基于相机阵列的光学组件缺陷在线检测方法

利用调焦方式可以实现焦距的连续变化从而对不同物距下的光学组件进行在线检测,但是调焦过程操作复杂且对调焦位移精度要求较高,景深内光学元件缺陷无法区分,难以实现真正意义上的在线检测。因此,本文提出了基于相机阵列的光学组件缺陷在线检测方法。首先建立了相机阵列的成像模型并给出了数字重聚焦表达式以及空间分辨率的表达式。接着利用MATLAB模拟相机阵列成像过程和数字重聚焦过程。最后进行实验验证,通过二维位移台带动相机对不同物距下的多个光学元件表面缺陷进行成像获得阵列相机图像,通过数字重聚焦算法得到不同物距下的光学元件表面缺陷分布信息。实验结果表明,基于相机阵列的光学组件缺陷在线检测技术能够同时对位于景深范围内的光学组件进行在线检测。该方法在光学元件缺陷在线检测方面有着一定的应用价值。     

光学图像中的抛物线目标特征检测方法研究

针对光学图像中的抛物线目标特征检测问题,提出一种基于最小二乘法的抛物线检测方法。采用Canny边缘检测算子进行边缘提取,对提取到的边缘点采用最小二乘法确定抛物线方程的系数,通过对二次曲线施加抛物线约束,使得算法不需要经过迭代就能得到最优解,并对抛物线方程进行坐标变换推导,计算出抛物线的参数。仿真数据实验和实际图像的实验表明:该算法能够检测到光学图像中的抛物线目标特征,各个参数的计算精度均在98%以上。     

A two-dimensional vision system with optical ranging capability

This paper describes an optical three-dimensional (3D) camera developed by combining a two-dimensional intensity image and non-contact time of flight (TOP) range image. The camera system consists of an amplitude modulated light source, an image dissector tube, a phase measuring circuit, and a host PC for system control. A semiconductor diode laser light source or a white light source and Kerr cell is used to generate continuously high frequency modulated light for bulk illumination of the scene to be viewed. A compact, 25 mm diameter image dissector camera (Hamamatsu N2730 or R4193) is used as a receiver to detect the modulated light and generate both a two-dimensional intensity image and a range image. The range is obtained by measuring the phase between the received signal and the transmitted signal. With the large detection area and the random access property of the image dissector camera, it is easy to scan systematically and electronically the illuminated area, within the frame, to identify the required object. Three different modulated frequencies (between 10 MHz and 45 MHz) have been used to obtain a range accuracy of 4 mm, over a distance of 10 m, within a time period of 10 ms per pixel. A demonstration of the three-dimensional vision system has been given, having an update time of 1 s during which a high resolution intensity image (300 × 300 pixels) is produced together with a limited number (100) of range measurements obtained for important features.     

单摄像机虚拟立体视觉测量技术研究

以双目立体视觉传感器三维测量模型为基础,提出了一种用于测量空间三维点坐标的低成本单摄像机模型。该模型利用光学成像,把单摄像机镜像为一对虚拟摄像机,在一个CCD像面上采集到同一物体存在视差的两幅图像,从而恢复空间点的三维信息。讨论了单摄像机传感器测量空间三维点坐标的基本原理,建立了单摄像机传感器的测量模型,克服了双摄像机系统中成本高、切换采集左右摄像机的图像使检测速度减慢等诸多缺陷,为空间三维点的精密测量提供了经济、快速、有效的测量途径。实验表明．传感器可实现约0．8％的相对测量精度．证明了本方案合理、有效。     

基于多目视觉技术的多点触摸屏系统的实现方法

设计了一种基于多目视觉技术的多点触摸屏系统。系统使用三个摄像机分别置于显示屏的左右两端和前方,利用计算机图像处理技术使普通显示屏具备触摸功能。为了有效减少图像处理时间,提高指尖坐标精度,提出了基于卡尔曼滤波器的指尖跟踪算法,通过卡尔曼滤波器预测指尖位置并以预测指尖点为中心划出矩形ROI(Region of Interest,感兴趣区域)进行指尖检测算法处理。实验证明,该系统能实时运行在嵌入式平台,并可实现较高的点击成功率,对于实现基于计算机视觉的触摸屏技术有着重要的研究意义。     

Real and Pseudo Pedestrian Detection Method with CA-YOLOv5s Based on Stereo Image Fusion

