兔死后晶状体赤道部形状时序性变化规律研究

目的探讨应用数字图像分析技术量化分析个体死后晶状体赤道部形状时序性变化规律的可行性,为利用眼组织推断死亡时间(postmortem interval, PMI)提供一种补充手段。方法建立兔空气栓塞死亡模型。在22±1℃、30%相对湿度室内,于兔死后96 h内每隔6 h分离晶状体,应用智能手机结合简易光源装置采集晶状体赤道部图像,基于Matlab软件设计相关程序代码,分割赤道部图像并提取赤道部边缘周长、面积及圆形度3项形状特征数值,建立PMI推断线性回归模型。结果兔死后96 h内,晶状体赤道部边缘周长、面积随PMI呈先上升后下降趋势,而圆形度随PMI呈线性上升趋势。建立的3项参数与PMI的线性回归模型均具有显著统计学意义(P0.05),且利用圆形度推断PMI的线性回归模型决定系数最高(R~2=0.925)。结论个体死后晶状体赤道部形状变化呈一定时间规律性。研究建立的晶状体赤道部数字图像分析方法及指标为客观、量化推断PMI提供参考依据。     

夜间红外行人图像在可见光行人图像库的检索算法

本文旨在弥补夜间红外模式监控下行人图像与白天监控下行人图像间的模态差异和异构特性，较好实现多摄像头下红外人像与可见光人像数据间的检索和比对任务。设计一种局部参数共享的双路卷积神经网络模型，提取具有全局粗粒度和局部细粒度的两种人像特征，通过对两种特征分别进行特征度量，实现跨模态人像检索模型的训练和优化。结果表明，模型所提取的两种特征在测试数据集上具有较好的应用效果，准确率与目前主流算法相比具有竞争力。本研究能够服务视频侦查工作，提高现有的动态、静态人像比对的应用水平。     

基于多尺度特征融合的穿鞋足迹识别网络

近年来，随着人工智能的快速发展，赤足足迹和穿袜足迹的自动识别均取得了不错的效果，但因为需要脱鞋，应用场景受到了很大的限制。单枚穿鞋足迹识别由于鞋底花纹的多样性、随机性给智能识别造成了很大的阻碍，识别准确率普遍较低，成为了一个具有挑战性的任务。本文聚焦于不同人穿同一种鞋的足迹识别，提出一种基于多尺度特征融合的穿鞋足迹识别网络。采集大量的足迹样本，通过旋转、平移、加噪声等方式模拟犯罪现场足迹图像可能产生的变化，制作了足迹数据集，然后以ResNet为骨干网络，利用双向金字塔特征融合模块充分融合足迹图像的深层特征和浅层特征，最后针对鞋底花纹变化造成网络识别准确率降低的问题，提出了用迁移学习解决的思路，让网络学习未知花纹与现有花纹之间的联系，快速拟合出模型。经过训练测试，本文提出的方法在自制足迹数据集上识别准确率达到了93.1%,CMC评价指标也明显优于其他网络。在新的足底花纹上，迁移学习对比从头开始训练，速度更快、准确率更高。大量的实验证明，本文提出的穿鞋足迹识别网络识别准确率更高，提出的迁移学习的方法在面对新的鞋底花纹时，能够实现更好的效果。     

高精度三维人脸图像数据库

3D人脸图像数据库广泛应用于计算机视觉、动画绘图设计、医学等很多领域。在法庭科学领域,采集三维人脸图像并建立数据库,可进行人像特征分类、统计人像特征的分布以及训练人像模型,这些分析是人像比对和识别的基础。与传统的二维数据库相比,三维(3D)人脸图像库能够提供更多信息,例如,三维人脸图像的空间结构和形状包含多视角轮廓。3D人脸图像采集方法包含多视角几何信息的方法、结构光的方法和3D扫描仪的方法,这些方法有不同的采集设备和环境。国内外已经建立了几个有代表性的3D人脸图像数据库,例如MPI实验室的MPI和BJUT的BJUT-3D,但这些库在分辨率和精度方面尚有不足。本文首先回顾了MPI、BJUT-3D数据库和它们的采集环境,然后对建立中国人的高精度3D人脸图像数据库进行探索研究。用彩色手持三维扫描仪(Artec Spider)采集了1100个3D人脸图像,这些图像包含彩色纹理和深度信息(几何形状和点云),每个人脸图像的几何形状的采样点数目超过2000万,三角面片数目超过4000万。与BJUT-3D人脸数据库在人脸形状、分辨率和纹理等方面的比较结果显示,本研究采集的人脸图像有更高的精度,在嘴巴、鼻子、眼睛等方面比其他数据库中的人脸图像显示了更多的细节。建立的数据库将会支持在3D人像识别和算法评估方面的进一步工作。     

