基于Kinect传感器多深度图像融合的物体三维重建

物体的三维重建技术一直是计算机视觉领域研究的热点问题,提出一种利用Kinect传感器获取的深度图像实现多幅深度图像融合完成物体三维重建的方法。在图像空间中对深度图像进行三角化,然后在尺度空间中融合所有三角化的深度图像构建分层有向距离场（hierarchical signed distance field）,对距离场中所有的体素应用整体Delaunay三角剖分算法产生一个涵盖所有体素的凸包,并利用Marching Tetrahedra算法构造等值面,完成物体表面重建。实验结果表明,该方法利用Kinect传感器采集的不同方向37幅分辨率为640480的深度图像完成目标物体的三维重建,仅需要48 s,并且得到非常精细的重建效果。     

Two-Phase Image Inpainting: Combine Edge-Fitting with PDE Inpainting

The digital image inpainting technology based on partial differential equations (PDEs) has become an intensive research topic over the last few years due to the mature theory and prolific numerical algorithms of PDEs. However, PDE based models are not effective when used to inpaint large missing areas of images, such as that produced by object removal. To overcome this problem, in this paper, a two-phase image inpainting method is proposed. First, some edges which cross the damaged regions are located and the missing parts of these edges are fitted by using the cubic spline interpolation. These fitted edges partition the damaged regions into some smaller damaged regions. Then these smaller regions may be inpainted by using classical PDE models. Experiment results show that the inpainting results by using the proposed method are better than those of BSCB model and TV model.     

三维面形测量中基于模版匹配的变形条纹修复

在基于条纹投影和相位分析的三维面形测量中,由于被测物体表面标志点或复杂面形的阴影遮挡存在,会造成变形条纹局部区域的条纹数据缺失,影响相位和高度信息的最终重建,需要人为地对缺失图像信息进行修复。提出了一种新的缺失条纹数据修复方法——基于模版匹配的图像修复算法,通过图像中已有条纹信息(特别是与待修复区域周围相位信息相似度较高的已知条纹信息)对缺失的变形条纹信息进行估算,实现数据修复。该方法修复效果好,运算过程无需人为参与,便于计算机自动实现,尤其适合于待修复图像整体结构明显、纹理清晰图像的数据修复,有助于提高被测物体相位计算质量和在此基础上的三维面形重建质量。     

Fusion of 3D laser scanner and depth images for obstacle recognition in mobile applications

The problem of obstacle detection and recognition or, generally, scene mapping is one of the most investigated problems in computer vision, especially in mobile applications. In this paper a fused optical system using depth information with color images gathered from the Microsoft Kinect sensor and 3D laser range scanner data is proposed for obstacle detection and ground estimation in real-time mobile systems. The algorithm consists of feature extraction in the laser range images, processing of the depth information from the Kinect sensor, fusion of the sensor information, and classification of the data into two separate categories: road and obstacle. Exemplary results are presented and it is shown that fusion of information gathered from different sources increases the effectiveness of the obstacle detection in different scenarios, and it can be used successfully for road surface mapping.     

基于优化估计的深度图像修复与误差补偿方法研究

针对Kinect传感器在获取深度图像时存在深度值随机跳变的不准确性问题,基于最优估计的思想,提出卡尔曼滤波与多帧平均法相结合的图像修复方法。首先利用卡尔曼滤波对多幅深度图像进行修复处理,实现Kinect传感器在采集信息过程中随着时间递推,深度值的跳变逐渐趋于平稳的效果;然后基于多幅图像平均法确定最终的深度图像,解决了Kinect获取深度值存在误差导致的不精确问题。实验结果表明,该算法的均方根误差为38.102 5,平均梯度为0.471 3,信息熵为6.191 8,与单幅图像修复效果相比,得到的深度图像边缘更加清晰。     

基于局部方向编码的遥感影像平行边缘识别

针对高分辨率遥感图像中道路、建筑物和水域等的特征识别难题,提出了一种基于边缘局部方向信息的平行边缘自动识别算法。该算法首先定义平行边缘由一系列相互平行的短直线组成;然后提出了交叉点共线约束的8-邻域边界追踪和9-像素滑动窗口内直线检测算法,实现了边缘连续线条局部方向信息编码;最后通过分析连续线条结构及方向编码规律,提出了主元分析及方向一致性判别准则进行平行特征识别。实验结果表明,该算法能够有效提取高分辨率遥感图像中具有最近邻关系的平行直线和曲线特征,平均识别准确率在95%以上,但算法执行速度有待提高。     

Fast 3D modeling in complex environments using a single Kinect sensor

Three-dimensional (3D) modeling technology has been widely used in inverse engineering, urban planning, robot navigation, and many other applications. How to build a dense model of the environment with limited processing resources is still a challenging topic. A fast 3D modeling algorithm that only uses a single Kinect sensor is proposed in this paper. For every color image captured by Kinect, corner feature extraction is carried out first. Then a spiral search strategy is utilized to select the region of interest (ROI) that contains enough feature corners. Next, the iterative closest point (ICP) method is applied to the points in the ROI to align consecutive data frames. Finally, the analysis of which areas can be walked through by human beings is presented. Comparative experiments with the well-known KinectFusion algorithm have been done and the results demonstrate that the accuracy of the proposed algorithm is the same as KinectFusion but the computing speed is nearly twice of KinectFusion. 3D modeling of two scenes of a public garden and traversable areas analysis in these regions further verified the feasibility of our algorithm.     

