基于间隙度的无人机林地航拍图像序列拼接方法

无人机林地航拍图像具有的分辨率高、数据量大、边缘丰富的特点,造成了特征点配准中误匹配率的增加,因此本文提出了一种新的无人机林地航拍图像序列拼接方法。分形中的间隙度可用于描述图像区域纹理的粗糙程度,本文首先利用间隙度特征选取图像中局部图像块作为候选区域查找特征点,减少了待配准的特征点数量,提高了特征点配准正确率;其次,采用全局拼接技术变换图像,减少传统拼接中矩阵连乘产生的误差的积累和传播。实验中选取了不同拍摄高度的两组图像序列,将本方法与传统的全局SURF特征方法和降采样图像拼接方法进行了对比,结果显示本方法可以有效拼接图像,同时不会损失原图像的精度信息,并从视觉效果和均方根误差两个角度证明了本文方法优于其他两种方法。      

基于CCD拼接的航空相机小波自动调焦方法研究

为精确地进行航空相机调焦,获取高清晰度的航空影像,结合CCD拼接的特点,提出一种新的航空相机调焦方法,便于航空成像设备进行自动调焦。考虑人类视觉系统的特点设置权值,利用图像的小波分解系数构造清晰度评价函数;分析小波基函数的性质,对不同频率成分的影像进行清晰度评价,从而使评价函数对图像的变化具有适应性,通过比较得出了利用symlet2小波进行3层分解获得的评价函数的效果最佳。为了提高清晰度评价函数的灵敏度,选取了细节丰富的区域作为评价区域,详细地阐述了航空相机利用CCD拼接结构的特点进行调焦的算法实现过程。试验证明,该方法获得的评价函数比传统的小波方法灵敏度高,可以满足航空相机的需要,大大提高了地物目标的识别和提取的精度。     

低空林地航拍图像拼接的改进缝合线算法

目的图像拼接效果的优劣主要取决于图像配准和图像融合两个步骤。图像配准误差导致的错切以及图像序列间的视差导致的鬼影、重影问题可通过图像融合算法减小或消除。目前图像融合算法中最佳缝合线算法的综合效果较好, 但没有考虑到拼接图的正射效果, 并且无人机低空飞行时树高相对飞行高度比值较大, 这在获取正射影像时是不可忽略的干扰因素。传统的数字正射影像(DOM)是基于数字高程模型(DEM)对单张影像进行数字微分纠正进而拼接成整个区域的正射影像图。但是, 地形高程数据和植物的高度数据获取困难, 而恢复出来的地形和植物高度与实际数据存在误差, 造成DOM在局部边缘出现扭曲、模糊问题。本文提出一种不需要DEM数据, 仅用图像信息使拼接结果图保留正射投影的改进缝合线算法。方法首先对SURF特征检测、匹配与筛选, 用RANSAC算法求得的单应性矩阵确定相邻图像重叠区域; 然后基于重叠区中像素点位置与相邻两图像中心点位置的距离差可以反映正射效果的思想, 将距离差引入能量函数, 同时设计了动态权值参数用来平衡颜色、结构和距离三者的重要程度, 利用动态规划思想搜索得到最佳缝合线; 最后在缝合线两侧进行多频带融合生成类似正射影像的无缝拼接图。结果实验图像来源于不同样地、不同飞行高度, 在相邻两幅图像以及同一条带航线图像上将本文的改进缝合线算法与其他3种缝合线算法以及Pix4D生成的数字正射影像进行对比。实验结果表明, 本文提出的缝合线改进算法能保留正射投影, 视觉效果优于现有的缝合线算法, 在城镇建筑图像的实验中局部效果优于Pix4D。结论本文针对无人机低空林地航拍图像拼接重影问题和拼接结果由于视角不同而产生非正射影像区域的问题, 实现了一种不需要DEM数据进行数字微分纠正但能生成类似正射影像效果的改进缝合线算法。实验结果显示, 本文算法优于目前的最佳缝合线算法, 能够保留正射投影, 效果类似DOM, 并且在保证物体边缘清晰方面优于目前商用软件生成的DOM。这有利于更准确地计算林地的郁闭度, 估算林地植被覆盖面积, 对跟踪识别地表动植物也具有一定的帮助。除林地图像之外, 本方法也可以推广到其他需要保留正射投影的低空航拍拼接应用领域, 如城镇航拍图像等。      

