细胞神经网络在重力异常分异中的研究及应用

介绍采用细胞神经网络CNN(cellular neural network)方法,对铬铁矿区内的矿体和围岩的重力异常进行分异。首先阐述了CNN方法的原理和算法,采用拟BP学习算法训练网络的权值,用全局误差函数求导方法推导权值的修正公式,讨论了如何根据目标异常训练适合该地质条件的网络的连接权值;其次将重力异常数据预处理,以达到适合CNN方法处理的数据格式和要求;最后由于该矿区内没有已知的重力数据作为网络训练的目标输出,根据相关地质图设置相应的地下构造模型。利用"点元"法分别正演出叠加异常和矿体异常,进而训练出适合全区的网络连接权值,实现了对全区重力异常的分异。应用结果表明,细胞神经网络方法较好地突出该矿区高异常和矿体的边界,只要选择了合适的网络连接权值,就能将横向叠加异常区分开,故CNN方法可以实现矿体和围岩的重力异常分异。

A method for gravity anomaly separation based on preferential continuation and its application

Based on the preferential continuation method proposed by Pawlowski （1995）, we propose a method and procedure for gravity anomaly separation with the preferential upward continuation operator in the case that the various sources are uncorrelated with one another and the continuation height is enough large. We also present a method for estimating optimum upward-continuation height, based on analyzing the characteristics of the preferential upward continuation operators of a synthesized gravity anomaly varying with different continuation heights. The method is tested on the raw Bouguer gravity data over an iron deposit. The result shows that the method separates the data into regional anomaly and residual anomaly efficiently and clearly.

Vector transformation and reduction to the pole for regional magnetic data with variable field direction

The vector transformation and pole reduction from the total-field anomaly are signifi cant for the interpretation.We examined these industry-standard processing procedures in the Fourier domain.We propose a novel iteration algorithm for regional magnetic anomalies transformations to derive the vertical-component data from the total-field measurements with the variation in the core-fi eld direction over the region.Additionally,we use the same algorithm to convert the calculated vertical-component data into the corresponding data at the pole and realize the processing of diff erential reduction to the pole(DRTP).Unlike Arkani-Hamed’s DRTP method,the two types of iterative algorithms have the same forms,and DRTP is realized by implementing this algorithm twice.The synthetic model’s calculation results show that the method has high accuracy,and the fi eld data processing confi rms its practicality.

Anomaly detection and segmentation based on multi-student teacher network

In automated industrial inspection,it is often necessary to train models on anomaly-free images and perform anomaly detection on products,which is also an important and challenging task in computer vision.The student-teacher network trains students to regress the output of the teacher,and uses the difference between the output of the student network and the pre-trained teacher network to locate anomalies,which has achieved advanced results in the field of abnormal segmentation.However,it is slow to predict a picture,and no anomaly detection is performed.A multi-student teacher network is proposed,which uses multiple student networks to jointly regress the output of the teacher network,and the minimum square difference between the output of students and teachers in each dimension is selected as the difference value.The information in the middle layer of the network is used to represent each area of the image and calculate the anomaly distance for anomaly segmentation,and the maximum abnormal score is used to represent the abnormal degree of the image for abnormal detection.Experiments results on MVTec anomaly detection show that the algorithm predicts a picture in 0.17 s and can output anomaly detection results at the same time,with image AUROC reaching 91.1%and Pixel AUROC reaching 94.5%.On the wall tile dataset produced by taking pictures of real scenes,image AUROC reached 89.7%,and Pixel AUROC reached 89.1%.Compared with the original student-teacher network,the proposed method can quickly complete anomaly segmentation and anomaly detection tasks at the same time with better accuracy,and it also has better results in real applications.

Associations between the Autumn Arctic Sea Ice and North American Winter Precipitation

Associations between the autumn Arctic sea ice concentrations (SICs) and North American winter precipitation were examined using singular value decomposition. The results show that a reduced SIC in the majority of the Arctic is accompanied by dry conditions over the Great Plains, the southern United States, Mexico, eastern Alaska, and southeastern Greenland, and by wet conditions over the majority of Canada, the northeastern United States, and the majority of Greenland. Atmospheric circulation anomalies associated with the SIC variability show a wave train structure that is persistent from autumn to winter and is responsible for the covariability between the autumn Arctic SICs and North American winter precipitation. This relationship suggests a potential long-term outlook for the North American winter precipitation.

Application of wavelet packet in gravity anomaly processing

Based on the advantages of the wavelet to separate regional field and local anomalies in MATLAB environment,a high-precision regional-residual separation was finally realized. Analytical continuation and trend surface analysis are conventional methods for gravity anomaly separation. But the wavelet packet analysis in analyzing gravity data can make the gravity anomaly to be computed at a higher precision. In this paper,wavelet packet method is used to process gravity anomaly data obtained in Laos,and the separation result is good. Daubechies wavelet series has a higher precision in the wavelet packet.

