THE GROUND CENSUSES ON RED-CROWNED CRANE IN KHANKA LAKES DISTRICT OF CHINA

Coordination of the aerial servey, the author carried ground census on the population size of Red-crowned crane(Grus japonensis) in Khanka Lakes District from 1984 to 1986. According to the census in spring of 1984, the total number of Red-crowned crane is 72 individuals and 20 nests. So that this ground census has recorded 59 individuals more than the aerial survey of 13 individuals, and the percentage of nesting individuals in this flock is 55.6%, which is obviously higher than 36.9% of the average ratio in China. After this, in fall of 1984 and in spring of 1985 and 1986, we repeated the same censuses three times in the same districts, the results indicated that the flock of the crane in Khanka Lakes District is stable basically, but some individuals interchange with opposite flock of USSR because of gcgraphical adjoin.So, we believe this flock in China and that flock in USSR belong to the same population and suggest two countries should act in concert to studing on this population in future. Meanwhile, the aerial servey cooperate with ground census, it is the best way for reducing missing crane and achieving to high precision of census. From this, total number of Red-crowned crane in China is 542 individuals (483+59 ind ividuals).

基于特征优化的Census立体匹配方法

针对传统Census立体匹配算法在弱纹理和边缘区域匹配精度较差的问题,提出一种基于特征信息优化的代价计算方法,在窗口中融入更多的差异信息以获得更精确的像素视差值。随后采用多方向路径独立的线扫描优化计算聚合代价以进一步提高匹配精度。为获得更好的遮挡区域匹配效果,提出一种基于差异填充的视差优化方法,对遮挡像素进行识别和视差填充。为提高算法的效率,提出一种基于降采样策略的算法运行模式,通过缩小视差搜索范围以减少硬件负荷。最后以五组标准图像为输入进行改进Census算法性能检验,结果显示,平均误匹配率为6.12%,较改进前降低了2.45%,算法效率平均提升17.7%。

改进Census变换与特征融合的立体匹配算法

针对局部立体匹配算法匹配精度较低问题,提出一种改进Census变换与特征融合的立体匹配算法。首先,使用变换窗口的邻域像素信息代替中心像素,解决传统Census变换过度依赖窗口中心像素问题;其次,引入图像的颜色信息与梯度信息构建融合代价计算函数,提高初始匹配代价的可靠性;为建立邻域像素间联系,引入单向动态规划思想进行代价聚合;最后,提出一种基于八方向的视差空洞填充方法对视差图进行优化。实验结果表明:该算法在Middlebury数据集上的平均误匹配率为3.77%,优于其他改进Census变换方法,具有较高的匹配精度。

基于双目视觉的输送带表面损伤定量分析方法

为了对输送带表面损伤进行定量分析以有效判断其损伤程度,并根据损伤程度选择检修方法和时间,为工作人员检修提供依据,提出了一种基于双目立体视觉的输送带表面损伤定量分析方法。该方法首先对双目相机进行相机标定,对输送带表面损伤图像进行图像处理校正、立体匹配,基于视差原理根据三角测量原理获取目标点的三维空间信息,通过数学计算模型对损伤程度进行定量分析,提出了改进的Census变换和SGM(semi-global matching)算法相结合的立体匹配算法。结果表明:提出的改进算法匹配效果有所提升,和Census变换算法及梯度算法进行对比,误匹配率分别降低了5.26百分点和3.92百分点,得出的视差图精度最高;通过提出的方法对输送带表面损伤进行定量分析,在输送带损伤长度、宽度、深度三者误差中,损伤深度误差最大为8.5%,仍在可接受范围。

基于双目立体视觉的连铸辊尺寸三维测量方法

针对连铸车间中,扇形段连铸辊尺寸人工测量效率低下的问题,提出了一种基于双目视觉系统的连铸辊尺寸测量方法。首先,对双目相机采集的图像进行预处理并采用Otsu法分割工件前后背景;接着,针对边缘检测精度不高的问题,将传统Canny边缘检测算法的梯度模板增加到8个以提取工件轮廓,结合多项式插值公式提取亚像素级别特征点;然后,在SAD与Census变换融合的立体匹配基础上引入RANSAC算法来消除错误匹配;最后,采用三角测量原理计算出零件的尺寸。实验结果表明,系统测量的平均相对误差为0.14%,测量方法具有较高的精度,其稳定性与精确性满足连铸辊的尺寸自动检测任务。