一种基于Mumford-Shah模型的红外图像边缘检测方法

 Chan-Vese模型是一种优秀的简化Mumford-Shah模型。然而Chan-Vese模型是以两个同质区域为基础建立的,这并不符合红外图像的特点,导致直接应用该模型处理红外图像时可能失败。针对这一问题,提出了一种适用于红外图像边缘检测的改进Mumford-Shah模型,并对该模型中目标边缘的保持、停止准则的建立及算法速度的提高作了详细讨论。实验表明,改进Mumford-Shah模型能够克服Chan-Vese模型在对红外图像边缘检测时不能跨越过渡区域的缺点,有效地检测出目标边缘。     

Optical display of true 3D objects in depth-priority integral imaging using an active sensor

In this paper, we propose a system combining the pickup process using an active sensor and the display process using depth-priority integral imaging (DPII) system to display true three-dimensional (3D) objects within large depth through real and virtual image fields. The active sensor provides depth map and color images of 3D objects. Using captured depth map and original color images, elemental images are computationally synthesized and displayed optically in DPII system. Proposed system provides scaling of 3D scenes for true 3D object. To show the usefulness of proposed system, we carry out the experiment for true 3D objects of three character patterns and present the experimental results.     

Interpolation algorithm for recovering the missing phase values in 3D measurement

In the three-dimensional (3D) phase measurement, some marks are usually adhered to the object in order to make the 3D registration process faster and easier. As covered with marks, local phase data are missing and have to be interpolated later. Considering the phase distribution nearby the marks, a gradient estimate (GE) interpolation algorithm is provided here. This algorithm recovers one pixel's missing phase value with the average of the estimated values which is calculated by gradients in eight directions nearby. Since this algorithm is a local processing, the missing phase values should be interpolated from the edge of the marks to the center. In the computer simulation and the practical experiment, compared with the same-size neighborhood mean (NM) algorithm and the Gerchberg-Saxton (GS) algorithm, this new algorithm achieves very good fit results with the least time. So it can be used as a practical tool for automatic missing phase interpolation.     

Defects’ geometric feature recognition based on infrared image edge detection

Edge detection is an important technology in image segmentation, feature extraction and other digital image processing areas. Boundary contains a wealth of information in the image, so to extract defects’ edges in infrared images effectively enables the identification of defects’ geometric features. This paper analyzed the detection effect of classic edge detection operators, and proposed fuzzy C-means (FCM) clustering-Canny operator algorithm to achieve defects’ edges in the infrared images. Results show that the proposed algorithm has better effect than the classic edge detection operators, which can identify the defects’ geometric feature much more completely and clearly. The defects’ diameters have been calculated based on the image edge detection results.     

Pixel-by-pixel absolute phase retrieval using three phase-shifted fringe patterns without markers

This paper presents a method that can recover absolute phase pixel by pixel without embedding markers on three phase-shifted fringe patterns, acquiring additional images, or introducing additional hardware component(s). The proposed three-dimensional (3D) absolute shape measurement technique includes the following major steps: (1) segment the measured object into different regions using rough priori knowledge of surface geometry; (2) artificially create phase maps at different z planes using geometric constraints of structured light system; (3) unwrap the phase pixel by pixel for each region by properly referring to the artificially created phase map; and (4) merge unwrapped phases from all regions into a complete absolute phase map for 3D reconstruction. We demonstrate that conventional three-step phase-shifted fringe patterns can be used to create absolute phase map pixel by pixel even for large depth range objects. We have successfully implemented our proposed computational framework to achieve absolute 3D shape measurement at 40 Hz.     

基于视觉SLAM和目标检测的语义地图构建

语义信息对于移动机器人理解环境内容、执行复杂任务至关重要,针对ORB-SLAM2构建的点云过于稀疏、缺乏语义信息、点云所占存储空间大等问题,提出将目标检测算法与视觉SLAM(同时定位与地图构建)技术紧密结合,构建环境的稠密点云语义地图.首先,通过目标检测网络YOLO v3及对象正则化准确获取物体的2D标签,并经过ORB...     

