In recent years, with the rapid development of deep learning, especially convolutional neural network technology, deep learning technology has been widely used in various image classification and recognition, including the detection and recognition of pulmonary nodules. However, due to the limitation of computer capability, the input image size of convolutional neural network for deep learning is usually fixed, and the image size is usually small. However, the size of CT images used to detect pulmonary nodules is 512×512 and the sample data for training convolutional neural network is relatively small. It is usually difficult for ordinary convolution neural networks to detect directly. As a special full convolution network, U-net can be used directly for large-scale image detection by replacing the full connection layer with the convolution layer. And U-net is suitable for small sample medical image detection. Therefore, the detection of pulmonary nodules with U-net has been tried in this paper. The detection and recognition of pulmonary nodules is divided into two steps. Firstly, a U-net network is used to detect and segment suspicious pulmonary nodules. Secondly, in order to improve the accuracy of pulmonary nodules detection and recognition, a convolution neural network(CNN) is used to classify and identify the pulmonary nodules segmented by U-net detection, so as to remove the false positive pulmonary nodules. The experimental results show that the pulmonary nodules detection and recognition method combined with U-net and CNN is not only suitable for small sample and large size chest CT image detection, but also can remove a large number of false positive pulmonary nodules, effectively improve the accuracy of pulmonary detection, which is of great significance for the classification and recognition of pulmonary nodules using deep learning technology.

In order to assist the doctors to diagnose lung cancer and improve the classification accuracy of benign and malignant pulmonary nodules, this paper proposes a novel intelligent diagnosis model which is aiming at CT imaging features of pulmonary nodules. Specifically, this model uses the convolutional neural network to extract the features of the pulmonary nodules, then uses the principal component analysis to reduce the dimension of the extracted features, and finally classifies the final features with particle swarm optimization optimized SVM. With regard to the pulmonary nodules extracted from the LIDC-IDRI database, 400 pulmonary nodules are used for training and 310 pulmonary nodules are used for testing, the classification accuracy rate is 91.94%. This model can provide objective, convenient and efficient auxiliary method for solving the classification problem of benign and malignant pulmonary nodules in medical images.

For the past several decades, offline handwritten character recognition is widely and deeply studied. The requirements of the identification results are constantly improving in practical applications. However, the recognition rates of the similar handwritten Chinese characters are not very high in different writing style, writing environment and writing mode. We propose a method of combining deep convolution neural network and support vector machine together. Using the deep convolution neural network to learn and extract Chinese characters features automatically, and then the extracted features are classified and identified by the support vector machine. Experiments show that the deep convolution neural network can extract the features effectively, which avoided the shortage of artificial feature extraction, then using the support vector machine to classify and identify that, the accuracy rate is further improved.

The requirement of the recognition result is also increasing in practical applications .In this paper , a new classifier cascade recognition model is proposed for the problem of offline handwritten Chinese character recognition.New model is the fusion of modified quadratic discriminant function ( MQDF) and deep belief network ( DBN).First to recognize and get result using MQDF , and calculate the reliability of the recognition result .If the reliability can meet the requirement , MQDF recognition result can be as the final result directly output .Otherwise using the DBN to make recognition again and getting the final recognition result .Experiments show that the MQDF and DBN fusion model proposed in this paper can achieve better accuracy than the single use of MQDF and DBN in the offline handwritten Chinese character recognition task , which is performed on the ETL-9 B handwritten Chinese character dataset.

For the past several decades, offline handwritten character recognition is widely and deeply studied. However, the requirements of the identification results are constantly improving in practical applications. This paper presents a new classifier cascaded model to improve the accuracy of offline handwritten Chinese character recognition. New model is the fusion of modified quadratic discriminant function (MQDF) and deep belief network (DBN). The main idea behind MQDF-DBN fusion model is that the significant difference on features and classification mechanisms between MQDF and DBN can complete each other. First to recognize and get result using MQDF, calculate the recognition confidence as evaluation criteria. If recognition confidence is high, the recognition result of MQDF will be output directly. Otherwise, using the DBN to make recognition again and getting the final recognition result. Experiment shows that the fusion model of MQDF and DBN proposed in this paper has achieved better accuracy than the single use of MQDF and DBN in the offline handwritten Chinese character recognition task, which is performed on the ETL-9B handwritten Chinese character dataset.

In order to make embedded and mobile products become more energy-efficient and more environmentally friendly, this paper proposed that adjusting loop structure in program and enhancing program locality could optimize device’s memory-accessing operation,accordingly decreased storage system’s and embedded deviced’s power dissipation.These methods were applied to the breathing sounds dynamic analysis and respiratory function monitoring system in anesthesia surgery developed by the laboratory,then on the energy consumption simulator HMSim,four groups of experiments were made.The results show that every group’s power consumption decreases in different degree.According to the calculation of equivalent method,the largest ratio of energy saving can achieve up to 25 .84%.All these indicate that the methods described by this paper are feasible and effective for reducing device’s power consumption.

