为实现对灰度不均匀脑核磁共振(MR)图像分割的同时进行有偏场估计并校正,提出一种基于局部高斯分布拟合(LGDF)模型的多相水平集方法.通过分析图像有偏场模型的局部特性,将有偏场乘性因子引入到图像局部灰度均值的表达中,从而使有偏场乘性因子成为新的能量函数的变量.能量函数的迭代最小化既实现了目标组织分割,又有效估计了有偏场.合成图像和仿真脑MR图像实验结果表明,本文方法比现有多种方法分割性能更好,且利用本文方法估计的有偏场校正后的图像有更好的视觉效果.

With the broad usage of diffusion tensor imaging (DTI) modality in clinical medical treatment, the DT image segmentation has become a research focus on medical image processing and analysis at home and abroad. In this paper, we reviewed various segmentation methods of DT images in recent years, and mainly investigated the state of the art and recent advances of those based on clustering, graph cuts and level set. Moreover, we discussed the computing procedure of each typical algorithm respectively, analyzed and compared segmentation objects, advantages, disadvantages and similarity metrics of these methods qualitatively. At the end, after summarizing main characteristics of existing methods, we prospected future development trend on DT image segmentation.

An improved adaptive genetic algorithm is proposed based on the principle of hormone modulation in endocrine system.An adaptive crossover operator and an adaptive mutation operator based on the downward form of Hill function are designed in the algorithm.The crossover rate and mutation rate are made self-regulated according to the standard deviation of fitness value in each generation.And the diversity is maintained at a reasonable level in the whole process of evolution to ensure the normal evolution of genetic algorithm.Experimental results of four test functions and 3D brain image segmentation show the improved genetic algorithm can maintain the diversity of the population effectively,and overcome the premature problem. The performances of the algorithm are better than those of the other three adaptive genetic algorithms and the traditional genetic algorithm.

An approach is proposed for fabric defect detection based on the improved conventional pulse coupled neural network(PCNN) model.For these too many parameters of conventional PCNN,it is difficult to get the adaptive parameters.The problem can be solved in the proposed way,in which optimal number of iteration to segment fabric defect image automatically is determined based on minimum difference of uniformity within region.Segmentations on various defect images are implemented with the proposed approach and the experimental results demonstrate its reliability and validity.

An algorithm is proposed to improve the performance of skin detection algorithms under poor illumination conditions. A hybrid skin detection model is addressed to solve these problems by combining two Gaussian models of skin under normal conditions and bright illumination. According to the distribution of the combined models, the algorithm automatically evaluates the skin segmentation result of an adaptive threshold algorithm based on a Gaussian model by estimating the illumination conditions of image. If the estimation result shows that the illumination condition is very different from the normal one, the skin color of the original image needs compensation, and then the algorithm feeds the compensated image back to the Gaussian model for finer skin detection. The experimental results show that our algorithm can cope with a complex illumination change and greatly improve skin classification performance under inferior illumination conditions.

To avoid the rotation and the interpolation steps in the image registration, and decrease the data dimension as well as the computational complexity, a cross-correlation registration algorithm based on the image rotational projection was proposed in this paper. It calculated the cross-correlation of the projection and the original image twice according to predictive estimation of the rotation angle in simulation figures. The rotation angle and offset corresponding to the image with bigger cross-correlation is the needed one. Simulations on standard images showed that the proposed algorithm is more accurate and faster than the Fourier transform based algorithm.

A new idea of tunable narrow waveguide band-elimination filter is proposed based on electric-driven control.Using step motor to tune the filter,a tunable narrow waveguide band-elimination filter is designed and realized.Because of the usage of step motor,the tuning of filter is very convenient and accurate.This device have higher Q value and bigger power capability.Furthermore,the fabrication of the filter is very easy.The results of experiment prove that this design is feasible and effective.

A new method of detection the edges of an image is presented in this article. The method uses a kind of twodimensional subband spectrum analysis (2D-SSA) filter that is based on subband decomposition, and it is very convenient to get the edge frequency spectrum of an image after certain preprocessing. Comparing with spatial methods, the method is less sensitive to noise. It is also superior to the conventional frequency methods. In conventional frequency methods, the bandwidth and central frequency of filter are fixed, and it needs to transform the whole image into frequency domain. While in this method, the bandwidth and central frequency can be adjusted flexibly, and it only uses a few pixels to implement FFT. So this method is a fast way to extract the edges of an image. The simulation results show its efficiency.

针对人耳识别中人耳的角度变化这个难点问题,提出一种结合Gabor小波和监督保局投影的人耳识别算法.由于Gabor特征维数高、冗余大,首先通过统计样本的边缘点再采样的方法对人耳进行稀疏的描述,然后利用类别可分离性判据评价Gabor展开系数的分类能力,选择最有利于识别的Gabor展开系数构造新的Gabor特征.在人耳库中的实验结果表明,采用文中算法提取的Gabor特征维数少、鉴别能力强,结合监督保局投影进行识别取得了很高的识别率,对于人耳角度的变化具有良好的鲁棒性.

The factors which cause additional losses of guidance optical fiber in wound state were analyzed.A mathematical model used to analyze the macro-bend losses in the cross region producing in the precision winding process was established.For an actual guidance optical fiber,the measured data of the fiber's additional losses under low temperature and the loss curves with radius were given in the paper.The simulation results were compared with the test data.It shows that the additional losses of optical fiber caused by bending and low temperature can meet the actual requirements of the fiber optical guidance system.The established model can be used to predict the change trend of fiber losses in the winding process with a certain tensile force.