With the development of convolutional neural networks, the effect of pedestrian detection has been greatly improved by deep learning models. However, the presence of pseudo pedestrians will lead to accuracy reduction in pedestrian detection. To solve the problem that the existing pedestrian detection algorithms cannot distinguish pseudo pedestrians from real pedestrians, a real and pseudo pedestrian detection method with CA-YOLOv5s based on stereo image fusion is proposed in this paper. Firstly, the two-view images of the pedestrian are captured by a binocular stereo camera. Then, a proposed CA-YOLOv5s pedestrian detection algorithm is used for the left-view and right-view images, respectively, to detect the respective pedestrian regions. Afterwards, the detected left-view and right-view pedestrian regions are matched to obtain the feature point set, and the 3D spatial coordinates of the feature point set are calculated with Zhengyou Zhang’s calibration method. Finally, the RANSAC plane-fitting algorithm is adopted to extract the 3D features of the feature point set, and the real and pseudo pedestrian detection is achieved by the trained SVM. The proposed real and pseudo pedestrian detection method with CA-YOLOv5s based on stereo image fusion effectively solves the pseudo pedestrian detection problem and efficiently improves the accuracy. Experimental results also show that for the dataset with real and pseudo pedestrians, the proposed method significantly outperforms other existing pedestrian detection algorithms in terms of accuracy and precision.     

Radial stereo imaging system for three-dimensional reconstruction

In this paper, we propose a radial stereo imaging system for three-dimensional reconstruction of an object. This system uses a novel optical device which consists of a camera and two reflected mirrors. It can reconstruct three-dimensional model of an object by capturing one image. Depending on the reflection of the mirrors, each world point in the scene has a pair of points which sit in the inner and outer of the captured image, respectively. We firstly analyze the imaging constraint of the proposed system, and then derive the geometrical relationship between world point and image point pair to calculate the three-dimensional coordinate values of an arbitrary point on the object, which means that the three-dimensional model of the object can be recovered. Experiments are carried out to prove our design, and the results show that our radial stereo imaging system is convenient and effective for the three-dimensional reconstruction of an object. This system can be used in variety of applications, such as virtual reality and commodity show.     

聚焦形貌恢复方法模型设计

采用基于拉普拉斯算符聚焦形貌恢复方法,提出了模拟目标深度测量的数值模型。数值模拟的核心是基于通过几何光学预测的理想图像的卷积与透镜广义孔径函数的多色点扩散函数,即用聚焦误差替代抛物线圆柱形貌或高斯函数。该模型可以使用基于聚焦形貌恢复方法的传感器真实组件参数、光源光谱、光学系统离差、相机的光谱灵敏度。提出了光学系统离差（消球差、消色差、色差）对确定目标表面形貌的精确度和可靠性的影响。结果表明,该模型可以有效提高实验效率,缩短时滞,降低成本。     

3D Reconstruction of Large-Depth Objects Using Fresnel Diffraction of a Ronchi Grating and Simultaneous Recording of Reference and Object Fringes

We propose an active triangulation based range-finding system that, besides its simplicity, has two advantages over similar existing systems. First, the system can acquire range data for large-depth objects since it generates an illumination pattern with large-depth of focus using Fresnel diffraction from a Ronchi grating projected over the entire object surface. Secondly, the system does not need preliminary recording of reference fringes since it displays simultaneously, on a single image, both reference and object deformed fringes. This system offers other properties: it is mechanically simple, uses everyday optical components, is low cost, can be operated via a PC and uses relatively simple image analysis software. Three dimensional reconstructions are illustrated for various objects having 10 cm maximum depth variations. The depth accuracy (typically 0.5 mm for an object located at 350 mm from the camera) is comparable with other systems, but a certain compromise has to be accepted in acquisition time (typically 1 minute). This system can find applications in research laboratories as a tool to provide range data with relatively good depth accuracy.     

多角度耦合分幅相机光学系统设计

为实现高速相机的分幅功能,本文提出一种采用多角度耦合分幅方式的高速相机光学系统。该系统分幅结构采用多组相同的光学系统,在平行于物面的圆周上均匀分布,分别从不同角度拍摄同一物面,在保证各组系统的物方视场相同的情况下,每组光学系统的光轴与物平面的夹角均相同,通过优化设计得到全视场的最佳成像。根据需求,使用光学设计软件设计了多角度耦合四分幅成像中长焦光学系统并绘制三维立体仿真模型,分析了每组像面像质、照度以及畸变等相关参数,调制传递函数MTF在频率为50 lp/mm处不低于0.5,F数为2,畸变小于0.4%,相较于常用的棱镜和反射棱锥分光方式,无需额外分光结构,像面照度提高4倍以上。结果表明成像质量理想,分幅相机系统各像面所成像一致性高。     

