带钢表面缺陷在线检测实时处理方法研究

设计了一种带钢表面缺陷在线检测系统,提出了一种判断在线采集图像中是否存在缺陷的方法,并设计了半封闭图像采集装置,实验证明通过差影图中非零像素点个数的统计,达到了较好的实验效果,证明该方法的可行性.     

基于向量的带钢缺陷检测算法

介绍了用计算机视觉检测系统中的图像处理技术检测带钢的边裂以及孔洞。首先识别带钢图像的轮廓,得到带钢边部、孔洞的轮廓曲线向量编码,然后利用向量的点积计算轮廓的弯曲程度,据此作为判断条件,检测边裂与孔洞缺陷。实验结果表明：该方法应用到实际检测系统中,能够可靠地检出带钢的边裂。     

基于Contourlet变换的连铸坯表面缺陷识别

根据连铸坯表面图像的特点,提出了一种基于Contourlet变换的连铸坯表面缺陷识别方法.通过Contourlet变换将样本图像分解成不同尺度和方向的子带,提取子带的Contourlet系数特征,并结合样本图像的纹理特征,得到一个高维的特征向量.利用监督核保局投影算法对高维特征向量进行降维,将降维后的低维特征向量输入支持向量机,对连铸坯表面图像进行分类识别.对现场采集到的裂纹、氧化铁皮、光照不均和渣痕四类样本图像进行实验,本文提出的识别方法对样本图像的识别率可达94.35%,优于基于Gabor小波的识别方法.      

非线性降维在冷轧带钢表面缺陷分类中的应用

将冷轧带钢表面缺陷图像中的所有像素作为高维空间中的特征向量,利用有监督非线性降维方法对其进行减维后再进行缺陷的分类。该方法解决了冷轧带钢表面缺陷自动分类中的特征提取和特征选择的困难,避免了分类器特征维数过高的问题,并可以用于动态数据的在线识别和聚类。用这种降维方法并结合K近邻分类器与支持向量机对现场采集到的缺陷样本数据集进行试验,结果表明经过降维预处理后,2种分类器的性能都得到了很大的提高。     

基于图像预处理的神经网络带钢缺陷检测

针对带钢表面缺陷的特点,提出了1种基于图像预处理消除光照不均等的干扰且用神经网络进行缺陷识别的检测方法.带钢缺陷的检测分为3步:首先,对采集的图像进行预处理,通过图像零均值化以消除光照对检测的影响,分别利用维纳滤波和sobel算子对图像进行滤波除噪和锐化处理;其次,通过最大类间方差法进行图像分割以及计算面积来判断是否存在缺陷;最后,在提取图像特征的基础上,通过设计人工神经网络识别带钢缺陷类型.实验表明,采用的方法能够有效抑制图像背景干扰,能够有效地实现带钢缺陷的快速检测.     

中厚板表面缺陷在线检测系统的分类器设计

研究了BP网络、LVQ1网络、LVQ2网络所构建的分类器的性能,将这3种分类器用于中厚板表面缺陷的自动分类中.从现场在线采集中厚板的表面缺陷图像,将每幅表面图像划分成64×64大小的子图像,对子图像进行FFT变换,得到子图像的幅值谱.将幅值谱中心区域内的像素灰度值作为特征量,分别输入给BP网络、LVQ1网络、LVQ2网络所构建的分类器模型,试验表明LVQ2网络能够得到理想的分类效果.     