Efficient registration of optical and IR images for automatic plant water stress assessment

Automatic registration of optical and IR images is a crucial step towards constructing an automated irrigation control system where plant water information is sensed via thermal imaging. The scene of the IR image is assumed to be completely included in the optical image and the alignment between the common scene in the two images may involve translation and rotation by a small angle, though a small scale difference may also be present. This automatic registration of data from two quite different, non-rigid imaging regimes presents several challenges, which cannot be overcome using common image processing techniques. In this paper, a fully automatic image registration algorithm for the alignment of optical and IR image pairs is described, where Pearson's cross-correlation between a pair of images serves as the similarity measure. A computationally efficient algorithm is designed and packaged as a software application. This work provides an intervention free process for extracting plant water stress information which can be fed into an automated irrigation scheduling program. The proposed algorithm is justified by the comparison of its registration performance with that of other potential algorithm techniques using several experimental data collections. Our results demonstrate the effectiveness of the proposed algorithm and efficiency of its application to the registration of IR and optical images.     

Efficient registration of optical and IR images for automatic plant water stress assessment

Automatic registration of optical and IR images is a crucial step towards constructing an automated irrigation control system where plant water information is sensed via thermal imaging. The scene of the IR image is assumed to be completely included in the optical image and the alignment between the common scene in the two images may involve translation and rotation by a small angle, though a small scale difference may also be present. This automatic registration of data from two quite different, non-rigid imaging regimes presents several challenges, which cannot be overcome using common image processing techniques. In this paper, a fully automatic image registration algorithm for the alignment of optical and IR image pairs is described, where Pearson's cross-correlation between a pair of images serves as the similarity measure. A computationally efficient algorithm is designed and packaged as a software application. This work provides an intervention free process for extracting plant water stress information which can be fed into an automated irrigation scheduling program. The proposed algorithm is justified by the comparison of its registration performance with that of other potential algorithm techniques using several experimental data collections. Our results demonstrate the effectiveness of the proposed algorithm and efficiency of its application to the registration of IR and optical images.     

基于加速鲁棒特征的木材显微图像自动配准方法

针对显微镜观测视野狭小而难以采集到全局图像的问题,提出了一种基于加速鲁棒特征(SURF)的木材显微图像自动配准方法.首先使用SURF检测并描述兴趣点,通过最近邻匹配得到匹配点对后,用双向匹配和RANSAC算法剔除错误匹配.然后利用最小二乘法和匹配结果进行模型参数估计,最后通过插值获得配准图像.对阔叶材显微图像配准实验结果表明,该方法具有较好的鲁棒性,无论图像是否有旋转,都可以实现自动的配准.比起尺度不变特征转换(SWT),由于用SURF得到的兴趣点数量更少,运算速度更快,总的匹配速度提升了5倍左右,缩短了整个配准过程的时间,算法更具有实时性.     

Near-optimal coverage trajectories for image mosaicing using a mini quad-rotor over irregular-shaped fields

Aerial images are useful tools for farmers who practise precision agriculture. The difficulty in taking geo-referenced high-resolution aerial images in a narrow time window considering weather restrictions and the high cost of commercial services are the main drawbacks of these techniques. In this paper, a useful tool to obtain aerial images by using low cost unmanned aerial vehicles (UAV) is presented. The proposed system allows farmers to easily define and execute an aerial image coverage mission by using geographic information system tools in order to obtain mosaics made of high-resolution images. The system computes a complete path for the UAV by taking into account the on-board camera features once the image requirements and area to be covered are defined. This work introduces a full four-step procedure: mission definition, automatic path planning, mission execution and mosaic generation.     