Spectrum analysis can quickly extract signal features, and is widely used in digital signal processing (DSP). Fast Fourier Transform (FFT) is chosen in the application for audio signal frequency spectrum analysis. The paper is based on FFT, aiming at audio signal acquisition, do optimization of the reverse position and butterfly algorithm, and then utilize the optimized algorithm to do frequency spectrum analysis about the audio signal in different formats, then realize the dynamic effect of mixed colors of lights. In the concrete realization of the system, through the micro controller STC12C5A60S2 to read a variety of digital music signal in SD card or U disk and other physical media by using the SPI protocol, then use the audio decoder VS1003 to covert digital signal to analog signal for playing. when playing music, use the A/D converter to do analog signal acquisition, then use the optimized FFT algorithm to analyses and extract the features, then through the micro controller and PWM technology to produce different duty ratio of the pulse signal, and then drive LED in a variety of color to realize dynamic mixed light.

GeoComputation is an important advanced computing paradigm in geographic information science.It is an indispensable part of the next generation of GIS(Geographic Information System).GeoComputation involves lots of processes such as data processing,spatial analysis,process simulation,etc.The process of GeoComputation is so complicated that usually consumes large amount of time.HPC(High Performance Computing) provides a powerful tool to solve the complex GeoComputation problems.To cope with the parallel implementation issues of the GeoComputation service process in HPC environment,a modeling method of the GeoComputation service process and a strategy for processing concurrent processes based on multithread were proposed.A prototype system of executing engine for the GeoComputation service process in HPC environment was designed and implemented,which can automatically resolve and dispatch the GeoComputation service process and keep the load balanced in concurrent conditions.The experimental results prove that the proposed method can effectively model GeoComputation service process with good practical performance,which can be widely used in geospatial data application fields.

Pay less attention to the behavior of users in the past and historical data for existing retrieval system problems,the author presents a geographical computing applications retrieval method based on user preferences.Under the background of geographic computing application retrieval,based on user preference model in internet information retrieval,analyze user searches for geographic computing applications history,and give full consideration to the impact of the last access time of geographical computing applications to determine user preferences,then extract user preference factor matrix and the time factor matrix,and calculate the most relevant geographic computing applications to the user's query.The experiment result shows that this method can availably improve the user geographical computing application retrieval efficiency,and provide strong support for the optimization of geographical computing application retrieval.

Solving traditional spatial analysis problems benefits from high performance geo-computation powered by parallel computing. Digital Terrain Analysis (DTA) is a typical example of data and computationally intensive spatial analysis problems and can be improved by parallelization technologies. Previous work on this topic has mainly focused on applying optimization schemes for specific DTA case studies. The task addressed in this paper, in contrast, is to find optimization methods that are generally applicable to the parallelization of DTA. By modeling a complex DTA problem with Cellular Automata (CA), we developed a temporal model that can describe the time cost of the solution. Three methods for optimizing different components in the temporal model are proposed: (1) a parallel loading/writing method that can improve the IO efficiency; (2) a best cell division method that can minimize the communication time among processes; and (3) a communication evolution overlapping method that can reduce the total time of evolutions and communications. The feasibilities and practical efficiencies of the proposed methods have been verified by comparative experiments conducted on an elevation dataset from North America using the Slope of Aspect (SOA) as an example of a general DTA problem. The results showed that the parallel performance of the SOA can be improved by applying the proposed methods individually or in an integrated fashion.

Taking account of the influence of different features on the grouping accuracy,the charactesistic selection and optimal experiment were performed by adopting the logistic regression analysis method,and the grouping comparison of the common peripheral lung cancer was carried out by methods of neural network and support vector machine.During the experiments,the application of both the neural network and the support vector Machine methods was adopted,and also the two methods in the application of lung cancer grouping were compared.The experimental results prove that under condition of small sample,the support vector machine method has a stronger generalizability than the neural network method.

In order to solve the difficulty of determining the growth rules in conventional regional growth algorithm and the slowly of support vector machine segmentation algorithm,an image segmentation method combined support vector machine and regional growth was proposed.Firstly,selected a certain numbers of sample point from target area and non-target area and trained the support vector machine classification,then used the trained classification search seed point and regional growing,the support vector machine classification was used as growth rules,the last,some necessary retrogressing were used for the edge and noise.The experimental results show that this algorithm is feasible and it performs better than conventional region growth segmentation algorithm and faster then conventional support vector machine segmentation algorithm.