基于改进的支持向量回归机算法的磁记忆定量化缺陷反演

针对焊缝缺陷磁记忆检测中存在定量化反演难题,建立了基于改进的支持向量回归机定量反演模型.以预制不同尺寸未焊透和夹渣缺陷的Q235焊接试样为试验材料,进行磁记忆扫描检测发现:缺陷位置的磁记忆信号特征参数随尺寸变化而呈现一定的变化规律,但同时存在分散性和不确定性.鉴于磁记忆信号样本的有限性、分散性和非线性,首先将提取到的磁记忆特征参数进行归一化处理,引入支持向量回归机建立焊缝缺陷磁记忆定量反演模型,并进一步利用模拟退火算法对支持向量回归机参数进行优化,使目标函数达到全局最优而非局部最优.最后,考虑到由磁记忆信号逆向反推缺陷的三维尺寸,存在解的不确定性,为此在缺陷单维尺寸反演模型的基础上,通过构建多层结构的支持向量回归机进行多尺寸反演输出,建立了基于模拟退火支持向量回归机的焊缝缺陷磁记忆定量反演模型,结果表明:未焊透缺陷尺寸反演最大相对误差为7.96%,夹渣缺陷为4.97%,为焊缝缺陷的磁记忆反演与定量化评价提供一种新的思路.      

基于图像零均值化的带钢缺陷检测

 针对带钢表面缺陷的特点,提出了一种基于图像零均值化的检测方法。首先,通过对测试图像进行零均值化,以消除光照对检测的影响;其次,利用维纳滤波对零均值化图像进行滤波除噪;在此基础上,采用Sobel进行锐化处理;最后,通过最大类间方差法进行图像分割,从而实现对带钢表面缺陷的检测。试验表明,本方法能够有效抑制图像背景干扰,有效地实现带钢缺陷的快速检测。     

超宽带钢典型板形缺陷向量提取方法

提出了从带钢板形数据中提取典型板形缺陷向量的方法,即先对带钢板形缺陷进行初级分类,再根据带钢应力之间的相似度提取出单位典型板形向量,最后采用k均值聚类算法求解出典型比值系数.针对提取出的带钢单位典型板形向量而设计出的单侧倒角工作辊,经过大规模的轧制试验,所轧宽带钢板形质量较好,综合平坦度均小于5 IU（1 IU=10_-5）,基本消除边中复合浪为主的复杂浪形缺陷.      

基于扫描线的带钢低对比度缺陷图像分割技术

在带钢表面质量视觉检测过程中,一些低对比度缺陷图像应用梯度算子进行分割时,往往找不到图像中的缺陷,从而出现漏检的情况.为了改善带钢图像中低对比度缺陷的分割效果,提出了一种基于扫描线的阈值图像分割算法,该算法可以有效实现平整花、点划伤等低对比度缺陷的分割,同时也能较好地实现对其他缺陷的分割.     

基于遗传算法的带钢表面缺陷特征降维优化选择

针对带钢表面的划痕、黑斑、翘皮、辊印、褶皱和压印6种典型缺陷,提取了样本图像的灰度、纹理和几何形状特征等32维特征向量。基于遗传算法对32维特征向量进行降维优化选择,选择了其中的20维以进行缺陷图像类型的分类。利用BP神经网络对降维前后的6种典型带钢表面缺陷分类进行对比识别,并同主成分降维方法进行了对比,验证了所提取的...     

Multiple pregnancy and microelements: zinc (preliminary report)

Incidence, causes and results of permanent ICA clipping during arterial cerebral aneurysm surgery were analyzed in randomized group of 470 patients. Permanent ICA clipping was performed in 6 cases (1.3% of all surgeries). The causes of permanent ICA clipping were ICA wall rupture in 4 (0.9%) cases, bleeding from aneurysms of the ophthalmic segment of the internal carotid artery (ICA) in 2 (0.4%). Atherosclerotic changes of ICA were found and verified by biopsy in all cases of ICA rupture. Two variants of rupture were identified. ICA aneurysm avulsion from the artery in the area of aneurysmal neck with vascular wall defect development in 2 (0.6%) cases; in the second variant, ICA rupture was caused by frontal lobe traction. In the study group aneurysms of ophthalmic segment of ICA were found in 19 cases: intraoperative bleeding rate was 31% (6 cases). ICA was clipped in 33% of all cases of intraoperative bleeding. Among 6 patients with permanent ICA clipping, 5 died. Deaths were caused by cerebral infarction in 4 cases and acute blood loss in 1 case.     