基于双目的目标快速测距算法

为保证测距精度的同时提高计算效率,针对特定的应用场景,提出一种在Census变换理论基础上改进的立体匹配算法。以降低计算的像素需求,优化测距方法为途径,针对双摄像头获取的视频图像,通过图像灰度化、像素值平均、像素定标比较以及像素点替换等图像分析方法,实现在特定应用场景下提高双目摄像头的数据计算效率,加快硬件的运算速度,达到算法快速、稳定、有效地运行的目的。根据实验结果分析,该算法在图像噪声抑制上具有较好的效果,在实际应用中展现了较好的稳定性与较高的计算效率。

基于改进Census变换与分区域聚合的立体匹配

立体匹配作为三维重建的关键步骤之一,针对基于Census变换的立体匹配算法在视差不连续区域以及弱纹理区域匹配精度较差,且易受光照不均与噪声干扰的问题,提出一种基于改进Census变换与分区域聚合的立体匹配算法。在Census变换阶段令窗口中所有像素点根据灰度值大小进行降序排序,选取灰度值为中位数的像素点替代窗口的中心像素;然后进行Hamming距离计算得到结果,再加权融合亮度或光照差的绝对值;接下来将代价聚合阶段的图像通过梯度大小划分为边缘和平滑区域,分别使用十字交叉域完成初始代价聚合;最终采用WTA找到初始视差值,经过一系列视差优化步骤获得最终视差图。在MiddleburyV3.0版本的测试平台上对该算法进行评测,实验结果表明,总体像素的视差平均误差为8.26%,相较于AD-Census算法下降了2.95%,具有较高的匹配精度,且对光照及噪声有较好的鲁棒性。

基于阈值和权重Census变换的煤矿井下图像立体匹配算法

双目图像立体匹配是实现煤矿井下无人辅助运输车自主避障和视觉侦察的关键技术,但煤矿井下高粉尘、光照条件不稳定等因素会导致视觉传感器采集到的图像产生椒盐噪声,使得立体匹配的误匹配率很高。为此,提出一种基于阈值和权重相结合Census立体匹配算法,减少椒盐噪声对立体匹配影响。具体改进包括:①先对支持窗口内所有像素的灰度值进行阈值处理,去除支持窗口内灰度值极大和极小的像素点,解决异常值对加权融合的影响;②再将中心点对应的4条斜对角像素进行加权融合代替中心点像素,选择中心像素点4条斜对角线上的像素点,步长取1~3,根据对应的步长分别赋予0.7、0.2、0.1的权重,将这12个像素点中有效的像素点分别乘上对应权重,然后除以有效权重之和,得到经过加权处理后的中心像素点值,解决传统算法对Census变换窗口中心点依赖的问题,从而提高匹配精度。试验结果表明:基于阈值和权重Census代价计算全部区域的误匹配率相比传统Census算法降低了5.64%;相比基于均值Census算法降低了1.71%,且在不同噪声下全部区域的误匹配率相比传统Census算法降低了15.93%;相比基于均值Census算法降低了16.62%;而在非遮挡区域,算法的误匹配率相对于传统Census算法降低17.19%,相对于基于均值Census算法降低18.11%。所提出的基于阈值和权重相结合Census立体匹配算法有效的增强了抗噪声的鲁棒性,降低了误匹配率,提高了匹配精度。

基于改进Census变换融合AD立体匹配算法研究

在立体匹配算法中,针对Census算法对于图像深度不连续、弱纹理区域匹配精度低、易受噪声干扰的问题,提出了一种改进Census变换结合自适应窗口的方法,利用梯度信息实现窗口的自适应选择,并将灰度绝对值计算(Absolute Difference,AD)和改进后的Census变换相融合,提升了算法的精度和抗噪声能力,在代价聚合阶段,通过不同的自适应窗口提供不同的自适应阈值,最后通过赢家通吃的算法(WTA)进行视差优化,进而得到最终的视差图。利用该改进算法对Middlebury平台提供的标准图像进行实验表明改进后的算法误匹配率降低,并有很好的抗噪声能力。