结合航空影像纹理和光谱特征的单木冠幅提取

随着航空摄影测量技术的不断发展与进步,为提高森林资源调查的工作效率,航空影像已经成功应用到林业资源监测中,但在单木冠幅提取上,研究多考虑影像光谱信息,使得分类结果存在偏差。本文提出同时结合航空影像的纹理及光谱特征,利用面向对象的影像分割方法,通过多次实验对比结果确定最优分割尺度,同时在结合正态分布法确定各光谱及纹理特征信息范围的基础上,提取单木冠幅信息。以2012年鹫峰国家森林公园航空像片为数据源,以ENVI5.0为数据处理平台,对影像进行面向对象的分割,提取试验区域内32株树木的冠幅,并结合传统外业实测数据以及立体像对观测数据进行精度分析。试验结果表明:文章所提出的方法试验精度达到90.05%,与传统立体像对量测方法精度相近,但数据获取速度快,满足林业调查基本需求。      

大豆冠层多源图像特征点配准方法研究

针对具有颜色信息的大豆冠层三维结构形态的重建问题,采用PMD摄像机与彩色摄像机相结合的多源图像采集系统获取大豆冠层多源图像,对大豆冠层多源图像特征点配准方法进行研究。以彩色图像和强度图像为研究对象,利用仿射变换实现彩色图像坐标系到PMD图像坐标系的转换;利用Harris算法检测图像特征点,采用基于归一化互相关系数法(NCC)实现特征点粗匹配。为克服传统RANSAC算法抽样次数较多及和数据检验时间较长的弊端,提出在特征点匹配阶段,按照可信度将特征点对排序,从可信度高的点对开始抽取的方法来优化经典RANSAC算法,进而实现特征点精匹配,最终完成多源图像特征点配准。为验证本研究提出的图像配准算法的有效性,将该算法与传统图像配准算法相对比,结果表明:室外和室内环境下,样本组的平准正确配准率分别为83%和87%,均优于传统图像配准算法,并满足快速配准大豆冠层多源图像特征点的要求。     

Three-dimensional representation and analysis of brain energy metabolism

Quantitative autoradiography of brain glucose metabolism has been combined with digital image processing to represent the brain as a three-dimensional (3-D) reconstruction of brain energy use. Autoradiographs contain enormous amounts of potentially useful data, but conventional analyses, based on tedious manual methods, can sample and analyze only a small portion of this information. Computer 3-D reconstruction provides a mechanism for observing and analyzing all the data; therefore, a system of computer programs was developed for this purpose. The programs use digital imaging methods for image registration, superimpose whole brain data sets, and allow resampling of the 3-D data in arbitrary planes for pixel-by-pixel comparisons among multiple 3-D sets. These programs operate on the mathematical properties of the images alone, obviating the need for manual image alignment. Various statistical analyses can be applied to the data directly to study the patterns of metabolic changes in different experiments. The system is applied to data from experiments on the influence of injectable anesthetics on cerebral glucose metabolism.     

基于色彩和纹理特征的建设用地遥感专题信息自动提取方法

提出了一种基于航空影像的建设用地信息自动提取方法.针对航空影像光谱信息少,色彩信息和纹理信息丰富的特点,该方法首先采用HSV色彩变换和纹理分析手段充分挖掘影像中所包含的色彩信息与纹理信息,在此基础上采用多特征阈值分割技术将色彩信息与纹理信息有机结合进行建设用地信息提取,并使用邻域分析方法对提取结果进行修正.通过对研究区域航空影像的处理,结果表明,该方法提取精度较高,且易于实现.     

Registration of multispectral 3D points for plant inspection

Machine vision technologies have shown advantages for efficient and accurate plant inspection in precision agriculture. Regarding the balance between accuracy of inspection and compactness for infield applications, multispectral imaging systems would be more suitable than RGB colour cameras or hyperspectral imaging systems. Multispectral image registration (MIR) is a key issue for multispectral imaging systems, however, this task is challenging. First of all, in many cases, two images needing registration do not have a one-to-one linear mapping in 2D space and therefore they cannot be aligned in 2D images. Furthermore, the general MIR algorithms are limited to images with uniform intensity and are incapable of registering images with rich features. This study developed a machine vision system (MVS) and a MIR method which replaces 2D-2D image registration by 3D-3D point cloud registration. The system can register 3D point clouds of ultraviolet (UV), blue, green, red and near-infrared (NIR) spectra in 3D space. It was found that the point clouds of general plants created by images of different spectral bands have a complementary property, and therefore a combined point cloud, called multispectral 3D point cloud, is denser than any cloud created by a single spectral band. Intensity information of each spectral band is available in a multispectral 3D point cloud and therefore image fusion and 3D morphological analysis can be conducted in the cloud. The MVS could be used as a sensor of a robotic system to fulfil on-the-go infield plant inspection tasks.     