图像降噪和增强对人脸识别系统识别性能的影响

目的研究图像降噪和图像增强方法对人脸识别系统识别性能的影响,以期为人脸识别系统应用过程中的图像处理方法选取提供理论指导和技术方案。方法收集33起人像鉴定领域实际案例中的人脸图像素材,研究以高斯滤波和小波变换为代表的图像降噪技术以及具有边缘保持和小波变换特性的单帧图像超分辨率增强技术对人脸识别系统识别性能的影响,并对不同图像处理方法对人脸识别性能的影响进行量化比较分析。结果本文研究的图像降噪技术均显著提高了人脸识别系统的识别准确性,而图像增强技术虽然提高了人脸图像显示效果,但对人脸识别系统的识别性能无正向促进作用。此外,高斯模糊图像处理的图像降噪方法虽然简单,但与本文研究的其他方法比较,其在人脸识别系统识别性能改善方面效果最显著。结论人脸图像质量对人脸识别系统的识别性能具有显著影响,可以通过图像处理技术改善人脸图像质量进而提高人脸识别系统的识别准确性。其中,图像降噪处理可以显著提高人脸识别系统的识别性能,且比图像增强技术更适合于实际人像鉴定应用中的人脸识别系统识别性能增强。     

利用残缺赤足足迹推算身高

目的由于主客观条件的影响,犯罪现场中常见的足迹多为不完整的残缺足迹,如何能够有效地利用残缺足迹,对案件破获往往具有至关重要的作用。方法从残缺赤足足迹的反映形态出发,选择足迹中明显、稳定的特征点,确定特征量,通过数理统计分析,建立残缺赤足足迹推算身高的经验公式。结果在4个特征量中,跖前缘突点到跟后缘突点的距离与身高的相关性最好,跖内、外缘突点到跟后缘突点的距离与身高具有一定的相关性。结论根据现场残缺赤足足迹的反映情况,也可以近似推断嫌疑人的身高特点,为侦查提供方向。     

平面赤脚印形状特征量化技术初探

赤足形状参量的提取是利用赤足迹进行个体识别技术研究的关键技术,传统的手工测量和手工描绘都会产生较大的误差,本文利用图像处理技术从足迹图中提取赤足形状的轮廓线,然后利用AutoCad软件测量特征点的参数,建立了利用计算机自动提取赤足形状参数的方法。     

基于行走轨迹的异常行为分析

<正>异常行为的分析对于智能监控系统有重要意义,目前智能监控系统能够从视频图像中提取运动目标,同时对提取出的运动目标进行跟踪和识别,并对其行为进行理解和描述。运动检测、目标识别和跟踪属于低层次的处理,属于图像处理的范畴;行为分析属于高层次的处理,属于人工智能的范畴。运动目标检测与跟踪方法简述运动检测的目的是在图像序列中将运动人体区域从复杂背景中提取出来。运动检测是运动跟踪和识别的基础,该阶段处理结果的质量直接影响到以后处理的效果。目前常用的运动检测方法有:背景减除法、时间差分法、光流法。     

重建耻骨联合三维结构用于活体年龄推断的技术方法

目的建立基于MRI成像的快速耻骨联合三维结构重建技术方法。方法利用3.0 T超导磁共振仪对成年男性志愿者骨盆进行T1小角度快速激发3D梯度回波序列扫描,进行二值化阈值分割、区域增长提取骨盆软组织图像;运算获得耻骨联合三维结构模型雏形,经逆向工程软件降噪处理建立耻骨联合三维结构模型,并与耻骨CT扫描三维结构重建模型进行比较。结果耻骨联合面沟嵴、下端、腹侧缘(斜面)、背侧缘(斜面)、耻骨结节等结构显示完整,无明显噪点,与CT扫描三维重建模型吻合较好。结论 MRI扫描成像可实现快速有效的耻骨联合三维结构重建,可为法医活体年龄推断提供安全无辐射作用的三维可视化影像学技术方法。     

外伤性鼓膜穿孔面积测定方法及41例分析

目的探索外伤性鼓膜穿孔面积大小的测量方法并分析41例病例。方法用图像分析软件Photoshop分析耳内窥镜获得的外伤穿孔鼓膜照片，计算穿孔与鼓膜面积比例并分析外伤性鼓膜穿孔的法医临床学特性。结果外伤性鼓膜穿孔通常为小穿孔，边缘血污部分外翻，形状不规则，左侧居多。结论耳内窥镜可以清晰记录外伤性鼓膜穿孔，图像处理软件可以定量分析鼓膜穿孔大小。     

