With the development of earth observation technologies, the acquired remote sensing images are increasing dramatically, and a new era of big data in remote sensing is coming. How to effectively mine these massive volumes of remote sensing data are new challenges. Deep learning provides a new approach for analyzing these remote sensing data. As one of the deep learning models, convolutional neural networks (CNNs) can directly extract features from massive amounts of imagery data and is good at exploiting semantic features of imagery data. CNNs have achieved remarkable success in computer vision. In recent years, quite a few researchers have studied remote sensing image classification using CNNs, and CNNs can be applied to realize rapid, economical and accurate analysis and feature extraction from remote sensing data. This paper aims to provide a survey of the current state-of-the-art application of CNN-based deep learning in remote sensing image classification. We first briefly introduce the principles and characteristics of CNNs. We then survey developments and structural improvements on CNN models that make CNNs more suitable for remote sensing image classification, available datasets for remote sensing image classification, and data augmentation techniques. Then, three typical CNN application cases in remote sensing image classification: scene classification, object detection and object segmentation are presented. We also discuss the problems and challenges of CNN-based remote sensing image classification, and propose corresponding measures and suggestions. We hope that the survey can facilitate the advancement of remote sensing image classification research and help remote-sensing scientists to tackle classification tasks with the state-of-art deep learning algorithms and techniques.

Habitat differentiation between polyploid and diploid plants are frequently observed, with polyploids usually occupying more stressed environments. In woody plants, polyploidization can greatly affect wood characteristics but knowledge of its influences on xylem hydraulics is scarce. The four Betula species in NE China, representing two diploids and two polyploids with obvious habitat differentiation, provide an exceptional study system for investigating the impact of polyploidization on environmental adaptation of trees from the point view of xylem hydraulics. To test the hypothesis that changes in hydraulic architecture play an important role in determining their niche differentiation, we measured wood structural traits at both the tissue and pit levels and quantified xylem water transport efficiency and safety in these species. The two polyploids had significantly larger hydraulic weighted mean vessel diameters than the two diploids (45.1 and 45.5 vs 25.9 and 24.5 μm) although the polyploids are occupying more stressed environments. As indicated by more negative water potentials corresponding to 50% loss of stem hydraulic conductivities, the two polyploids exhibited significantly higher resistance to drought-induced embolism than the two diploids (-5.23 and -5.05 vs -3.86 and -3.13 MPa) despite their larger vessel diameters. This seeming discrepancy is reconciled by distinct characteristics favoring greater embolism resistance at the pit level in the two polyploid species. Our results showed clearly that the two polyploid species have remarkably different pit-level anatomical traits favoring greater hydraulic safety than their congeneric diploid species, which have likely contributed to the abundance of polyploid birches in more stressed habitats; however, less porous inter-conduit pits together with a reduced leaf to sapwood area may have compromised their competitiveness under more favorable conditions. Contrasts in hydraulic architecture between diploid and polyploid Betula species suggest an important functional basis for their clear habitat differentiation along environmental gradients in Changbai Mountain of NE China.

Nitrogen (N) deposition is expected to have great impact on forest ecosystems by affecting many aspects of plant-environmental interactions, one of which involves its influences on plant water relations through modifications of plant hydraulic architecture. However, there is a surprising lack of integrative study on tree hydraulic architecture responses to N deposition, especially at the whole-plant level. In the present study, we used a 5-year N addition experiment to simulate the effects of six different levels of N deposition (20-120 kg ha(-1) year(-1)) on growth and whole-plant hydraulic conductance of a dominant tree species (Fraxinus mandshurica Rupr.) from the typical temperate forest of NE China. The results showed that alleviation of N limitation by moderate concentrations of fertilization (20-80 kg ha(-1) year(-1)) promoted plant growth, but further N additions on top of the threshold level showed negative effects on plant growth. Growth responses of F. mandshurica seedlings to N addition of different concentrations were accompanied by corresponding changes in whole-plant hydraulic conductance; higher growth rate was accompanied by reduced whole-plant hydraulic conductance (Kplant) and higher leaf water-use efficiency. A detailed analysis on hydraulic conductance of different components of the whole-plant water transport pathway revealed that changes in root and leaf hydraulic conductance, rather than that of the stem, were responsible for Kplant responses to N fertilization. Both plant growth and hydraulic architecture responses to increasing levels of N addition were not linear, i.e., the correlation between measured parameters and N availability exhibited bell-shaped curves with peak values observed at medium levels of N fertilization. Changes in hydraulic architecture in response to fertilization found in the present study may represent an important underlying mechanism for the commonly observed changes in water-related tree performances in response to N deposition.