RGB-D SLAM增强现实原木检尺系统构建与测试

将内嵌有ToF相机、面阵相机及IMU的智能手机作为硬件系统,RGB-D SLAM技术实时获取的深度图、位姿等为数据源,构建了RGB-D SLAM增强现实楞堆原木检尺系统。首先设计了基于ToF影像实时估计RGB影像像素深度的方法,实现对待测原木端面几何坐标的初步估计;其次,设计了散形分区去噪算法实现原木端面点云的精确过滤,设计了原木端面曲率估计算法实现对过滤点云可靠性判别;然后,基于PCA等算法实现原木长、短直径方向向量估计,并基于该向量对原木长、短直径进行了估计;最后,以所构建算法为基础在智能手机平台上搭建了增强现实楞堆检尺系统,实现智能手机对原木进行实时检尺、增强现实场景对测量结果实时监督。新型检尺系统通过对6个楞堆334根原木进行了检尺实验,以评估该设备的测量精度。结果显示：原木平均直径估计值的偏差及均方根误差分别为-0.13 cm(-0.35%)及1.05 cm(3.34%);原木径阶化直径估计值的偏差及均方根误差分别为-0.10 cm(-0.22%)及1.33 cm(4.43%);原木材积估计值的偏差及均方根误差分别为-0.007 m³(-0.27%)及0...     

Keeping watch on the unwatchable: technological solutions for the problems generated by ecosystem-based management

Ecosystem‐based management is an emerging paradigm influencing the management of commercial fisheries. Increasingly, developed nations are adopting explicit legislation and policy governing the assessment and management of their fisheries against criteria of ecological sustainability. Yet the ability to evaluate ecosystem impacts of fisheries is compromised by a general lack of understanding of marine ecosystem function (beyond the population level) and a lack of robust and practical indicators for ecosystem health and management. Recent technological advances can assist in developing criteria, including structural analyses of seafloor communities potentially impacted by fishing gears (e.g. demersal trawling). Similarly, advances in fishing gear technology, including improved selectivity and the development of gears which have a more benign environmental impact, can mitigate some of the ecological impacts of fishing. Such technological advances are summarized in the context of contemporary fisheries management.     

基于视频处理的分层流持液率测量

截面持液率是气液两相流的基本参数之一。在气液两相实验系统上进行实验,采用高速摄像机系统获取了水平管气液两相流流型视频,提出了一种针对两相流动视频实现持液率检测的新方法。该方法针对自然光下拍摄的气液两相分层流图像气液区域与气液界面模糊、边缘难以直接提取的难点,采用新的递变规律加强图像的灰度使画面效果清晰化,通过中间线计算法提取气液边界线,进而通过提取的气液界面距离管底高度数据,计算管路持液率,其中包括同一帧图像不同截面的持液率以及同一截面持液率随时间的变化情况。实践证明：在自然光干扰较大的拍摄条件下,该方法能够获得连续、清晰的气液相边界,可推广应用于其他流型图像的分析和检测。（图9,参8）。     

Detection of antigenic proteins expressed by lymphocystis virus as vaccine candidates in olive flounder, Paralichthys olivaceus (Temminck & Schlegel)

Although the major capsid proteins (MCPs) of lymphocystis disease virus (LCDV) have been characterized, little is known about the host-derived immune response to MCPs and other LCDV antigenic proteins. To identify antigenic proteins of LCDV that could be used as vaccine candidates in olive flounder, Paralichthys olivaceus, we analysed the viral proteins responsible for its virulence by applying immuno-proteomics. LCDV proteins were separated by one-dimensional gel electrophoresis, transferred to polyvinylidene difluoride membrane, and probed with homogeneous P. olivaceus antisera elicited by LCDV natural infection and vaccination with formalin-killed LCDV. Four immune-reactive proteins were obtained at 68-, 51-, 41- and 21 kDa using antisera collected from natural infection while two proteins at 51- and 21 kDa exhibited response to antisera from vaccinated fish, indicating that the latter two proteins have vaccine potential. Using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry and nanoelectrospray MS/MS, the 51 and 21 kDa proteins were identified as MCP and an unknown protein, respectively.     

广角镜头的摄像机非线性标定技术

在路面检测的计算机视觉系统中,采用了线结构光照明和摄像机采集外界三维信息。为了实现在大范围内的测量,摄像机采用了广角镜头摄像,同时也引入了比较大的图像畸变。为了精确标定摄像机,采用比较接近实际情况的几何模型,并采用非线性迭代解方程的方法对摄像机进行标定,求解摄像机的内外参数。实验精度可以达到0.1mm。该方法简洁实用,可以采用这种方法在实际当中对系统进行标定。     

