Land cover classification with the focus on chlorophyll-rich vegetation detection plays an important role in urban growth monitoring and planning, autonomous navigation, drone mapping, biodiversity conservation, etc. Conventional approaches usually apply the normalized difference vegetation index (NDVI) for vegetation detection. In this paper, we investigate the performance of deep learning and conventional methods for vegetation detection. Two deep learning methods, DeepLabV3+ and our customized convolutional neural network (CNN) were evaluated with respect to their detection performance when training and testing datasets originated from different geographical sites with different image resolutions. A novel object-based vegetation detection approach, which utilizes NDVI, computer vision, and machine learning (ML) techniques, is also proposed. The vegetation detection methods were applied to high-resolution airborne color images which consist of RGB and near-infrared (NIR) bands. RGB color images alone were also used with the two deep learning methods to examine their detection performances without the NIR band. The detection performances of the deep learning methods with respect to the object-based detection approach are discussed and sample images from the datasets are used for demonstrations.

There is an emerging interest in using hyperspectral data for land cover classification. The motivation behind using hyperspectral data is the notion that increasing the number of narrowband spectral channels would provide richer spectral information and thus help improve the land cover classification performance. Although hyperspectral data with hundreds of channels provide detailed spectral signatures, the curse of dimensionality might lead to degradation in the land cover classification performance. Moreover, in some practical applications, hyperspectral data may not be available due to cost, data storage, or bandwidth issues, and RGB and near infrared (NIR) could be the only image bands available for land cover classification. Light detection and ranging (LiDAR) data is another type of data to assist land cover classification especially if the land covers of interest have different heights. In this paper, we examined the performance of two Convolutional Neural Network (CNN)-based deep learning algorithms for land cover classification using only four bands (RGB+NIR) and five bands (RGB+NIR+LiDAR), where these limited number of image bands were augmented using Extended Multi-attribute Profiles (EMAP). The deep learning algorithms were applied to a well-known dataset used in the 2013 IEEE Geoscience and Remote Sensing Society (GRSS) Data Fusion Contest. With EMAP augmentation, the two deep learning algorithms were observed to achieve better land cover classification performance using only four bands as compared to that using all 144 hyperspectral bands.

Online learning is crucial to robust visual object tracking as it can provide high discrimination power in the presence of background distractors. However, there are two contradictory factors affecting its successful deployment on the real visual tracking platform: the discrimination issue due to the challenges in vanilla gradient descent, which does not guarantee good convergence; the robustness issue due to over-fitting resulting from excessive update with limited memory size (the oldest samples are discarded). Despite many dedicated techniques proposed to somehow treat those issues, in this paper we take a new way to strike a compromise between them based on the recursive least-squares estimation (LSE) algorithm. After connecting each fully-connected layer with LSE separately via normal equations, we further propose an improved mini-batch stochastic gradient descent algorithm for fully-connected network learning with memory retention in a recursive fashion. This characteristic can spontaneously reduce the risk of over-fitting resulting from catastrophic forgetting in excessive online learning. Meanwhile, it can effectively improve convergence though the cost function is computed over all the training samples that the algorithm has ever seen. We realize this recursive LSE-aided online learning technique in the state-of-the-art RT-MDNet tracker, and the consistent improvements on four challenging benchmarks prove its efficiency without additional offline training and too much tedious work on parameter adjusting.

Rukban is a desert area crossing the border between Syria and Jordan, and thousands of Syrian refugees fled into this area since the Syrian civil war in 2014. In the past few years, the number of refugee shelters for the forcibly displaced Syrian refugees in this area has increased rapidly. Estimating the location and number of refugee tents has become a key factor to maintain the sustainability of the refugee shelter camps. Manually counting the shelters is labor-intensive and sometimes prohibitive given the large quantities. In addition, these shelters/tents are usually small in size, irregular in shape, and sparsely distributed in a very large area and could be easily missed by the traditional image-analysis techniques, making the image-based approaches also challenging. In this letter, we proposed a deep fully convolutional neural network (FCN) model to extract automatically the refugee shelters/tents in the worldview-2 (WV-2) satellite images. In addition, we transferred knowledge in the pretrained VGG-16 model to improve the detection accuracy and network training convergence. We compared the proposed approach with the traditional spectral angle mapper (SAM) method, deep convolutional neural network (CNN) models, and the mask Region-based CNN (R-CNN) model. The experimental results show that the FCN model improved the overall accuracy by 4.49%, 3.54%, and 0.88% compared with the CNNs, SAM, and mask R-CNN models, and improved the precision by 34.61%, 41.99%, and 11.87%, respectively.

Current transfer learning models study the source data for future target inferences within a major view, the whole source data should be used to explore the shared knowledge structure. However, human resources are constrained, the source domain data is collected as a whole in the real scene. However, this is not realistic, this data is associated with the target domain. A generalized empirical risk minimization model (GERM) is proposed in this article with discriminative knowledge-leverage (KL). The empirical risk minimization (ERM) principle is extended to the transfer learning setting. The theoretical upper bound of generalized ERM (GERM) is given for the practical discriminative transfer learning. The subset of the source domain data can be automatically selected in the model, and the source domain data is associated with the target domain. It can solve with only some knowledge of the source domain being available, thus it can avoid the negative transfer effect which is caused by the whole source domain dataset in the real scene. Simulation results show that the proposed algorithm is better than the traditional transfer learning algorithm in simulation data sets and real data sets.