Spatial patterns, driving forces, and urbanization effects of China＇s internal migration： County-level analysis based on the 2000 and 2010 censuses

China has witnessed unprecedented urbanization over the past decades. The rapid expansion of urban population has been dominantly contributed by the floating population from rural areas, of which the spatiotemporal patterns, driving forces, and multidimensional effects are scrutinized and evaluated in this study by using the latest national censuses conducted in 2000 and 2010. Analysis based on the county-level data comes to conclusions as follows. The spatial pattern of floating population has remained stable over the first decade of the new century. The top 1% cities with the largest floating population received 45.5% of all migrants in China. As the rapid development of mega-city regions, the coastal concentration areas of floating population tended to geographically united as a whole, whereas the spatial distribution of migrants within each region varied significantly. The migrant concentration area in the Yangtze River Delta was the largest and its expansion was also the most salient. However, the floating population has growingly moved into provincial capitals and other big cities in the inland regions and its gravity center has moved northward for around 110 km during the study period. The spatial pattern of floating population has been formed jointly by the state and market forces in transitional China and the impacts of state forces have been surpassed by those of market forces in the country as a whole. The attractiveness of coastal cities and counties to the floating population comes mainly from the nonagricultural employment opportunities and public services, reflecting that long-distance and long-term migrants have moved coastward not only to gain employment but also to enjoy city life. By contrast, in the central and western regions, places with a higher economic development level and at a higher administrative level are more attractive to floating populations, demonstrating that the state remains to play an important role in allocating economic resources and promoting regional development in inland China. As the main body of new urban residents, the floating population has contributed substantially to the elevation of the urbanization levels of migrant-sending and -receiving places, by 20.0% and 49.5% respectively. Compared with ex- tensively investigated interprovincial migrants, intra-provincial migrants have higher intention and ability to permanently live in cities and thus might become the main force of China＇s urbanization in the coming decades. The internal migration has also reshaped China＇s urban system in terms of its hierarchical organization and spatial structure.

Semi-urbanization and evolving patterns of urbanization in China: Insights from the 2000 to 2010 national censuses

Based on the prefecture-level data of the 2000 and 2010 national censuses, the spatial evolution of China's semi-urbanization is analyzed in this study. The stages of urbanization are re-examined by considering semi-urbanization. Nine types of urban development are presented according to the relations between semi-urbanization and urbanization, and China's urbanization is divided into five stages, namely, high incoordination, incoordination, low coordination, coordination, and high coordination. Results show that China's semi-urbanization rate varies significantly from one area to another; its order in 2010 from the highest to the lowest value was as follows: east, middle, west, and northeast. Urbanization and semi-urbanization rates in inland cities increase much faster than those in coastal cities. In addition, semi-urbanization displays a spatial pattern similar to that of urbanization across China, with the sole exception of the northeastern region. Through a spatial autocorrelation analysis, the spatial concentration of semi-urbanization is determined to be increasing. High-value concentration areas are expanding in the coastal east, whereas low-value concentration areas are growing in the northeast. Lastly, the evolution of China's urbanization model suggests a weakening trend of coordination between urbanization and semi-urbanization over the studied decade. Semi-urbanization can be viewed as a special production of China's hukou system, which restricts the permanent settlement of migrants in cities. As such, China's semi-urbanization trend is expected to exhibit a reversed U-shaped pattern as urbanization and citizenization develop.

匹配代价融合与分层迭代优化的半全局立体匹配

立体匹配是双目立体视觉中的核心步骤之一.在户外场景中,传统的立体匹配算法难以达到较高的匹配精度,并且边缘设备既要低成本又要高效率.针对实际应用中存在的上述问题,提出了一种改进的半全局立体匹配算法.首先,整体采用分层迭代的匹配策略,减少计算复杂度的同时可以提高立体匹配精度.其次,使用改进的代价计算方式使得初始代价更为准确.最后,使用并行优化加速计算.实验结果表明,该算法在KITTI2012和KITTI2015数据集上的误匹配率可以达到4.72%和6.04%,使用分辨率为1800×1500的图像测试效率,在256视差条件下,完全优化后的算法时间效率可以提高23.6倍.该算法可有效提高边缘CPU设备的立体匹配效率,并且视差图的误匹配率可以达到主流经典算法的水平.