Familial aorto-cervicocephalic arterial dissections and congenitally bicuspid aortic valve

BACKGROUND: A primary arteriopathy is often implicated in the etiology of spontaneous cervicocephalic arterial dissections, but its exact nature usually remains unknown. We describe the familial occurrence of spontaneous arterial dissections and congenitally bicuspid aortic valve (BAV) and propose a common developmental defect in these families. SUMMARY OF REPORT: In the first family, a 63-year-old man suffered an extracranial internal carotid artery (ICA) dissection, and his 43-year-old cousin with BAV suffered an intracranial vertebral artery (VA) dissection. Two other family members had pathologically proven BAV. In the second family, a 31-year-old woman suffered bilateral extracranial ICA and VA dissections. Her father, at age 46, suffered an aortic dissection associated with cystic medial necrosis and BAV. Her paternal uncle died from an aortic dissection at age 59. In the third family, a 39-year-old woman suffered extracranial ICA and VA dissections, and her brother died at age 48 from an aortic dissection associated with a BAV. CONCLUSIONS: The familial occurrence of spontaneous arterial dissections and BAV suggests a common developmental defect. The aortic valvular cusps and the arterial media of the aortic arch and its branches are derived from neural crest cells, suggesting that a neural crest defect may be the underlying abnormality in these families.     

Combined screening for autoantibodies to IA-2 and antibodies to glutamic acid decarboxylase in first degree relatives of patients with IDDM. The DENIS Study Group. Deutsche Nikotinamid Interventions-Studie

To determine the value of antibodies to the intracytoplasmic domain of the tyrosine phosphatase IA-2 (anti-IA-2ic) and glutamic acid decarboxylase (GADA) for identification of subjects at risk for insulin-dependent diabetes mellitus (IDDM) we investigated 1238 first degree relatives of patients with IDDM for the presence of anti-IA-2ic and GADA and compared the results with cytoplasmic islet cell antibodies (ICA). Anti-IA-2ic were observed in 54 (4.4%) first degree relatives, in 51 of 86 (59.3%) ICA positive relatives and in 3 of 4 individuals who developed overt IDDM within a follow-up period of 1 to 28 months. GADA were found in 78 of 1238 (6.3%) first degree relatives. They were detected in 22 of 35 (62.9%) sera with ICA alone and in 1 of 3 subjects with anti-IA-2ic in the absence of ICA. Of the 1238 subjects 37 (3.0%) sera were positive for all three antibodies. Both anti-IA-2ic and GADA were positively correlated with high levels of ICA. Anti-IA-2ic and GADA were detected in 39.1 and 47.8% of subjects with ICA of less than 20 Juvenile Diabetes Foundation units (JDF-U) but in 66.7 and 76.2% of individuals with ICA of 20 JDF-U or more, respectively (p < 0.05). The levels of ICA and GADA in first degree relatives with at least one additional marker were significantly higher than in subjects with ICA alone (p < 0.005) or GADA alone (p < 0.03). The combination of anti-IA-2ic and GADA identified 84.9% of all ICA positive subjects and 93.7% of individuals with high level ICA (> or = 20 IDF-U). All 4 individuals who progressed to IDDM had either IA-2ic or GADA. Our data indicate that primary screening for anti-IA-2ic and GADA provides a powerful approach with which to identify subjects at risk for IDDM in large-scale population studies which may represent the basis for the design of new intervention strategies.     

