Use of Noise to Augment Training Data: A Neural Network Method of Mineral–Potential Mapping in Regions of Limited Known Deposit Examples

One of the main factors that affects the performance of MLP neural networks trained using the backpropagation algorithm in mineral-potential mapping isthe paucity of deposit relative to barren training patterns. To overcome this problem, random noise is added to the original training patterns in order to create additional synthetic deposit training data. Experiments on the effect of the number of deposits available for training in the Kalgoorlie Terrane orogenic gold province show that both the classification performance of a trained network and the quality of the resultant prospectivity map increasesignificantly with increased numbers of deposit patterns. Experiments are conducted to determine the optimum amount of noise using both uniform and normally distributed random noise. Through the addition of noise to the original deposit training data, the number of deposit training patterns is increased from approximately 50 to 1000. The percentage of correct classifications significantly improves for the independent test set as well as for deposit patterns in the test set. For example, using ±40% uniform random noise, the test-set classification performance increases from 67.9% and 68.0% to 72.8% and 77.1% (for test-set overall and test-set deposit patterns, respectively). Indices for the quality of the resultant prospectivity map, (i.e. D/A, D × (D/A), where D is the percentage of deposits and A is the percentage of the total area for the highest prospectivity map-class, and area under an ROC curve) also increase from 8.2, 105, 0.79 to 17.9, 226, 0.87, respectively. Increasing the size of the training-stop data set results in a further increase in classification performance to 73.5%, 77.4%, 14.7, 296, 0.87 for test-set overall and test-set deposit patterns, D/A, D × (D/A), and area under the ROC curve, respectively.     

Use of Fuzzy Membership Input Layers to Combine Subjective Geological Knowledge and Empirical Data in a Neural Network Method for Mineral-Potential Mapping

Use of GIS layers, in which the cell values represent fuzzy membership variables, is an effective method of combining subjective geological knowledge with empirical data in a neural network approach to mineral-prospectivity mapping. In this study, multilayer perceptron (MLP), neural networks are used to combine up to 17 regional exploration variables to predict the potential for orogenic gold deposits in the form of prospectivity maps in the Archean Kalgoorlie Terrane of Western Australia. Two types of fuzzy membership layers are used. In the first type of layer, the statistical relationships between known gold deposits and variables in the GIS thematic layer are used to determine fuzzy membership values. For example, GIS layers depicting solid geology and rock-type combinations of categorical data at the nearest lithological boundary for each cell are converted to fuzzy membership layers representing favorable lithologies and favorable lithological boundaries, respectively. This type of fuzzy-membership input is a useful alternative to the 1-of-N coding used for categorical inputs, particularly if there are a large number of classes. Rheological contrast at lithological boundaries is modeled using a second type of fuzzy membership layer, in which the assignment of fuzzy membership value, although based on geological field data, is subjective. The methods used here could be applied to a large range of subjective data (e.g., favorability of tectonic environment, host stratigraphy, or reactivation along major faults) currently used in regional exploration programs, but which normally would not be included as inputs in an empirical neural network approach.     

长江三峡狮子口重力滑动构造

围绕着长江三峡库处区狮子口ＮＮＷ向线性影像带的构造性及其地壳稳定性评价问题进行了专题研究，以区域地质调查为基础，运用构造解析方法，采用多学科，多手段的综合研究，确认该影像带影像特征最明显的地段乃是一个滑褶型多层式得力滑动构造，本文还分析了该系统形成的介质条件，斜坡构造及动力环境。     

Fault geometry and mechanics of marly carbonate multilayers: An integrated field and laboratory study from the Northern Apennines,Italy

Sealing layers are often represented by sedimentary sequences characterized by alternating strong and weak lithologies. When involved in faulting processes, these mechanically heterogeneous multilayers develop complex fault geometries. Here we investigate fault initiation and evolution within a mechanical multilayer by integrating field observations and rock deformation experiments. Faults initiate with a staircase trajectory that partially reflects the mechanical properties of the involved lithologies, as suggested by our deformation experiments. However, some faults initiating at low angles in calcite-rich layers (θ_i = 5°–20°) and at high angles in clay-rich layers (θ_i = 45°–86°) indicate the important role of structural inheritance at the onset of faulting. With increasing displacement, faults develop well-organized fault cores characterized by a marly, foliated matrix embedding fragments of limestone. The angles of fault reactivation, which concentrate between 30° and 60°, are consistent with the low friction coefficient measured during our experiments on marls (μ_s = 0.39), indicating that clay minerals exert a main control on fault mechanics. Moreover, our integrated analysis suggests that fracturing and faulting are the main mechanisms allowing fluid circulation within the low-permeability multilayer, and that its sealing integrity can be compromised only by the activity of larger faults cutting across its entire thickness.     

基于深度学习的海温观测数据质量控制应用研究

基于国家海洋信息中心质量控制后的西太平洋10度方区约100万站次温盐实测历史调查资料,对经过26种严格质量控制方法的综合海温质量符进行分类分析,首次将深度学习技术应用于海洋数据质量控制多分类(multiclass classification)算法与应用研究.通过人工合成少数类样本和加权损失函数方法减少多数类的频率来降...     