基于改进Census变换的鲁棒立体匹配算法

针对局部立体匹配算法对噪声与光照变化敏感及在弱纹理区域匹配效果不佳的情况,提出一种基于改进Census变换与自适应参数引导滤波的立体匹配算法。该算法融合HSV通道值计算代价,对噪声与光照变化有较好的鲁棒性;在支持窗口内分别使用欧式距离加权与颜色加权获得Census变换参考值;集成AD与Census代价提高了单像素匹配代价的稳定性,降低了原Census变换对中心像素的依赖程度。在代价聚合过程中使用峰度系数对正则化参数进行自适应处理,通过视差计算获得视差图。在VS2017平台上对Middlebury数据库提供的图像对进行匹配实验,该算法在标准图像、加噪声图像及光照变化图像的各区域平均误匹配率分别是7.80%、10.72%和9.89%。结果表明:该算法可以降低误匹配率,同时能更好地适应噪声与光照变化。

基于融合代价和优化引导滤波的立体匹配算法

现有局部立体匹配算法与全局立体匹配算法相比,计算量更小、速度更快,能达到实时匹配的要求,但存在误匹配率较高、视差结果精度较低等问题。提出一种新的局部立体匹配算法,依据像素梯度信息、像素梯度的平均值及标准差计算多指标梯度代价,使算法对图像局部结构的表达更加全面。根据像素的灰度差异程度划分等级,提出7等级编码的精细化Census变换匹配代价,以有效识别图像信息之间的细微差异,弥补Census变换在相似纹理区域精度较低的不足。将多指标梯度代价和精细化Census变换匹配代价按一定权重进行融合,形成初始匹配代价,从而充分发挥两种代价的优越性。在代价聚合阶段,对引导滤波的线性系数进行自适应优化,解决固定参数引起图像局部过于平滑或平滑不足的问题,并使用优化后的引导滤波模型进行代价聚合,改善代价聚合的效果。使用胜者为王算法计算视差获得初始视差图,最后通过左右一致性检测和加权中值滤波进行视差优化,得到更为理想的视差图。在Middlebury V3立体评估平台上的实验结果表明,所提算法在非遮挡区域的加权平均误匹配率为15.7%,与Cens5、IGF、ISM等算法相比具有较高的精度。

基于自适应窗口和改进Census变换的半全局立体匹配算法

针对基于Census变换的立体匹配算法匹配精度不理想、易受噪声干扰的问题,提出一种自适应窗口结合改进Census变换的方法。通过自适应选择变换窗口大小,并在Census变换中引入噪声容限参数,且将其改进Census代价与RGB颜色差值代价加权融合,再采用四路径代价聚合策略完成初始代价聚合,最后用WTA算法计算初始视差后通过后处理优化得到最终视差。实验结果表明所提算法误匹配率较低,抗噪性能有较好提升。

基于主动双目立体视觉的干排渣机渣量检测方法研究

实现燃煤电站锅炉底部渣量在线检测,进而实现干排渣机冷却风量有效调整,对提高干排渣机运行的可靠性和经济性具有重要意义。提出一种基于主动双目视觉的炉渣体积流量在线检测方法,通过线型激光照射干排渣机输送带上炉渣表面,有效地增强了图像的亮度、纹理等特征,进一步利用改进的自适应Census算法实现了双目视觉系统准确匹配。在此基础上,开发了炉渣体积流量检测系统,开展了评价研究,并在某660 MW超临界机组上进行了现场应用。结果表明:该系统能够准确可靠的获得灰渣的深度信息,误差小于8%,整体精度较高。660 MW超临界机组现场应用结果进一步验证了渣量在线检测系统的可靠性和准确性,满足工业现场应用要求。