Cytoplasmic islet cell antibodies recognize distinct islet antigens in IDDM but not in stiff man syndrome

Cytoplasmic islet cell antibodies are well-established predictive markers of IDDM. Although target molecules of ICA have been suggested to be gangliosides, human monoclonal ICA of the immunoglobulin G class (MICA 1-6) produced from a patient with newly diagnosed IDDM recognized glutamate decarboxylase as a target antigen. Here we analyzed the possible heterogeneity of target antigens of ICA by subtracting the GAD-specific ICA staining from total ICA staining of sera. This was achieved 1) by preabsorption of ICA+ sera with recombinant GAD65 and/or GAD67 expressed in a baculovirus system and 2) by ICA analysis of sera on mouse pancreas, as GAD antibodies do not stain mouse islets in the immunofluorescence test. We show that 24 of 25 sera from newly diagnosed patients with IDDM recognize islet antigens besides GAD. In contrast, GAD was the only islet antigen recognized by ICA from 7 sera from patients with stiff man syndrome. Two of these sera, however, recognized antigens besides GAD in Purkinje cells. In patients with IDDM, non-GAD ICA were diverse. One group, found in 64% of the sera, stained human and mouse islets, whereas the other group of non-GAD ICA was human specific. Therefore, mouse islets distinguish two groups of non-GAD ICA and lack additional target epitopes of ICA besides GAD. Longitudinal analysis of 6 sera from nondiabetic ICA+ individuals revealed that mouse-reactive ICA may appear closer to clinical onset of IDDM in some individuals. Mouse-reactive ICAs, however, remained absent in 36% of the patients at diagnosis of IDDM.     

Sensitivity and specificity for detection of islet cell cytoplasmic antibodies using rat pancreatic sections

We determined islet cell cytoplasmic antibodies (ICA) using rat pancreatic sections as a test substrate substitutive for human pancreatic sections by indirect immunofluorescent technique. ICA were measured in sera from 58 patients with insulin-dependent diabetes mellitus (IDDM), 456 with non-insulin-dependent diabetes mellitus (NIDDM), 50 patients with autoimmune diseases, and 110 healthy controls. Seventeen of 58 patients with IDDM showed recent-onset (within 3 months). ICA were also measured in some samples using blood group O human pancreatic sections, and the ICA titers were compared with those measured using rat pancreatic sections. The prevalence of ICA was 55.2% (32/58) in patients with IDDM, 1.5% (7/456) in those with NIDDM, 0% (0/50) in those with autoimmune diseases, and 0.9% (1/110) in the healthy controls. Of the 17 recent-onset IDDM ICA were positive in 14 (82.3%). In comparative study of titers for ICA using rat pancreatic sections or human pancreatic sections, rat pancreatic sections yielded ICA titers as high as human pancreatic sections did. These results demonstrate that ICA assay using rat pancreatic sections was disease-specific, and that antigenicity of the substrate was favorable to ICA. Rat pancreas presents the advantage of greater availability, while providing an identical substrate for ICA. In conclusion, rat pancreatic sections are useful substrate for detecting ICA.     

Estrogen and progesterone receptors in endometrial carcinoma: comparison of immunohistochemical and biochemical analysis

In 159 endometrial carcinomas, estrogen (ER) and progesterone receptors (PR) were determined biochemically by dextran-coated charcoal (DCC) assay and immunohistochemically (ICA) on frozen sections. ICA receptor content was estimated by a total histologic score (HSCORE), including all tissue components, and by a cancer HSCORE, including malignant cells only. These scores were closely correlated. A single biopsy was found to be representative for each tumor. ER-DCC status was positive in 90.3% and PR-DCC status in 92.2% of the tumors. ER total HSCORE was positive in 47% and PR total HSCORE in 89% of tumors. ER and PR correlated inversely with tumor grade (p < 0.001). Correlations were found between ER and PR content determined by either method (DCC: r = 0.77; ICA: r = 0.50), as well as between DCC and ICA content (ER: r = 0.52; PR: r = 0.76). The association between DCC and ICA was affected by the tumor grade: the DCC values decreased relatively more than total HSCOREs with increasing grade. The sensitivity of ICA against DCC assay was 56% for ER and 86% for PR. Maximal agreement between receptor status as determined by ICA and by DCC would result from a DCC cutoff level of 130 fmol/mg for ER and 114 fmol/mg for PR.     