一种基于图像的套印参数测量方法

提出了一种新的基于图像套印参数的检测方法.该方法在应用蜂窝分裂法实现图像量化的基础上,确定色标分割的颜色中心和阈值距离.经过计算色标质心间横向和纵向距离得到印刷套印偏差参数.经过计算色标与标准图像颜色的色差,建立色度与密度的转换关系,将色度值转换为对应的密度值,由网点参数与供墨系统的数学模型,计算出被检测处的实际墨量.理论分析与实验结果均表明:套印偏差检测精度高,有利于提高套印的准确性,油墨的质量和供墨量的实时检测,有利于实现墨量的自动控制,保证印刷图像整个画面阶调和色调的正确复制.     

基于互信息和混合优化算法的多模医学图像配准

针对互信息函数的多极值问题,提出了一种基于混合优化算法的多模医学图像配准方法.对于多模医学图像,以互信息作为相似性测度,使用混合优化算法搜索出最佳配准变换参数,将待配准图像进行变换,从而达到配准的目的.实验表明,该算法能避免陷入局部最优值,配准结果精度达到亚像素级.     

基于小波分解的多模态图像配准方法研究

在小波分解的基础上,采用最大互信息(MMI)配准测度和POWELL优化算法对多模态图像配准,多分辨率配准方法具有可以有效地避免优化算子陷入局部极值、算子收敛速度快和配准精度高等优点。CT/MRI和CT/PET图像配准结果证明基于小波分解的多模态配准方法的有效性。     

Automatic segmentation of relevant textures in agricultural images

One important issue emerging strongly in agriculture is related with the automatization of tasks, where the optical sensors play an important role. They provide images that must be conveniently processed. The most relevant image processing procedures require the identification of green plants, in our experiments they come from barley and corn crops including weeds, so that some types of action can be carried out, including site-specific treatments with chemical products or mechanical manipulations. Also the identification of textures belonging to the soil could be useful to know some variables, such as humidity, smoothness or any others. Finally, from the point of view of the autonomous robot navigation, where the robot is equipped with the imaging system, some times it is convenient to know not only the soil information and the plants growing in the soil but also additional information supplied by global references based on specific areas. This implies that the images to be processed contain textures of three main types to be identified: green plants, soil and sky if any. This paper proposes a new automatic approach for segmenting these main textures and also to refine the identification of sub-textures inside the main ones. Concerning the green identification, we propose a new approach that exploits the performance of existing strategies by combining them. The combination takes into account the relevance of the information provided by each strategy based on the intensity variability. This makes an important contribution. The combination of thresholding approaches, for segmenting the soil and the sky, makes the second contribution; finally the adjusting of the supervised fuzzy clustering approach for identifying sub-textures automatically, makes the third finding. The performance of the method allows to verify its viability for automatic tasks in agriculture based on image processing.     

The application of small unmanned aerial systems for precision agriculture: a review

Precision agriculture (PA) is the application of geospatial techniques and sensors (e.g., geographic information systems, remote sensing, GPS) to identify variations in the field and to deal with them using alternative strategies. In particular, high-resolution satellite imagery is now more commonly used to study these variations for crop and soil conditions. However, the availability and the often prohibitive costs of such imagery would suggest an alternative product for this particular application in PA. Specifically, images taken by low altitude remote sensing platforms, or small unmanned aerial systems (UAS), are shown to be a potential alternative given their low cost of operation in environmental monitoring, high spatial and temporal resolution, and their high flexibility in image acquisition programming. Not surprisingly, there have been several recent studies in the application of UAS imagery for PA. The results of these studies would indicate that, to provide a reliable end product to farmers, advances in platform design, production, standardization of image georeferencing and mosaicing, and information extraction workflow are required. Moreover, it is suggested that such endeavors should involve the farmer, particularly in the process of field design, image acquisition, image interpretation and analysis.     