基于生物视觉特征和视觉心理学的视频显著性检测算法

提出了一种空域和时域相结合的视频显著性检测算法.对单帧图像,受视觉皮层层次化感知特性和Gestalt视觉心理学的启发,提出了一种层次化的静态显著图检测方法.在底层,通过符合生物视觉特性的特征图像（双对立颜色特征及亮度特征图像）的非线性简化模型来合成特征图像,形成多个候选显著区域;在中层,根据矩阵的最小Frobenius-范数（F-范数）性质选取竞争力最强的候选显著区域作为局部显著区域;在高层,利用Gestalt视觉心理学的核心理论,对在中层得到的局部显著区域进行整合,得到具有整体感知的空域显著图.对序列帧图像,基于运动目标在位置、运动幅度和运动方向一致性的假设,对Lucas-Kanade算法检测出的光流点进行二分类,排除噪声点的干扰,并利用光流点的运动幅度来衡量运动目标运动显著性.最后,基于人类视觉对动态信息与静态信息敏感度的差异提出了一种空域和时域显著图融合的通用模型.实验结果表明,该方法能够抑制视频背景中的噪声并且解决了运动目标稀疏等问题,能够较好地从复杂场景中检测出视频中的显著区域.     

利用无人机多源传感器估算马铃薯植株氮含量

快速准确地获取作物的植株氮含量(PNC)信息,是农业精细化管理的关键和数字农业发展的研究热点。近年来,随着无人机和传感器技术的发展,利用多种传感器信息监测作物理化参数逐渐引起国内外学者的关注。以马铃薯为研究对象,首先,基于无人机获取了马铃薯现蕾期、块茎形成期、块茎增长期、淀粉积累期和成熟期的高光谱影像和数码影像,同时采集各生育期的地面数码影像,并实测了株高(H)、PNC和11个地面控制点(GCPs)的三维空间坐标。其次,利用无人机数码影像结合GCPs生成试验区域的数字表面模型(DSM),分别从无人机数码影像和DSM中提取马铃薯的地面覆盖度(VC_uav)和株高(H_dsm),并利用地面数码影像计算的覆盖度(VC)和实测H验证提取的VC_uav和H_dsm的精度。然后,根据高光谱反射率数据计算绿边参数(GEPs),构造GEPs×H_dsm×VC_uav,GEPs/(1+VC_uav),(GEPs+VC_uav)×H_dsm和GEPs/(1+H_dsm) 4种融合特征参数(FFPs),对高光谱影像信息和数码影像信息进行融合。最后,将各生育期提取的GEPs和构造的FFPs分别与PNC作相关性分析,筛选最优绿边参数（OGEP）和最优融合特征参数(OFFP)构建5个生育期的PNC线性估算模型,并根据相关性较高的GEPs和FFPs利用偏最小二乘(PLSR)和人工神经网络(ANN) 2种回归方法构建PNC的多参数估算模型,结果表明：(1)基于无人机数码影像提取的H_dsm和VC_uav具有较高的精度,可以代替实测H和VC估算作物理化参数。(2)与GEPs相比,前4个生育期,构造的大部分FFPs与PNC的相关性更高,能更好地反映马铃薯的氮营养状况。(3)马铃薯5个生育期,OFFP估算PNC的效果优于OGEP。(4)与单参数模型相比,基于GEPs和FFPs利用PLSR和ANN 2种方法构建的模型精度和稳定性均明显提高,其中,以FFPs为模型因子利用ANN方法构建的模型效果最好。该研究表明融合高光谱绿边参数和高清数码相机传感器提取的株高和覆盖度信息能显著提升PNC的估算精度,可为马铃薯氮营养状况的动态无损监测和多源传感器信息的应用提供参考。     

准同心广义经纬相机的成像模型及高精度标定

对于大尺度运动目标的参数测量,固定相机存在着视场与空间分辨率之间的矛盾,并且当测量空中或海上目标时相机的标定也十分困难。跟踪式相机(如光电经纬仪)能解决此问题,但存在体积大、成本高和操作复杂等缺点。结合两者特点取长补短,将相机固定在二维旋转平台上,尽可能实现同心放置,组成准同心广义经纬相机进行中远场高精度测量,测量过程中转台实时跟踪目标并为相机提供外参数。此方法并没有光电经纬仪非常严格的同心同轴的安装要求。在合理假设基础上建立了经纬相机成像模型,提出了线性求解及平差优化相机参数的方法。大量仿真与实际测量验证了模型和标定方法的正确性和高精度。广义经纬相机测量系统组合巧妙、装拆和操作简单、体积小、成本低、可全视场测量且测量精度高,有广泛且重要的应用前景。     

