In recent years, deep learning (DL), a rebranding of neural networks (NNs), has risen to the top in numerous areas, namely computer vision (CV), speech recognition, and natural language processing. Whereas remote sensing (RS) possesses a number of unique challenges, primarily related to sensors and applications, inevitably RS draws from many of the same theories as CV, e.g., statistics, fusion, and machine learning, to name a few. This means that the RS community should not only be aware of advancements such as DL, but also be leading researchers in this area. Herein, we provide the most comprehensive survey of state-of-the-art RS DL research. We also review recent new developments in the DL field that can be used in DL for RS. Namely, we focus on theories, tools, and challenges for the RS community. Specifically, we focus on unsolved challenges and opportunities as they relate to (i) inadequate data sets, (ii) human-understandable solutions for modeling physical phenomena, (iii) big data, (iv) nontraditional heterogeneous data sources, (v) DL architectures and learning algorithms for spectral, spatial, and temporal data, (vi) transfer learning, (vii) an improved theoretical understanding of DL systems, (viii) high barriers to entry, and (ix) training and optimizing the DL.

Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community

There is a growing demand for detailed and accurate landslide maps and inventories around the globe, but particularly in hazard-prone regions such as the Himalayas. Most standard mapping methods require expert knowledge, supervision and fieldwork. In this study, we use optical data from the Rapid Eye satellite and topographic factors to analyze the potential of machine learning methods, i.e., artificial neural network (ANN), support vector machines (SVM) and random forest (RF), and different deep-learning convolution neural networks (CNNs) for landslide detection. We use two training zones and one test zone to independently evaluate the performance of different methods in the highly landslide-prone Rasuwa district in Nepal. Twenty different maps are created using ANN, SVM and RF and different CNN instantiations and are compared against the results of extensive fieldwork through a mean intersection-over-union (mIOU) and other common metrics. This accuracy assessment yields the best result of 78.26% mIOU for a small window size CNN, which uses spectral information only. The additional information from a 5 m digital elevation model helps to discriminate between human settlements and landslides but does not improve the overall classification accuracy. CNNs do not automatically outperform ANN, SVM and RF, although this is sometimes claimed. Rather, the performance of CNNs strongly depends on their design, i.e., layer depth, input window sizes and training strategies. Here, we conclude that the CNN method is still in its infancy as most researchers will either use predefined parameters in solutions like Google TensorFlow or will apply different settings in a trial-and-error manner. Nevertheless, deep-learning can improve landslide mapping in the future if the effects of the different designs are better understood, enough training samples exist, and the effects of augmentation strategies to artificially increase the number of existing samples are better understood.

/pdf/evaluation-of-different-machine-learning-methods-and-deep-5rqqhxf2cs.pdf

Evaluation of Different Machine Learning Methods and Deep-Learning Convolutional Neural Networks for Landslide Detection

Agricultural robots for field operations: Concepts and components

1 The goal of an N recommendation system is to accurately estimate the gap between the N provided by the soil and the N required by the plant. Accurately estimating this gap depends on the ability of the recommendation system to accurately estimate fi eld or subfi eld specifi c economically optimal nitrogen rates (EONR). Current recommendation systems are not as accurate as needed to provide consistently reliable estimates of N needs across years at the fi eld or subfi eld scale. Uncontrollable factors like temperature, rainfall timing, intensity and amount, and interactions of temperature and rainfall with factors such as N source, timing and placement, plant genetics, and soil characteristics combine to make N rate recommendations for an individual fi eld or rates for subfi elds a process guided as much by science as by the best professional judgement of farmers and farm advisors. Substantial evidence has accumulated that EONRs can vary widely across fi elds, within fi elds and over years in the same fi eld for a wide range of crops and geographies. Examples Strengths and Limitations of Nitrogen Rate Recommendations for Corn and Opportunities for Improvement

https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=1537&context=agron_pubs

Strengths and limitations of Nitrogen rate recommendations for corn and opportunities for improvement

In recent years, deep learning (DL), a re-branding of neural networks (NNs), has risen to the top in numerous areas, namely computer vision (CV), speech recognition, natural language processing, etc. Whereas remote sensing (RS) possesses a number of unique challenges, primarily related to sensors and applications, inevitably RS draws from many of the same theories as CV; e.g., statistics, fusion, and machine learning, to name a few. This means that the RS community should be aware of, if not at the leading edge of, of advancements like DL. Herein, we provide the most comprehensive survey of state-of-the-art RS DL research. We also review recent new developments in the DL field that can be used in DL for RS. Namely, we focus on theories, tools and challenges for the RS community. Specifically, we focus on unsolved challenges and opportunities as it relates to (i) inadequate data sets, (ii) human-understandable solutions for modelling physical phenomena, (iii) Big Data, (iv) non-traditional heterogeneous data sources, (v) DL architectures and learning algorithms for spectral, spatial and temporal data, (vi) transfer learning, (vii) an improved theoretical understanding of DL systems, (viii) high barriers to entry, and (ix) training and optimizing the DL.