LonWorks technology is introduced.The function analysis of intelligent home network is carried out.A hardware and software design to control the home nodes is presented.The remote reading of the meters,security,home control and other functions are achieved by the technology.A SCM is adopted in the core part of the home controller as a master processor to perform the monitoring and control.FT3120 smart transceiver solidified ShortStack firmware becomes ShortStack micro-processor taken as a slave processor to complete the communication function.

In order to solve the low accuracy diagnosis of metastases and non-metastases tumid lymph nodes in the lung cancer N stage with chest CT images,effective image features of lymph nodes need to be found for quickly and accurately differentiating metastases and non-metastases tumid lymph nodes.First,tumid lymph nodes are extracted from chest CT images using interactive segmentation.Second,the multi-resolution histograms of tumid lymph nodes are directly calculated to receive a high-dimensional features sample set with spatial information.Then the classifier for differentiating metastases and non-metastases tumid lymph nodes is constructed with making full use the advantage of SVM which is good at dealing with high dimensional data sets.Finally,the performance of classification is evaluated by testing the trained SVM with the test sample set.The test results by 96 cases show that it takes 1.91 s for computing 200 dimensional features of 100 lymph nodes,1.36 s for training and testing the SVM classifier.Receiver Operating Characteristic(ROC) analysis of the classification performance shows that the sensitivity is 76%,specificity is 64%,accuracy is 70%,and the Area Under Curve(AUC) is nearly 0.6525.Image spatial information can effectively express the characteristics of lymph nodes,the classification accuracy of metastases and non-metastases tumid lymph nodes is up to 70% without medical signs,and the classification speed is about 10 times than traditional texture methods.It provides a feasible,simple,objective method for improving the accuracy of the lung cancer N stage in medical imaging diagnosis.

In order to improve the accuracy of the diagnosis with medical signs in medical imaging diagnostics,a computer-aided diagnosis system is developed quickly and accurately to find out the differentce of benign and malignant Solitary Pulmonary Nodwles(SPNs) in chest CT images based the image features of SPNs. Firstly,SPNs are extracted from chest CT images using the interactive segmentation,and the multi-resolution histograms of SPNs are directly calculated to receive a high-dimensional feature sample set with spatial information of SPNs.Then,the classifier for differentiating benign and malignant SPN is constructed by using a Support Vector Mechine(SVM). Finally,the performance of classification is evaluated by testing the trained SVM with the test sample set. The test results of 214 cases show that it takes 4.83 s for computing 768 dimensional features of 240 SPNs and 2.24 s for training and testing the SVM classifier. The receiver operating characteristic (ROC) analysis of classification performance of the proposed approach shows that the sensitivity is 73.33%,specificity is 70%,accuracy is 71.67%,and the area under curve (AUC) is nearly 0.786 4. Obtained results show that the image spatial information can effectively express the characteristics of SPNs. The system classification accuracy of benign and malignant SPNs is up to 71.67% without medical signs,and the classification speed is about 50 times faster than that of traditional texture methods. It provides a feasible,simple and objective method for solving the problem in medical imaging diagnosis of the SPNs.

In this paper,a linear resistance nerwork analysis method based on optimization technology is proposed.Current distribution should minimize the total energy loss of network system is utilized to define the optimum target function,and Kirchhoff's Current Law(KCL) is utilized to define the constraint condition.Then the linear resistive network,as an example,is calculated.

针对工控网络的结构特点,提出了一种具有透明传输和简单并联特征的专用调制解调器(MODEM).其采用多进制频移键控(MFSK)的调制方法,在电话线路通频带内,无需数据压缩即可获得18.9 kbps的平均传输速率.系统以80C196KB CPU为核心,数据端口采用三线制(TXD,RXD,GND)最简方式,无需控制信号事先建立载波,实现了主从式工控网络挂灯笼连接方式,可满足1.0 km以上工控系统交换数据的需要.

The system can measure the concentrations of phosphate,nitrite, nitrate,and ammonia.The auto-analysis system communicates based on field bus,employs programmable logic controller (PLC) to directly control the analysis devices, and is managed by the rack PC with embedded Windows CE operation system (OS) and Kingview Embedded 6.0 configuration software.

In this paper, the principle of a NAVTEX receiver and the character of its signal are introduced. The idea of using DSP to design new-typed NAVTEX receiver and using software radio algorithm to make the receiver software are utilized. The capability and function of this receiver are improved by making use of DSP to realize digital filter demodulation and self-detected technology. In the aspect of system functions, its parameter can be adjusted easily and also can be updated easily.

本文介绍了一种基于嵌入式Windows CE操作系统和PLC、组态王6.0组态软件的车辆尾气自动检测系统.中央控制单元以运行Windows CE操作系统和组态王6.0组态软件的工控机为控制站,PLC作为现场控制单元,传感器获取的测量数据由数据采集模块传送给工控机,实现对污染物检测装置的中央控制.