望远系统像倾斜的数字图像测量技术的研究

利用CCD摄像机采集像倾斜分划图像,建立线性回归数学模型,通过"目标移动法"和"亚像素定位边缘法"实现望远系统像倾斜的高精度自动测量。该方法可消除传统像倾斜测量方法中的主观误差的影响,图像处理误差小于10″,使测量的结果客观化,易于实现光电自动测量。     

Chan-Vese模型两相自适应窄带的阴影区检测方法

为得到快速高精度的声呐图像阴影区检测效果,提出Chan-Vese模型两相自适应窄带检测方法。利用各向异性二阶邻域马尔可夫模型估计声呐图像的纹理特征参数,实现原始图像平滑去噪;由块方式的k-均值聚类算法确定图像的初始两类分割,初步确定阴影区大致位置,并根据此大致位置,自适应初始化零水平集函数,来减少人为干预,提高检测速度;在此基础上,提出建立Chan-Vese模型两相窄带水平集进行声呐图像检测,完成局部寻优,排除全局图像中孤立区对检测的影响,使阴影区检测结果更加精确。通过对真实声呐图像的检测实验结果分析,验证提出的检测方法能够去除原始图像的部分噪声,提高检测精度和速度,有一定的自动性和适应性。      

Phase-coded aperture for optical imaging

For the optical spectrum region, we describe a novel phase-coded aperture imaging system that can be used in a computational imaging camera. The optical design includes a phase-only screen followed by a detector array. A specific diffraction pattern forms at the detector array when the wavefront from a point source object passes through the phase screen. Since diffraction effects cannot be ignored in the optical regime, an iterative phase retrieval method is used to calculate the phase coded screen. Correlation type processing can be applied for the image recovery. Computer simulation results are presented to illustrate the excellent imaging performance of this camera.     

Imagery of local defects in multilayer components by short coherence length interferometry

We propose to image local defects inside multidielectric optical components by using a special configuration Linnik interference microscope, along with a CCD camera and a dedicated detection, to extract the amplitude scattered by the defects in the interference image. The use of a short coherence length source allows one to obtain interference only from a thin slice (~1microm) within the observed object. The object is tilted to permit the use of a dark-field configuration to enhance the defect contrast. We describe the experimental setup and the detection scheme. Images that exhibit local point defects on the interfaces of various multilayer optical components (laser mirrors) are presented.     

基于光流的运动小目标检测算法

运动目标的检测是目标识别与跟踪的关键技术之一。光流技术是一种以物体的运动特征来检测目标的方法,它的提出为运动小目标的检测开辟了新的空间。在一个搜索跟踪系统中使用光流技术检测和跟踪空中小目标,目标大概为5～10个像素,而且背景复杂,相机抖动,普通分割算法无法得到小目标。在目标的运动明显异于背景的情况下,通过利用基于光流的目标检测算法来检测出小目标,同时运用高斯金字塔模型,提高算法的运算速度。试验结果表明提出的基于光流的检测算法在背景运动的红外图像中取得了较好的效果。     

Improved spatial phase detection for profilometry using a TDI imager

Performance of a time delay and integration (TDI) imager is analogous to a drum/periphery camera. Coupled to a modulated line generating laser diode, it is possible to use a TDI imager to encode the shape profile of a rotating cylindrical object as deformed gratings in a digital image. Among the various methods for extracting the phase from a deformed grating, the spatial phase detection (SPD) method is fast and easy to use. Errors are introduced in this method if the phase gradient is high over a grating interval. Here we propose a method to reduce the phase gradient by using a TDI feature which makes it possible to record images at variable lateral magnification.     

移动特征靶标的摄像机径向畸变标定

针对成像测量系统中镜头径向畸变影响测量精度的问题,提出了一种基于物面移动同心圆特征靶标的径向畸变标定方法。该方法先将固定在二维精密平台上的同心圆靶标置于垂直物面的特定位置,然后采集靶标图像,同时用最小二乘法以拟合得到的圆直径为条件,按一定方式移动特征靶标,直到拟合值达到极值或者在一定误差范围内。记录该幅图像,则其拟合得到的圆心坐标便是畸变中心,同时利用该幅图像,根据等差值半径和摄像机成像模型的半径的成像关系求出其畸变多项式系数。为提高特征靶标的移动效率,提出了坐标轮换最优化移动的方案。实验结果表明,该方法对畸变中心的标定精度可达0.6pixel,畸变多项式系数有效数字重复误差小于0.02,并可实现两者的一靶标定,且利用该法获得的参数能实现对畸变图像的准确校正。