A Comprehensive Survey of Deep Learning in Remote Sensing: Theories, Tools and Challenges for the Community

3 Soil properties and weather conditions are known to affect soil nitrogen (N) availability and 4 plant N uptake. However, studies examining N response as affected by soil and weather 5 sometimes give conflicting results. Meta-analysis is a statistical method for estimating treatment 6 effects in a series of experiments to explain the sources of heterogeneity. In this study, the 7 technique was used to examine the influence of soil and weather parameters on N responses of 8 corn (Zea mays L.) across 51 studies involving the same N rate treatments which were carried out 9 in a diversity of North American locations between 2006 and 2009. Results showed that corn 10 response to added N was significantly greater in fine-textured soils than in medium-textured 11 soils. Abundant and well-distributed rainfall and, to a lesser extent, accumulated corn heat units 12 enhanced N response. Corn yields increased by a factor of 1.6 (over the unfertilized control) in 13 medium-textured soils and 2.7 in fine-textured soils at high N rates. Subgroup analyses were 14 performed on the fine-textured soil class based on weather parameters. Rainfall patterns had an 15 important effect on N response in this soil texture class, with yields being increased 4.5-fold by 16 in-season N fertilization under conditions of “abundant and well-distributed rainfall.” These 17 findings could be useful for developing N fertilization algorithms that would allow for N 18 application at optimal rates taking into account rainfall pattern and soil texture, which would lead 19 to improved crop profitability and reduced environmental impacts. 20

Corn response to nitrogen is influenced by soil texture and weather

Extracting and identifying objects in very high resolution imagery has been a popular research topic in remote sensing. Since the beginning of this decade, deep learning techniques have revolutionized computer vision providing significant performance gains compared to traditional “shallow” techniques in various challenging vision problems. The training of deep neural networks usually requires very large training datasets. The advantage of using deep features is to exploit already trained Convolutional Neural Networks (CNN) in order to produce high level features without the burden of having to train a CNN from scratch. In this paper, we are investigating the use of deep features for the detection of small objects (cars and individual trees) in high resolution Pleiades imagery. Preliminary results show good detection performance and are very encouraging for future applications.

Object detection in pleiades images using deep features

A fuzzy Inference System (FIS) was developed to generate recommendations for spatially variable applications of nitrogen (N) fertilizer using soil, plant and precipitation information Experiments were conducted over three seasons (2005-2007) to assess the effects of soil electrical conductivity (ECa), nitrogen sufficiency index (NSI), and precipitations received in the vicinity of N fertilizers application, on response to N measured at mid-season growth Another experiment was conducted in 2010 to understand the effect of water supply (WS) on response to N, using a spatially variable irrigation set-up Better responses to N were observed in the case of high ECa, low NSI and high WS In the opposite cases (low ECa, high NSI or low WS), nitrogen fertilizer rates can be reduced Using fuzzy logic, expert knowledge was formalized as a set of rules involving ECa, NSI and cumulative precipitations to estimate economically optimal N rates (EONR)

Fuzzy logic approach for spatially variable nitrogen fertilization of corn based on soil, crop and precipitation information

This paper presents the software package REFLECT for the retrieval of ground reflectance from high and very-high resolution multispectral satellite images. The computation of atmospheric parameters is based on the 6S (Second Simulation of the Satellite Signal in the Solar Spectrum) routines. Aerosol optical properties are calculated using the OPAC (Optical Properties of Aerosols and Clouds) model, while aerosol optical depth is estimated using the dark target method. A new approach is proposed for adjacency effect correction. Topographic effects were also taken into account, and a new model was developed for forest canopies. Validation has shown that ground reflectance estimation with REFLECT is performed with an accuracy of approximately ±0.01 in reflectance units (for the visible, near-infrared, and mid-infrared spectral bands), even for surfaces with varying topography. The validation of the software was performed through many tests. These tests involve the correction of the effects that are associated with sensor calibration, irradiance, and viewing conditions, atmospheric conditions (aerosol optical depth AOD and water vapour), adjacency, and topographic conditions.

Ground Reflectance Retrieval on Horizontal and Inclined Terrains Using the Software Package REFLECT

Information-based crop management, such as variable rate technology, allows for changing rates of fertilization according to local needs. Fertilizer prescription maps can be derived from crop growth status assessed through proximal canopy sensing technologies. Typically, a vegetation index, such as the NDVI, is used to estimate a Nitrogen Sufficiency Index (NSI=NDVI/NDVIN-rich). NSI is related to N intake in comparison to the measurements in N-rich reference areas created artificially with high N rates applied at planting. This approach requires N-rich areas representative of different field conditions, which is not always practical. To circumvent this limitation `naturally N-rich' reference areas linked to natural field characteristics can be identified. The goal of this present work was to show that the NSINLR estimated using less laborious Natural Local References (NLR) has a benefit comparable to the more traditional NSIN-rich from N-rich plots. Experiments were performed on corn fields located in the Monteregie area of Quebec. The NDVI values were measured using 5 tractor-mounted GreenSeeker sensors. The spatial variability of the soil was analyzed using maps of apparent electrical conductivity (ECa) obtained using a Veris 3100 system. The NSINLR, set at the 90th percentile of the NDVI values obtained around each sensed area, was compared to NSIN-rich. The NSI values estimated using both techniques were strongly correlated and demonstrated a high capacity to estimate the optimal N rate for corn.

Yacine Bouroubi

Papers

Corn response to nitrogen is influenced by soil texture and weather

Object detection in pleiades images using deep features

Fuzzy logic approach for spatially variable nitrogen fertilization of corn based on soil, crop and precipitation information

Ground Reflectance Retrieval on Horizontal and Inclined Terrains Using the Software Package REFLECT

Estimating Nitrogen Sufficiency Index using a Natural Local Reference approach