高能闪光照相中Sobel算子的边缘检测方法

 针对高能闪光照相中投影图像的特点,对数字图像处理中具有代表性的Sobel边缘检测算法进行了分析,发现该算法存在检测出的边缘粗和人为进行阈值的选取会造成边缘点的误判的缺点;为此提出了一种阈值确定方法和基于最大梯度的定位准则算法。改进的算法实测结果表明：该方法有效地解决了Sobel算子检测出的边缘过粗的问题,得到的边缘较细,定位精确,提高了边缘检测精度,可用于高能辐射照相中的边缘提取。     

基于双目视觉的边缘检测算法

提出了一种基于立体视觉的物体边缘检测的方法。先对立体图像对进行基于图割的立体匹配方法求取场景的视差图，然后再用Canny的边缘检测方法对视差图进行边缘检测。立体视觉方法有效解决了单目视觉检测方法中的一些难点，利用了物体在空间的深度信息，对复杂背景下的物体和具有复杂纹理物体的边缘检测有很高的鲁棒性。实验结果表明该边缘检测方法优于传统的单目视觉边缘检测方法。     

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.     

一种改进的Sobel图像边缘检测算法

边缘检测在数字图像处理和计算机视觉中有着重要的应用。对数字图像处理中具有代表性的Sobel边缘检测算法进行了分析。针对该算法存在检测出的边缘粗且对噪声极其敏感的缺点,提出了一种改进算法。该算法对实际图像中出现的边缘类型进行了数学模型描述,然后把连续型的边缘模型作为研究对象,重新构造了对图像边缘方向进行检测的模板。针对Sobel边缘检测基于一阶导数极大值或二阶导数零交叉而带来的边缘定位准确度不高的缺点,对图像梯度图进行了细化处理。仿真结果表明:该算法对图像噪声干扰有较强的抑制能力,提取的边缘定位准确、结构细腻。     

真实场景下视频运动目标自动提取方法

视频运动目标跟踪逐渐成为研究热点并应用到军事民用等领域,为了能够从真实场景中快速准确地提取视频跟踪单运动目标或多运动目标,提出了一种新的运动目标自动提取方法。首先通过自适应阈值获得滤波后的相邻帧差值图像。其次,为了消除差值图像中噪声的影响,标记此二值图像的连通像素来检测出运动目标所在的区域,并与边缘检测出的空间信息结果比较得到运动目标模型。最后,将图像分成若干区域,在每个分区域内依次连接每个运动目标模型的最外围轮廓点,由此构成目标闭合轮廓。利用得到的连续边界,对运动目标进行提取。实验结果表明,该算法能够有效地自动提取速度不同的单运动目标,同时能够提取多运动目标。     

基于平稳小波变换的有噪弱目标检测方法的研究

利用平稳小波变换的多尺度边缘检测算法分别对模糊图像、低对比度图像和加入噪声的图像进行了边缘检测,验证了多尺度二次B样条小波的检测效果,也比较了三种小波局部模极大值方法在抗噪性、计算量及检测效果等方面的性能,并且针对对比度低,受噪声污染严重的目标图像,提出一种能够根据不同背景计算出自适应阈值的新方法,使其在抗噪的同时又能较好地提取出微弱目标边缘。实验证明,利用多尺度二次B样条小波边缘检测算法能有效地排除噪声干扰,准确地提取出微弱边缘,可以实现3%对比度下的有噪图像的目标探测问题。     

Mixed noise removal algorithm of EBAPS image under low illumination condition based on FPGA北大核心CSCD

In order to solve the problem that single median filtering and gaussian filtering algorithm is not effective in suppressing impulse noise and poisson noise simultaneously in low illumination image, and the edge detail protection is insufficient, an open and close mix-median-gaussian (OCMMG) filtering algorithm based on field programmable gate array (FPGA) was proposed. Firstly, the minimum four-direction difference was used to detect the anomaly degree of each pixel point, the weight was allocated according to the threshold of pulse noise discrimination, and the first step was filtering. Then, the four-direction edge detection algorithm was used to extract image edges, and the second step was filtered according to the set edge confidence characterization value. Finally, the images collected by electron bombarded active pixel sensor (EBAPS) under the condition of 1×10−3 lx illumination were processed by FPGA in real time. The experimental results show that the FPGA processing results are consistent with the software simulation processing results. Compared with the median filtering and gaussian filtering algorithm, the peak signal-to-noise ratio (PSNR) of the algorithm is improved by 3.23% and 16.34%, the structural similarity is improved by 14.66% and 33.86%, and the edge retention index is improved by 0.49% and 4.21%, respectively, which can effectively remove the mixed noise of EBAPS image and meet the real-time requirements. © 2022 Editorial office of Journal of Applied Optics. All rights reserved.