Progression from low-grade dysplasia to malignancy in patients with Barrett's esophagus diagnosed by two or more pathologists

AIM To evaluate annual incidence of low grade dysplasia(LGD) progression to high grade dysplasia(HGD) and/or esophageal adenocarcinoma(EAC) when diagnosis was made by two or more expert pathologists.METHODS Studies evaluating the progression of LGD to HGD or EAC were included. The diagnosis of LGD must be made by consensus of two or more expert gastrointestinal pathologists. Articles were searched in Medline, Pubmed, and Embase. Pooled proportions were calculated using fixed and random effects model. Heterogeneity among studies was assessed using the I2 statistic. RESULTS Initial search identified 721 reference articles, of which 53 were selected and reviewed. Twelve studies(n = 971) that met the inclusion criteria were included in this analysis. Among the total original LGD diagnoses in the included studies, only 37.49% reached the consensus LGD diagnosis after review by two or more expert pathologists. Total follow up period was 1532 patient-years. In the pooled consensus LGD patients, the annual incidence rate(AIR) of progression to HGD and or EAC was 10.35%(95%CI: 7.56-13.13) and progression to EAC was 5.18%(95%CI: 3.43-6.92). Among the patients down staged from original LGD diagnosis to No-dysplasia Barrett's esophagus, the AIR of progression to HGD and EAC was 0.65%(95%CI: 0.49-0.80). Among the patients down staged to Indefinite for dysplasia, the AIR of progression to HGD and EAC was 1.42%(95%CI: 1.19-1.65). In patients with consensus HGD diagnosis, the AIR of progression to EAC was 28.63%(95%CI: 13.98-43.27). CONCLUSION When LGD is diagnosed by consensus agreement of two or more expert pathologists, its progression towards malignancy seems to be at least three times the current estimates, however it could be up to 20 times the current estimates. Biopsies of all Barrett's esophagus patients with LGD should be reviewed by two expert gastroenterology pathologists. Follow-up strict surveillance programs should be in place for these patients.

VARIATION OF THE MONSOON TROUGH INTENSITY IN THE SOUTH CHINA SEA AND WESTERN NORTH PACIFIC OCEAN AND ITS RELATIONSHIP TO TROPICAL CYCLONE ACTIVITIES

Using the daily average of the NCEP/DOE AMIP-II reanalysis data from 1979 to 2005 and the characteristics of monsoon troughs in the western North Pacific,we established an intensity index and a location index to describe the activity of the monsoon troughs in three different regions and their impacts on tropical cyclones generated therein(MTTCs).The behavior of the monsoon troughs was analyzed.The following conclusions are obtained:(1)The established monsoon trough intensity index has a positive correlation to the location index,indicating that stronger monsoon trough intensity corresponds to more northward location.(2)Monsoon trough intensity exhibits significant interannual variation,with obvious periods of 4–5 years prior to 1994 and 2–3 years afterwards.(3)The affecting factors on monsoon trough intensity are different with areas.The preceding SST anomaly results in anomalous atmospheric circulation, leading to the anomaly of monsoon trough intensity in different areas.(4)The frequency of cyclogenesis and location anomalies of the MTTC are closely related to the intensity and location of the monsoon trough. Most of the anomalously less MTTC years coincide with the years with a weak general monsoon trough and weak regional monsoon troughs.The anomalously more MTTC years are associated with both a strong general monsoon trough and a weak general monsoon trough combined with a strong one over the South China Sea,though with a larger probability for the latter.(5)The interseasonal variation of the intensity of monsoon troughs provides favorable conditions for TC generation and development.The monsoon trough is in the active periods of both quasi-biweekly 10 to 20 day and 30 to 60 day oscillations,which is favorable for MTTC occurrence.

Characteristic Analysis of Anomalies of Strong Earthquakes along the Tancheng-Lujiang Fault Zone and Both Its Sides

This article made a systematic statistical analysis on the duration,spatial distribution,form,range,type and reliability of precursory anomalies based on the observations of precursors of 24 moderate-strong earthquakes occurring along the Tancheng-Lujiang fault zone and its bilateral areas in Shandong,Liaoning,Hebei,Jiangsu and Jiangxi Provinces in the period from 1966 to 2005. Results show that:( 1) For M7. 0 and M6. 0earthquakes,med-term abnormities are the greatest in number,short-term abnormities come second,and imminent anomalies are the least. For M5. 0 earthquakes,short-term abnormities are often the most in number,imminent anomalies come second,and the med-term ones are the least.( 2) Precursory anomalies are periodical,part of the med and short-term anomalies are shown as short-term and imminent anomalies during their development.( 3) The development of precursor shows quasi-synchronism; the closer to earthquake occurrence time,the more turning changes of anomalies will appear.( 4)Anomalies mainly occur within an epicenter distance of 100 km, showing a relative concentrated distribution in space,and the shorter the distance to epicenter,the higher the anomaly station / item rate.( 5) The spatial distribution of anomalies is relatively correlated to the strike of structures,having a good consistency with the structure.( 6)There is no significant correlation between duration of abnormities and the epicenter distance.( 7) The higher the magnitude of the earthquake,the greater the amplitude of the abnormities measured with the same observation approach.( 8) In terms of the itemsof the observations,fluid abnormities are the greatest in number,followed by deformation and electromagnetism observations.( 9) For all observation items,med and short-term abnormities make up the majority and short-term and imminent anomalies the minority.( 10) In fluid,deformation and electromagnetism observations,fairly reliable anomalies are about double the number of reliable anomalies.