In order to improve the real-time performance of the real-time HLA(high level architecture) in the application of massive data communication volume,multi-thread processing was adopted,thread pool structure was introduced into the system,different threads to handle corresponding message queues was utilized to respond different message requests.Furthermore,an allocation strategy of semi-complete deprivation of priority was adopted,which reduces thread switching cost and processing burden in the system,provided that the message requests with high priority can be responded in time,thus improves the system's overall performance.The design and experiment results indicate that the method proposed in this paper can improve the real-time performance of HLA in distributed system applications greatly.

Image segmentation problem often demands the incorporation of as much prior information as possible to help the segmentation algorithms extract the tissue of interest.The model of image segmentation based on statistical shape prior level set was reviewed.The feature of mode is the energy function of the model composed by two terms.The first one is data term based on the image gradient or region gray intensity,the second one is the shape prior term which provides robustness against missing shape information due to cluttering,occlusion and gaps.How to construct the implicit shape model which aims to extract a compact representation for the structure of interest from a set of training examples,how to construct the evolve model to constrain an implicit surface to follow global shape consistence while preserving its ability to capture local deformation were discussed intensively.The key problems such as shape registration and the correspondence problem were introduced.Finally the open issues and possible future research directions were pointed.

Noise-jamming and multi-false target jamming are the most common jamming style,but the two style are of strong pertinency.However,nosie-jamming accompanyed with multi-false target jamming is a important jamming style,it can achieve good results aiming at the radar of antiaircraft weapons system.This thesis set up the jamming mathematical models of noise-jamming and multi-false target jamming and analyse the afffects that nosie-jamming and multi-false target jamming arouse,simulation result indicates that nosie-jamming accompanyed with multi-false target jamming can improve airplane defense-breaking probability.

As the 3D datasets are usually in large scalar,the capability of a single CPU to rendering is not sufficient to achieve interactivity.Direct volume rendering via GPU has positioned itself as an efficient tool for the display and visual analysis of volumetric scalar fields.A rapid PC hardware based visualization method for large-scale datasets method was proposed.At last we demonstrated the effectiveness of our method with several data sets.It was proved that the proposed method can generate high-quality visual representations on normal PC.

In order to improve the real-time performance of the real-time HLA(high level architecture) in the application of massive data communication volume,multi-thread processing was adopted,thread pool structure was introduced into the system,different threads to handle corresponding message queues was utilized to respond different message requests.Furthermore,an allocation strategy of semi-complete deprivation of priority was adopted,which reduces thread switching cost and processing burden in the system,provided that the message requests with high priority can be responded in time,thus improves the system's overall performance.The design and experiment results indicate that the method proposed in this paper can improve the real-time performance of HLA in distributed system applications greatly.

The exact knowledge of the blood vessel geometry plays an important role, not only in clinical applications (stroke diagnosis, detection of stenosis), but also for deeper analysis of hemodynamic functional data, such as fMRI strongly depending on the vessel structure. Such vessel geometries can be obtained by different MR angiographic. First we present algorithms for automatic vessel reconstructions from different MRA angiographic modalities. Moreover, we show that simulations using computational fluid dynamics (CFD) can be used to validate the vessel geometry, reconstructed from time-of-flight (TOF) angiograms. CFD simulations are based on phase-contrast angiography (PC-MRA) data, since these data contain rheological information (phases) besides merely amplitudes as is the case for TOF measurements. Parts of the rat brain vessel system are carefully modeled consisting of a main tube and second order branches. By analyzing velocity changes up and downstream of bifurcations, we show that CFD can be used to help detecting missing vessels in the TOF based reconstruction. We demonstrated this by artificially deleting a branch from the reconstruction and compared the flow in both resulting CFD simulations. Finally the simulations help to understand the effects of secondary branches on the flow in the main tube.

Using the example of airborne radar and jammer,the EMC cause of airborne devices is analyzed and theory model of electromagnetic mutual jam is established.Under the different jamming circumstance and modes,a calculate measurement of airborne radar detection range is put forward and radar range equation in jamming condition is developed when blanketing jamming and self-protection jamming occur.On the basis of common calculate measurement of radar detection probability,the relation between radar detection probability and target distance is deduced in self-protection jamming and simulations are accomplished.It is a reference for solving.

A series of experimental pressure measurements inside internal flow fields employing the intensity-based method of pressure sensitive paint technique were depicted based on self-established optical pressure measurement system and indigenously developed pressure sensitive paint.These experimental measurements include the applications inside both the transonic cascade wind tunnel and the nationally unique axial dual counter-rotate compressor facility(ADCCF) of Northwest Polytechnical University(NPU).The test articles involved a single vane of a large bending angle in transonic cascade wind tunnel and the oriented vane inside ADCCF,and both the surfaces of interest were correlated with the suctions.The traditional pressure measurement using electronic pressure scanning has been simultaneously conducted in the cascade wind tunnel for comparison.The contrast of both results from optical pressure measurement and electronic pressure scanning has revealed that the feasible application of optical pressure measurement system in internal flow fields and in practical industrial engineering.

为提高目标识别率,在目标图像融合过程中引入Markov随机场建立类别的先验分布模型,针对模型中参量β的选取问题,提出了基于各类各向异性的期望最大化-最大后验概率-多层次马尔可夫随机场集中式与分布式两种图像融合算法.实验证明,两种融合算法都既可以提高分类准确度,又能够增大抗噪能力,且二者又有不同的特色,可以根据实际要求(如,运算速度、分类准确度、计算负荷等)进行应用选择,用以提高对特定目标进行自动检测与识别的准确性.

The pressure-sensitive-paint (PSP) technique is new kinds of dynamic-measurement technique with the characteristic of non-intrusion and the continuous pressure measure on whole model surface. It is now commonly used in stationary wind-tunnel tests in abroad. In this paper,presents the basic principles,the main components of the PSP measurement systems,paint calibration,how to use this technique to measure pressure,were introduced,and the developing of PSP technique was described.