A method to measure the rice kernel chalkiness objectively

Rice kernel chalkiness is an impor-tant quality character.Being the un-transparent portions in grain en-dosperm,chalkiness iS always mea-sured by some subjective eye-judgingmethods domestically and interna-tionally.Results measured by suchmethods aye subjective,inaccurate,and unstable.This research is in-

Non-Rigid Object Tracking by Anisotropic Kernel Mean Shift

Mean shift,an iterative procedure that shifts each data point to the average of data points in its neighborhood,has been applied to object tracker.However,the traditional mean shift tracker by isotropic kernel often loses the object with the changing object structure in video sequences,especially when the object structure varies fast.This paper proposes a non-rigid object tracker by anisotropic kernel mean shift in which the shape,scale,and orientation of the kernels adapt to the changing object structure.The experimental results show that the new tracker is self-adaptive and approximately twice faster than the traditional tracker,which ensures the robustness and real time of tracking.

Predicting the fault-proneness of class hierarchy in object-oriented software using a layered kernel

A novel kernel learning method for object-oriented (OO) software fault prediction is proposed in this paper. With this method, each set of classes that has inheritance relation named class hierarchy, is treated as an elemental software model. A layered kernel is introduced to handle the tree data structure corresponding to the class hierarchy models. This method was vali-dated using both an artificial dataset and a case of industrial software from the optical communication field. Preliminary experi-ments showed that our approach is very effective in learning structured data and outperforms the traditional support vector learning methods in accurately and correctly predicting the fault-prone class hierarchy model in real-life OO software.

Rebound of Region of Interest (RROI), a New Kernel-Based Algorithm for Video Object Tracking Applications

This paper presents a new kernel-based algorithm for video object tracking called rebound of region of interest (RROI). The novel algorithm uses a rectangle-shaped section as region of interest (ROI) to represent and track specific objects in videos. The proposed algorithm is constituted by two stages. The first stage seeks to determine the direction of the object’s motion by analyzing the changing regions around the object being tracked between two consecutive frames. Once the direction of the object’s motion has been predicted, it is initialized an iterative process that seeks to minimize a function of dissimilarity in order to find the location of the object being tracked in the next frame. The main advantage of the proposed algorithm is that, unlike existing kernel-based methods, it is immune to highly cluttered conditions. The results obtained by the proposed algorithm show that the tracking process was successfully carried out for a set of color videos with different challenging conditions such as occlusion, illumination changes, cluttered conditions, and object scale changes.

基于匝道合流数据的自动驾驶汽车安全性测试评价方法

针对自动驾驶汽车测试场景不明确、评价模型主观性强等问题,研究了高速匝道汇入场景下的典型测试场景提取方法和自动驾驶汽车匝道汇入安全性客观评价方法。深入分析了匝道汇入功能场景下的逻辑场景要素,对自然驾驶数据中的自车速度、车间距离、前车车速等逻辑场景要素进行聚类,得到两类典型的匝道汇入测试场景用于自动驾驶汽车的仿真测试。构建多层次自动驾驶汽车评价体系,引入基于自然驾驶数据的核密度估计模型来获取指标最优阈值,建立以最优阈值为参考序列、以层次分析法(AHP)和客观赋权法(CRITIC)为权重输入的灰色关联理评价模型,对自动驾驶汽车在汇入过程中的安全性进行客观评价。评价结果表明:基于核密度估计的灰色关联理论模型评价结果与主观模糊综合分析模型的评价结果相似率达98.01%,验证了客观模型的有效性。

对象驱动的Linux内核crash分类技术研究

Crash(程序崩溃)分析是漏洞挖掘与利用的关键阶段,精准的crash分类是crash分析和漏洞利用的前提.针对现有的Linux内核crash存在大量重复的问题,本文提出一种对象驱动的Linux内核crash分类方法.该方法将内核crash与内核对象的关系建模为二部图结构,从而将crash分类问题转化为内核对象的相似性对比问题.首先,通过对crash执行后向污点分析提取crash相关的内核对象;其次,构造内核对象调用图计算内核与根本原因的相关性度量;最后,基于上述结果构造二部图实现crash相似性比较算法.基于上述方法,本文开发出了Linux内核crash分类的原型系统.通过在真实的数据集上进行实验,验证了系统的有效性和可用性,弥补了现有分类方法粒度较粗,存在误报较多的问题.

KCPStack:张量分解的卷积核分层矩阵压缩方法

针对现有张量分解卷积核压缩方法难以兼顾时空轻量化、过于依赖卷积瓶颈结构等问题,提出一种具有可观压缩与加速能力的卷积核分层矩阵压缩方法(KCPStack)。首先,在矩阵乘法视角下,将卷积核按通道拆分为2阶克罗内克规范多项式(KCP)分解,所得因子张量组合为两层权重矩阵,使卷积计算转换为具有较高推理效率的双层轻量卷积结构;其次,对比所提KCPStack方法与其他典型张量分解卷积核压缩方法的参数约减空间复杂度与推理计算时间复杂度;最后,基于RK3588神经处理单元进行KCPStack方法的部署,面向实际场景目标检测识别需求开发相关应用。实验结果表明:与现有张量分解方法相比,在张量秩相同或者参数量相当的前提下,所提KCPStack方法具有最快的推理计算效率;在图像分类标准数据集CIFAR-10和ImageNet上,KCPStack方法能够将精度损失控制在1%左右,最高可减少85.0%的参数量和79.8%的计算量;在目标检测识别标准数据集COCO上,KCPStack方法相对于基线模型的平均精度下降不超过1%;采用所提KCPStack方法对实际场景进行目标检测识别,在RK3588神经处理单元上能达到95.4%的平均精度和35帧/s的图像处理帧率,内存开销仅为33.1 MB。