数码照相法估算植被盖度

植被盖度是表征生态系统特征的重要参数之一,其准确测度对于分析土壤侵蚀强度、生态环境现状具有重要意义。该文选择黄土高原退耕还林还草区野外测量的25个典型样方, 探讨了数码照相法估算植被盖度的模型和方法,提出了中心投影引起的边缘畸变问题和植被信息识别问题的有效解决方案。前者采用控制点几何校正的方法解决,后者基于目视判读、统计汇总红(R)、绿(G)、蓝(B)三基色的DN（digital number）值分布规律,采用编程方式解决。研究结果表明,植被信息的三基色组成分别符合DNGDNRDN B、DNGDNBDNR和DNBDNRDNG 3种规律,非植被信息的三基色构成遵循DNRDNGDNB的规律,且采用上述规律,编程得到草地样方的植被盖度精度较高。可见,该文提出的估算植被盖度的数码照相法可移植到数码相机中,具有明显的应用价值。      

EFFECT OF REGION OF INTEREST SELECTION AND UPTAKE MEASUREMENT ON GLOMERULAR FILTRATION RATE MEASURED BY 99mTc-DTPA SCINTIGRAPHY IN DOGS

Determinations of different methods of measurement of uptake of 99mTc-DTPA using scintigraphy of glomerular filtration rate (GFR) were made from 29 studies on 10 healthy beagle dogs. GFR was measured by calculating the percentage dose uptake (integral method) and rate of uptake (slope method) of 99mTc-DTPA using manual kidney regions of interest (ROI) and automatic kidney and background ROIs at different time periods of the uptake phase. These results were compared using linear regression analysis to the GFR obtained from 99mTc-DTPA plasma clearance using multiple blood samples. The best correlation coefficient between percentage DTPA uptake and GFR by DTPA clearance (r = 0.84, P < 0.001) was derived from time intervals between 30s-120s with a perirenal background ROI at 1 or 2 pixels out from the kidney ROI using automatic kidney ROI at 20% threshold. With the slope method, the best correlation coefficient (r = 0.85, P < 0.001) was obtained from time intervals between 30s-peak with the background ROI at 2 pixels out from the kidney ROI using automatic ROI at 35% threshold. The offset was higher, and the correlation varied more with different ROIs and the method was unreliable at time intervals extending beyond the peak radioactivity. Manual kidney ROIs with automatic background ROIs had slightly lower correlations. With DTPA renography both integral and slope uptake method with automatic kidney and background ROIs are accurate methods to estimate the GFR, but that the integral method is much more stable to variations in ROI size and the duration of the uptake phase of the renogram.     

基于RGB-D SLAM手机的森林样地调查系统研究

基于RGB-D SLAM手机构建了森林样地调查系统,该系统实现了样地构建、每木检尺及林分/样地参数的估计功能,并在测量过程中使用增强现实展示测量结果,且提供了重新测量的交互方式,使观测者在观测过程中能够检测结果的可靠性,并保证所获取样地信息的完整性。该系统在18块半径为7. 5 m的圆形样地中进行了测试。结果显示,平均胸径估计值的偏差(BIAS)及均方根误差(RMSE)分别为0. 36、0. 69 cm,平均树高估计值的BIAS及RMSE分别为0. 06、0. 63 m,蓄积量估计值的BIAS及RMSE分别为8. 595 9、25. 735 8 m3/hm2,横断面积估计值的BIAS及RMSE分别为0. 949 7、1. 987 3 m2/hm2,株树密度估计值的BIAS及RMSE分别为-3、13株/hm2,坡度估计值的BIAS及RMSE分别为0. 30°、0. 88°,坡向估计值的BIAS及RMSE分别为-0. 44°、7. 61°。其中,坡向估计具有较大的RMSE,是由于当坡度较小时,即使SLAM系统估计位姿有较小漂移,仍会导致该值产生较大偏差,但整体而言坡向仍是无偏的。     

Comparison of four methods for estimating leaf area index based on terrestrial three-dimensional laser scanning

The leaf area index (LAI) of 16 sample plots was estimated based on terrestrial three-dimensional laser scanning. The point-cloud data of stand canopy were first scaled and projected onto a hemisphere according to Lambert azimuthal equal-area projection or stereographic projection, and the resulting hemispherical point-cloud images were used to extract the canopy porosity coefficients. Then, single-angle inversion and Miller formula inversion methods were used, respectively, to calculate the effective leaf area indices with canopy porosity coefficients. Results showed that the effective LAIs estimated by single-angle inversion method with Lambert projection and stereographic projection were within the range of 2.14~5.36 and 1.83~4.67, respectively. The effective LAIs obtained by Miller formula inversion method with Lambert projection and stereographic projection were within the range of 1.84~4.67 and 1.68~4.34, respectively. As a comparison, the LAI measured with a fish-eye camera ranged from 1.55 to 3.87. The LAI values estimated with four different calculation methods were linearly correlated with those measured by a fish-eye camera. The highest coefficient of determination (R²) 90.28% was obtained by the Miller formula inversion method combined with stereographic projection, and Duncan’s new multiple range test also further showed that this method had a relatively higher precision compared to other three methods.     

一种车载双目视觉系统的标定

为了在线感知汽车行驶前方道路,搭建车载双目视觉系统,从不同拍摄位置和角度获取多组标定板图像。利用传统标定方法,在MATLAB应用程序Stereo Camera Calibrator工具箱中对图像进行校准及合理性验证,最终得到左右两个相机需要标定的内参数和外参数。标定结果将为视觉系统在车辆环境感知方面的研究奠定基础。