, SRM involves two steps: soft class value estimation, which is similar to the image super resolution of image restoration, and land cover allocation. A new SRM approach utilizes a deep image prior (DIP) strategy combined with a super resolution convolutional neural network (SRCNN) to estimate fine resolution fraction images for each land cover type; then, a simple and efficient classifier is used to allocate subpixel land cover types under the constraint of the generated fine fraction images. The proposed approach can use prior information of input images to update network parameters and no longer require training data. Experiments on three different cases demonstrate that the subpixel classification accuracy of the proposed DIP-based SRM approach is significantly better than the three conventional SRM approaches and a transfer learning-based neural network SRM approach. In addition, the DIP-SRM approach performs very robustly about small-area objects within multiple land cover types and significantly reduces soft classification uncertainty. The results of this paper provide an extension for utilizing SRCNN to address SRM issues in hyperspectral images.

A new subpixel mapping (SPM) method based on a super-resolution convolutional neural network (SRCNN) is proposed to generate subpixel land cover maps for hyperspectral images. The SRCNN is used to restore the image spatial resolution from a coarse input image, which is equivalent to interpolation. First, an efficient subpixel convolutional neural network, which is a state-of-the-art SRCNN, is utilized to calculate the subpixel soft class value via a transfer learning strategy. Then, a classifier is used to transform the subpixel soft class values to hard-classified land cover maps with the constraint of fraction images. Experiments on three different hyperspectral images demonstrate that the SPM accuracy of the proposed SRCNN-based method is significantly better than those of three traditional SPM methods. In addition, the SRCNN-based SPM method has a simplified calculation process, does not require training data, and is less time consuming. This article provides a new solution for SPM of hyperspectral images.

Multi-temporal Landsat-8 satellite images with fine spatial resolution (i.e., 30 m) are crucial for modern digital soil mapping (DSM). Generally, cloud-free images covering bare topsoil are common choices for DSM. However, the number of effective Landsat-8 data is greatly limited due to cloud contamination coupled with the coarse temporal resolution, and interference of material covering topsoil in most of the months, hindering the development of accurate DSM. To address this issue, temporally dense Landsat images were predicted using a spatio-temporal fusion method to improve DSM. Specifically, the recently developed virtual image pair-based spatio-temporal fusion (VIPSTF) method was adopted to produce simulated Landsat-8 time-series, by fusing with 500 m Moderate Resolution Imaging Spectroradiometer (MODIS) time-series with frequent observations. Subsequently, the simulated Landsat-8 data were used for distinguishing different soil classes via a random forest (RF) model. Training and validation samples of soil classes were collected from legacy soil data. Our results indicate that the simulated data were beneficial for improving DSM owing to the increase in class separability. More precisely, after combining the observed and simulated data, the overall accuracy (OA) and kappa coefficient (Kappa) were increased by 3.099% and 0.047, respectively. This research explored the potential of the spatio-temporal fusion method for DSM, providing a new solution for remote sensing-based DSM.

As computer performance continues to improve, new deep learning algorithms emerge in an endless stream. Object recognition (target detection) is one of the influential research directions in the field of computer vision. The traditional object recognition method has the following problems: 1. The generation of the target suggestion frame has a cumbersome effect on the detection speed, accuracy and redundancy; 2. Artificially extracting the image features cannot guarantee the quality of the feature; 3. Using the traditional machine learning method for feature classification is low; 4.slow detection speed and low accuracy. In this paper, the DenseNet structure is used to improve the recognition accuracy, and the SSD is used to improve the detection speed. At the same time, with the classification detection technology and the jump connection technology used in the network, the experiment shows that the target detection efficiency is further improved.