Yet another method for marking incisal edges of teeth for bitemark analysis

Over the years many methods have been proposed and presented for marking or "highlighting" the incisal edges of anterior teeth on a dental stone model -- one of the critical steps in production of transparent overlays for bitemark analysis. A method is presented here that is an extension and refinement of the "paint the edges" method shown at least as early as 1966 by Gustafson. This method uses "invisible ink" and ultraviolet illumination to produce a very high contrast image of the incisal edges of the teeth on the model. The advantages of this new method are that it is nondestructive, it does not leave visible marks on the models, operator subjectivity in selecting the portions of anterior teeth to be highlighted is reduced, and the author submits that this technique lends itself to easy construction of overlays using computer image-processing software such as Adobe Photoshop and ImageJ.     

Homicidal Snake Bite in Children

Snake bites are common in many regions of the world. Snake envenomation is relatively uncommon in Egypt; such unfortunate events usually attract much publicity. Snake bite is almost only accidental, occurring in urban areas and desert. Few cases were reported to commit suicide by snake. Homicidal snake poisoning is so rare. It was known in ancient world by executing capital punishment by throwing the victim into a pit full of snakes. Another way was to ask the victim to put his hand inside a small basket harboring a deadly snake. Killing a victim by direct snake bite is so rare. There was one reported case where an old couple was killed by snake bite. Here is the first reported case of killing three children by snake bite. It appeared that the diagnosis of such cases is so difficult and depended mainly on the circumstantial evidences.     

人体体表损伤面积及关节活动角度智能测量与辅助鉴定系统开发应用

本系统由图像采集、分析处理、档案管理、报告生成、联机帮助五个子系统组成。图像采集采用摄像录入，可以支持数码相机、扫描仪录入。面积计算采用模式识别技术标定边界“曲面坐标网格法”直接计算损伤面积。关节活动角度测量允许交点不在图像内，用终止用与起始角相减获得关节活动角度，对四肢及脊柱活动均能较准确测量。     

Ear identification: a pilot study

Although several papers have recently been devoted to establishing the validity of identification using the ear, this part of the human body still remains underexploited in forensic science. The perfect overlap of two images of the same ear is not really possible, but photographs of the ears as a reliable means of inferring the identity of an individual are poorly treated in the literature. In this study, we illustrate a simple, reproducible method, which divides the photograph of an ear into four parts-helix, antihelix, concha, and lobe-by means of a suitable grid of four straight lines. Although the division does not follow exact anatomical features, their edges do join anatomical points which are more easily identifiable. Measurement of certain areas of these parts can be combined to produce a code allowing personal identification. This method produces false-positive identifications of <0.2%. Last, the repeatability and reproducibility aspects of the method are tested.     

Breath alcohol measurement variability associated with different instrumentation and protocols

Breath alcohol measurement has variability resulting from instrumental, procedural and biological components. Reliable estimates of the standard deviation (S.D.) are necessary for calculating uncertainty in the form of confidence intervals. These estimates are concentration dependant and can be obtained from models derived from duplicate breath test data. Duplicate data from four state jurisdictions (Alabama using the Drager 7110, Minnesota using the Intoxilyzer 5000, Washington using the BAC Datamaster and Wisconsin using the Intoximeter EC/IR) were analyzed to derive predictive models for standard deviation as a function of concentration. All jurisdiction/instrument combinations yielded forensically acceptable variation while showing a general linear increase in standard deviation with concentration. This is consistent with a multiplicative error model. Jurisdictions using the same instruments but not performing duplicate analyses could also employ the straightforward models derived here to estimate standard deviations for their measurement results.     

区域金融发展、居民金融资产与收入分配——以重庆市为例

基于重庆市1990~2009年的数据,运用VAR模型和脉冲响应函数,实证分析了金融发展、居民金融资产与收入分配的关系。重庆市金融发展与收入分配之间非同阶单整变量,并不存在长期均衡关系。但在一阶差分VAR模型基础上所做的脉冲响应函数分析表明,金融规模扩大导致收入差距呈现由缩小到扩大再到缩小的过程;金融效率的提高则使收入差距先扩大后缩小。城乡人均储蓄比和城乡房产比都扩大了收入差距,而农村居民非农就业水平的提高则有助于缩小收入差距。     

A Study on the Estimation of Postmortem Interval Based on Environmental Temperature and Concentrations of Substance in Vitreous Humor