Comparison of isradipine and enalapril effects on regional carotid circulation in patients with hypertension with unilateral internal carotid artery stenosis

This randomized, double-blind, placebo-controlled study was aimed at detecting cerebrovascular effects of isradipine and enalapril in patients with moderate hypertension depending on the presence and grade on unilateral stenosis of internal carotid artery (ICA). We evaluated carotid vascular resistance by using Doppler analysis and regional cerebral blood flow (rCBF) by using 133Xe-clearance technique before and after a single 5-mg oral dose of isradipine, enalapril, or placebo. Their effects were randomly and consecutively tested in 73 patients with essential hypertension subdivided into three groups: without carotid occlusive lesions, with moderate (50-75%), and with severe (76-99%) unilateral asymptomatic ICA stenosis. There were no differences in age, gender, and antihypertensive effects of the drugs between these three subgroups. Three major variants of cerebrovascular drug effects were observed: absence of changes (variant I), decrease in carotid vascular resistance with increase in rCBF and elimination of side-to-side asymmetry (variant II), and increase in carotid vascular resistance with further reduction of rCBF ipsilaterally ICA stenosis, and increased side-to-side asymmetry (variant III). Frequency of variant III was significantly higher in patients with severe ICA stenosis. Enalapril produced variant I of cerebrovascular effects in most patients examined; variant III was observed only in 13% of patients with severe ICA stenosis. Isradipine produced variant I of cerebrovascular effects much less frequently than did enalapril. For this drug, variant II was most typical in patients without ICA stenosis and with moderate ICA stenosis. In 43.5% of patients with severe ICA stenosis, however, isradipine produced reduction of cerebral perfusion. Presumably the presence of ICA stenosis, especially >75%, increases the risk of cerebrovascular disorders in antihypertensive therapy. In patients with severe ICA stenosis, treatment with enalapril appears to be safer than that with isradipine.     

Cerebral haemodynamics in internal carotid artery trial occlusion

The purpose of this study was to analyse the cerebral haemodynamic changes brought about by trial occlusion of the internal carotid artery (ICA). Sixteen patients with surgically inaccessible cerebral aneurysms, carotid cavernous fistulas or neck neoplasms were monitored with transcranial Doppler ultrasonography (TCD) during 90-120 s angiographic ICA balloon occlusion or ICA closure with a Selverstone clamp. The blood velocity (V) was registered continuously in both middle cerebral arteries (MCA) while the pulsatility index (PIMCA) and haemodynamic tension (Uhem MCA) were calculated. ICA closure led to an instantaneous drop in the ipsilateral VMCA, PIMCA and Uhem MCA. The VMCA thereafter increased gradually until reaching a stable level. The subjects were grouped into those with initial drops in VMCA to > or = 60% of pre-occlusion value (group 1) and those that fell to < 60% (group 2), respectively. In group 1 autoregulatory mechanisms made the PIMCA decline further, while the Uhem MCA remained unaltered during ICA closure. In group 2, however, the PIMCA did not change further, while the Uhem MCA increased slightly. The cerebral haemodynamic features during ICA test occlusion were thus essentially different in the two groups. On re-opening the ICA, there was an overshoot in VMCA and Uhem MCA. Contralaterally, the VMCA was increased during ICA occlusion. Seven of the patients later had their ICA closed permanently. While none of five group 1 patients developed haemodynamic complications, two group 2 individuals experienced haemodynamic stroke. Assuming ICA sacrifice is feasible when test occlusion results in an ipsilateral initial reduction in VMCA to > or = 60% of pre-occlusion value, the corresponding limit for the Uhem MCA is > or = 40%. In the pre-operative evaluation of the haemodynamic risk related to ICA loss, TCD emerges as a reliable method. It also seems to allow for the reduction of test occlusion time to 90-120 s.