Machine learning for pore-water pressure time-series prediction: Application of recurrent neural networks

Knowledge of pore-water pressure(PWP)variation is fundamental for slope stability.A precise prediction of PWP is difficult due to complex physical mechanisms and in situ natural variability.To explore the applicability and advantages of recurrent neural networks(RNNs)on PWP prediction,three variants of RNNs,i.e.,standard RNN,long short-term memory(LSTM)and gated recurrent unit(GRU)are adopted and compared with a traditional static artificial neural network(ANN),i.e.,multi-layer perceptron(MLP).Measurements of rainfall and PWP of representative piezometers from a fully instrumented natural slope in Hong Kong are used to establish the prediction models.The coefficient of determination(R^2)and root mean square error(RMSE)are used for model evaluations.The influence of input time series length on the model performance is investigated.The results reveal that MLP can provide acceptable performance but is not robust.The uncertainty bounds of RMSE of the MLP model range from 0.24 kPa to 1.12 k Pa for the selected two piezometers.The standard RNN can perform better but the robustness is slightly affected when there are significant time lags between PWP changes and rainfall.The GRU and LSTM models can provide more precise and robust predictions than the standard RNN.The effects of the hidden layer structure and the dropout technique are investigated.The single-layer GRU is accurate enough for PWP prediction,whereas a double-layer GRU brings extra time cost with little accuracy improvement.The dropout technique is essential to overfitting prevention and improvement of accuracy.     

Simulation of landscape spatial layout evolution in rural-urban fringe areas: a case study of Ganjingzi District

In recent years, the rapid expansion of urban spaces has accelerated the mutual evolution of landscape types. Analyzing and simulating spatio-temporal dynamic features of urban landscape can help to reveal its driving mechanisms and facilitate reasonable planning of urban land resources. The purpose of this study was to design a hybrid cellular automata model to simulate dynamic change in urban landscapes. The model consists of four parts: a geospatial partition, a Markov chain (MC), a multi-layer perceptron artificial neural network (MLP-ANN), and cellular automata (CA). This study employed multivariate land use data for the period 2000–2015 to conduct spatial clustering for the Ganjingzi District and to simulate landscape status evolution via a divisional composite cellular automaton model. During the period of 2000–2015, construction land and forest land areas in Ganjingzi District increased by 19.43% and 15.19%, respectively, whereas farmland, garden lands, and other land areas decreased by 43.42%, 52.14%, and 75.97%, respectively. Land use conversion potentials in different sub-regions show different characteristics in space. The overall land-change prediction accuracy for the subarea-composite model is 3% higher than that of the non-partitioned model, and misses are reduced by 3.1%. Therefore, by integrating geospatial zoning and the MLP-ANN hybrid method, the land type conversion rules of different zonings can be obtained, allowing for more effective simulations of future urban land use change. The hybrid cellular automata model developed here will provide a reference for urban planning and policy formulation.     

SAR image despeckling with a multilayer perceptron neural network

Speckle noise in synthetic-aperture radar (SAR) images severely hinders remote sensing applications; therefore, the appropriate removal of speckle noise is crucial. This paper elaborates on the multilayer perceptron (MLP) neural-network model for SAR image despeckling by using a time series of SAR images. Unlike other filtering methods that use only a single radar intensity image to derive their parameters and filter that single image, this method can be trained using archived images over an area of interest to self-learn the intensity characteristics of image patches and then adaptively determine the weights and thresholds by using a neural network for image despeckling. Several hidden layers are designed for feedforward network training, and back-propagation stochastic gradient descent is adopted to reduce the error between the target output and neural-network output. The parameters in the network are automatically updated in the training process. The greatest advantage of MLP is that once the despeckling parameters are determined, they can be used to process not only new images in the same area but also images in completely different locations. Tests with images from TerraSAR-X in selected areas indicated that MLP shows satisfactory performance with respect to noise reduction and edge preservation. The overall image quality obtained using MLP was markedly higher than that obtained using numerous other filters. In comparison with other recently developed filters, this method yields a slightly higher image quality, and it demonstrates the powerful capabilities of computer learning using SAR images, which indicate the promising prospect of applying MLP to SAR image despeckling.     

卫星遥感接收数据的管理与软件开发

卫星遥感接收数据是遥感影像应用的重要数据来源,因此对特殊格式的海量接收数据的存储和管理成为卫星遥感应用的重要环节。采用多层存储模式对数据分类、分级存储,建立空间数据库管理空间数据和属性数据。基于ArcEngine开发技术建立卫星接收数据管理平台,实现空间数据的快速查询和属性数据的自动入库等高效管理功能。     

用多层感知器模型由吸收光谱反演浮游植物色素

浮游植物吸收光谱已逐渐成为高光谱水色遥感的可获取参量。文章采用了多层感知器模型, 由珠江口担杆群岛附近水体的浮游植物吸收光谱进行了色素浓度的反演, 感知器的输入量是浮游植物吸收光谱, 输出量分别对应叶绿素a、叶绿素b、叶绿素c、光保护类胡萝卜素和非光保护类胡萝卜素五大类主要色素的浓度。分析结果表明, 叶绿素a和叶绿素c估算结果的平均相对偏差比较低, 在测试数据集中两者的偏差分别为19.06%和15.90%; 光保护类胡萝卜素和非光保护类胡萝卜素的估算浓度的相对偏差比较高, 对于测试数据而言, 分别为37.62%和36.96%; 叶绿素b浓度在测试数据集中的估算相对偏差约为27.47%。五大类色素在测试数据集和训练数据集的估算偏差比较接近, 已训练好的多层感知器可用于担杆岛水体中色素信息的反演。同时, 此色素反演方法也为遥感监测水体浮游植物种群动态提供了重要的手段。