A method to determine postmortem interval (PMI) based on environmental temperature and the concentrations of vitreous humor (VH) molecules were explored. Rabbit carcasses were placed in a chamber at 5, 15, 25, or 35°C, and 80–100 μL of VH was collected with the double‐eye alternating micro‐sampling method every 12 h. A Roche DPPI biochemical analyzer was used to measure the concentrations of six substances in VH samples. The interpolation function model and mixed‐effect model were employed for data fitting to establish equations for PMI estimation. The concentrations of K⁺, P, Mg²⁺, creatinine (CRE), and urea nitrogen (UN) exhibited an upward trend with increasing PMI in all temperature groups, while the concentration of Ca²⁺ showed a downward trend. Validation results using K⁺ and Mg²⁺ ions revealed that the mixed‐effect model provided a better estimation than the interpolation function model using the data from our experiment. However, both models were able to estimate PMI using temperature and VH molecule concentrations.     

Ancestry Estimation Using Cranial and Postcranial Macromorphoscopic Traits

Ancestry estimation methods using macromorphoscopic (MMS) traits commonly focus exclusively on cranial morphology. The objective of this study was to demonstrate the value of postcranial MMS traits, highlighting a combined cranial/postcranial trait approach to ancestry estimation using quadratic discriminant function and a variety of machine learning classification models including artificial neural networks (aNN), random forest models, and support vector machine. Eight cranial and eleven postcranial MMS traits were collected from the Terry and Bass Skeletal Collections (American Black = 81; American White = 173). Our classification models using cranial and postcranial traits correctly classified 88–92% of the sample, improving classification accuracies by nearly fifteen percent over models relying exclusively on cranial data. These same results demonstrate the importance of a multivariate statistical framework incorporating cranial and postcranial data and the nearly unlimited potential of machine learning models to improve the accuracy of ancestry estimates over traditional methods of analysis. To facilitate implementation in casework, one of the more robust models (aNN) is incorporated into a web-based application, ComboMaMD Analytical, to facilitate cranial and postcranial MMS traits analysis for ancestry estimation.     

Analysis of ecstasy tablets: comparison of reflectance and transmittance near infrared spectroscopy

Calibration models for the quantitation of commonly used ecstasy substances have been developed using near infrared spectroscopy (NIR) in diffuse reflectance and in transmission mode by applying seized ecstasy tablets for model building and validation. The samples contained amphetamine, N-methyl-3,4-methylenedioxy-amphetamine (MDMA) and N-ethyl-3,4-methylenedioxy-amphetamine (MDE) in different concentrations. All tablets were analyzed using high performance liquid chromatography (HPLC) with diode array detection as reference method. We evaluated the performance of each NIR measurement method with regard to its ability to predict the content of each tablet with a low root mean square error of prediction (RMSEP). Best calibration models could be generated by using NIR measurement in transmittance mode with wavelength selection and 1/x-transformation of the raw data. The models build in reflectance mode showed higher RMSEPs using as data pretreatment, wavelength selection, 1/x-transformation and a second order Savitzky-Golay derivative with five point smoothing was applied to obtain the best models. To estimate the influence of inhomogeneities in the illegal tablets, a calibration of the destroyed, i.e. triturated samples was build and compared to the corresponding data of the whole tablets. The calibrations using these homogenized tablets showed lower RMSEPs. We can conclude that NIR analysis of ecstasy tablets in transmission mode is more suitable than measurement in diffuse reflectance to obtain quantification models for their active ingredients with regard to low errors of prediction. Inhomogeneities in the samples are equalized when measuring the tablets as powdered samples.     

Localization and quantification of histamine in injured skin as parameters for the timing of wounds.

Localization and estimation of the histamine (HA) content in skin wound edges in 86 Sprague-Dawley rats and three cases of human injuries were carried out by a microfluorimetric method specific for this amine which forms a complex with o-phthalaldehyde (OPT). Distribution and density of the mast cells in the same areas were observed at the same time by toluidine blue stain. In all skin specimens with antemortem wounds, both the epidermis and upper dermis exhibit extracellular yellowish fluorescence of the HA-OPT complex. The fluorescent zone spreads in the wound edges with the lapse of time in vital injuries. The HA content increases gradually up to 30 min and then the yellow histamine fluorescence in areas 0-200 microns from the wound edge decreases. None of these features can be observed in normal skin and postmortem-injured skin. Mast cell degranulation can be demonstrated in all antemortem-injured skin. No statistical relationship exists between the number of mast cells and the HA-OPT fluorescence in either ante- or postmortem-injured groups. This study indicates that skin HA microfluorimetry by the OPT method is of practical value for distinguishing ante- from the postmortem wounds and for timing antemortem wounds.