基于改进SGM的工件高度定位研究

针对工业环境中光照条件不稳定、图像纹理信息缺乏等因素导致目标视差图获取困难的问题,提出一种基于改进半全局立体匹配(semi-global matching,SGM)算法的工件高度定位方法。采用Census变换窗口内像素的平均值代替原变换的中心像素值,使算法对噪声更具鲁棒性,减少了原算法对中心像素的依赖。引入Sobel算子对图像边缘进行检测,依据左右视图的梯度差值划分像素点类别,对路径聚合中的惩罚系数进行自适应调整,通过后续优化处理,得到了最终视差图。为验证本文所述工件高度定位系统的可行性和有效性,搭建机器人实验平台,对所提算法进行实验验证。实验结果表明,本文的改进算法与传统SGM算法相比,误匹配率下降约为8.76%,匹配精确度显著提高,工件高度的平均定位误差约为0.708 mm,满足系统抓取要求。该研究为机械臂精确定位工件高度提供了思路。

Physical fitness before and during the COVID-19 pandemic:Results of annual national physical fitness surveillance among 16,647,699 Japanese children and adolescents between 2013 and 2021

Background:Limited nationally representative evidence is available on temporal trends in physical fitness(PF)for children and adolescents during the coronavirus disease 2019(COVID-19)pandemic.The primary aim was to examine the temporal trends in PF for Japanese children and adolescents before and during the COVID-19 pandemic.The secondary aim was to estimate the concurrent trends in body size(measured as body mass and height)and movement behaviors(exercise,screen,and sleep time).Methods:Census PF data for children in Grade 5(aged 10-11 years)and adolescents in Grade 8(aged 13-14 years)were obtained for the years 2013-2021 from the National Survey of Physical Fitness,Athletic Performance,and Exercise Habits in Japan(n=16,647,699).PF and body size were objectively measured,and movement behaviors were self-reported.Using sample-weighted linear regression,temporal trends in mean PF were calculated before the pandemic(2013-2019)and during the pandemic(2019-2021)with adjustments for age,sex,body size,and exercise time.Results:When adjusted for age,sex,body size,and exercise time,there were significant declines in PF during the pandemic,with the largest declines observed in 20-m shuttle run(standardized(Cohen’s)effect size(ES)=-0.109 per annum(p.a.))and sit-ups performance(ES=-0.133p.a.).The magnitude of the declines in 20-m shuttle run and sit-ups performances were 18-and 15-fold larger,respectively,than the improvements seen before the pandemic(2013-2019),after adjusting for age,sex,body size,and exercise time.During the pandemic,both body mass and screen time significantly increased,and exercise time decreased.Conclusion:Declines in 20-m shuttle run and sit-ups performances suggest corresponding declines in population health during the COVID-19pandemic.

基于动态自适应加权Census变换的立体匹配算法

为了解决匹配过程中在弱纹理区域的误匹配问题,提出了在不同搜索窗口下自适应加权的Census变换的立体匹配方法。采用Census变换得出搜索窗口内对中心像素影响不同的特异点,分别将邻域像素与中心像素灰度差值进行处理,得到搜索窗口内邻域特异点对中心像素的影响权重,使用权重值加权求和的方式代替计算汉明距离,得出初始匹配代价。为了增强视差图,采用4路径聚合方式与赢家通吃算法(Winner-take-all,WTA),利用子像素拟合、一致性检查、颜色信息加权中值滤波和视差填充等步骤优化视差。为了验证算法效果,将提出的自适应加权Census算法与ADSG、SGM、LPSM和SGBM2进行比较,在Middlebury 2014数据集上进行了测试实验。验证了所提算法的有效性,提高了匹配的准确率,视差图更加平滑、边缘清晰。

基于改进Census变换的双目立体匹配算法

针对传统Census变换窗口中心点灰度值易受噪声影响而导致全局匹配精度低等问题,设计了一种基于改进Census变换的双目立体匹配算法.首先采用双边滤波的模板值替代传统Census变换窗口中心点的灰度值,为了增强初始代价计算的可靠性,加入Sobel算子,将其与灰度绝对误差和算法的匹配代价进行代价融合;然后选择动态十字交叉域建立相邻视差的联系;最后运用赢家通吃策略选择最佳的视差,并采用左右一致性检测和引导滤波优化视差图.实验结果表明,在无噪声情况下,改进算法的平均误匹配率比传统Census变换的降低了约46%,含噪声情况下降低了约53%,说明改进算法有较好的匹配精度和抗噪能力.