Imaging from manned ultra-light and unmanned aerial vehicles for estimating properties of spring wheat

This study investigates an imaging system based on a Rikola hyperspectral (HSI) and Nikon D800E (CIR) cameras installed on a manned ultralight aircraft Bekas Ch-32 for applications involving precision agriculture. The efficiency of this technical solution is compared with that of using Canon PowerShot SX260HS camera images acquired from helicopter-type unmanned aerial vehicle (UAV) to accomplish similar tasks. The criteria for comparison were the suitability of acquired images for modelling chlorophyll concentration in spring wheat and for estimating the normalized difference red edge (NDRE) index, which is conventionally obtained using OptRx proximal sensors. Hyperspectral image values used as explanatory variables in ordinary least squares regression explain 68 and 61% of the variance in chlorophyll concentration and NDRE, respectively and outperform other images. The advantage of hyperspectral imagery became negligible when applying geographically weighted regression to improve global regression models. The use of ultralight aircraft as a sensor platform for precision agriculture aimed aerial photography projects is suggested as currently the most cost-effective solution in Lithuania.     

基于改进C-V模型的木材表面缺陷图像分割

木材表面缺陷会严重影响木材的质量、性能和使用价值,对木材表面缺陷分割检测有利于提高木材的利用率,节约现有木材资源,缓解森林资源短缺的压力。针对传统的C-V(Chan-Vese)模型算法不能分割灰度不均匀图像的缺点,本文采用C-V模型与形态学结合的方法与传统的C-V模型算法进行对比试验。与此同时,根据C-V模型和C-V模型结合形态学方法的不足之处,在C-V模型基础上,引入局部拟合函数和高斯核函数,提出了一种基于C-V模型的改进算法,能够有效地克服C-V模型的不足。通过对木材表面缺陷图像分别采用传统C-V模型算法、C-V模型与形态学结合的方法和改进的C-V模型算法进行多组针对单一目标的木材表面缺陷图像的对比试验。结果表明:C-V模型能够将虫眼和活节缺陷图像分割出来,但是对纹理干扰强烈的死节缺陷图像分割困难;运用C-V模型与形态学结合的方法,可以有效地消除分割结果中的细小空洞和噪声,但是仍无法抵抗死节缺陷图像中木材自身纹理的干扰,难以将死节缺陷完整地分割出来;改进的C-V模型算法对木材表面缺陷图像的分割能够减少迭代次数,缩短分割时间,使分割轮廓线更加光滑和完整。通过采用改进C-V模型算法对多目标木材表面缺陷图像进行试验,能够更好地验证改进算法的优越性、有效性和可行性。      

Estimation of surface soil organic matter using a ground-based active sensor and aerial imagery

Active canopy sensors are currently being studied as a tool to assess crop N status and direct in-season N applications. The objective of this study was to use a variety of strategies to evaluate the capability of an active sensor and a wide-band aerial image to estimate surface soil organic matter (OM). Grid soil samples, active sensor reflectance and bare soil aerial images were obtained from six fields in central Nebraska before the 2007 and 2008 growing seasons. Six different strategies to predict OM were developed and tested by dividing samples randomly into calibration and validation datasets. Strategies included uniform, interpolation, universal, field-specific, intercept-adjusted and multiple-layer prediction models. By adjusting regression intercept values for each field, OM was predicted using a single sensor or image data layer. Across all fields, the uniform and universal prediction models resulted in less accurate predictions of OM than any of the other methods tested. The most accurate predictions of OM were obtained using interpolation, field-specific and intercept-adjusted strategies. Increased accuracy in mapping soil OM using an active sensor or aerial image may be achieved by acquiring the data when there is minimal surface residue or where it has been excluded from the sensor’s field-of-view. Alternatively, accuracy could be increased by accounting for soil moisture content with supplementary sensors at the time of data collection, by focusing on the relationship between soil reflectance and soil OM content in the 0–1 cm soil depth or through the use of a subsurface active optical sensor.