系统故障演化过程中关键事件的确定方法研究

为研究系统故障演化过程中关键事件的确定方法,提出了一种基于核局部保持投影(Kernel Locality Preserving Projections,KLPP)的关键事件确定方法。首先论述了系统故障演化过程、关键事件和因素空间,随后提出了关键事件确定方法,最后进行了实例分析。研究认为系统故障演化过程具有复杂的结构和层次,经历事件是演化测量得到的对象,其中具有决定作用的就是关键事件。关键事件是描述演化过程的基础,可通过因素空间中的对象分布进行确定。通过KLPP方法对对象分布特征进行研究,实现近邻对象分析,得到特征对象。这些特征对象对应的经历事件即为关键事件。按照测量时刻升序排列特征对象即为所求,最终作为描述演化过程的空间故障网络的节点。实例分析得到了预期结果,并说明了方法的特点和研究意义。

融合空谱特征的MR-KRVFL高光谱地物识别模型研究

针对高光谱图像复杂空谱特性影响地物识别模型分类精度的问题,提出一种融合空谱特征的流形正则化核随机向量函数连接网络(MR-KRVFL)高光谱图像地物识别方法。首先,对高光谱图像进行熵率超像素分割(ERS),获取对应的同质区域;其次,利用主元分析(PCA)对同质区域进行降维并提取其空谱联合特征;最后,基于空谱特征信息,构造核随机向量函数连接网络(KRVFL)地物识别模型,并对模型进行流形正则化约束,提高高光谱图像地物识别模型的泛化性能。将该模型应用于Indian Pines和Pavia University高光谱数据集,分类精度达到了96.84%和98.83%,证明所提模型的有效性。

面向无人机视角下小目标检测的YOLOv8s改进模型

从无人机视角进行目标检测,面临图像目标小、分布密集、类别不均衡等难点,且由于无人机的硬件条件限制了模型的规模,导致模型的准确率偏低。提出一种融合多种注意力机制的YOLOv8s改进模型,在骨干网络中引入感受野注意力卷积和CBAM(concentration-based attention module)注意力机制改进卷积模块,解决注意力权重参数在感受野特征中共享问题的同时,在通道和空间维度加上注意力权重,增强特征提取能力;通过引入大型可分离卷积注意力思想,改造空间金字塔池化层,增加不同层级特征间的信息交融;优化颈部结构,增加具有丰富小目标语义信息的特征层;使用inner-IoU损失函数的思想改进MPDIoU(minimum point distance based IoU)函数,以innerMPDIoU代替原损失函数,提升对困难样本的学习能力。实验结果表明,改进后的YOLOv8s模型在VisDrone数据集上mAP、P、R分别提升了16.1%、9.3%、14.9%,性能超过YOLOv8m,可以有效应用于无人机平台上的目标检测任务。

A Cluster-Based Method for Marine Sensitive Object Extraction and Representation

Within the context of global change, marine sensitive factors or Marine Essential Climate Variables have been defined by many projects, and their sensitive spatial regions and time phases play significant roles in regional sea-air interactions and better understanding of their dynamic process. In this paper, we propose a cluster-based method for marine sensitive region extraction and representation. This method includes a kernel expansion algorithm for extracting marine sensitive regions, and a field-object triple form, integration of object-oriented and field-based model, for representing marine sensitive objects. Firstly, this method recognizes ENSO-related spatial patterns using empirical orthogonal decomposition of long term marine sensitive factors and correlation analysis with multiple ENSO index. The cluster kernel, defined by statistics of spatial patterns, is initialized to carry out spatial expansion and cluster mergence with spatial neighborhoods recursively, then all the related lattices with similar behavior are merged into marine sensitive regions. After this, the Field-object triple form of < O, A, F > is used to represent the marine sensitive objects, both with the discrete object with a precise extend and boundary, and the continuous field with variations dependent on spatial locations. Finally, the marine sensitive objects about sea surface temperature are extracted, represented and analyzed as a case of study, which proves the effectiveness and the efficiency of the proposed method.

基于SRKDA的系统故障演化过程分解方法研究

为研究系统故障演化过程中可能蕴含的多种演化特征,对演化过程的分解进行研究,提出基于谱回归核判别分析(SRKDA)的演化过程分解方法。首先介绍演化过程的特点和分解原理,其次论证对象集合对演化过程的可表示性,给出分解方法流程,最后进行实例分析。研究结果表明:分解演化过程本质上是对象与系统功能状态对应关系的确定,各对象集合都对应了各自的子演化过程;线性和非线性条件下对象可表示各种功能状态;对象标签矩阵须满足标签值的均匀分布特征;使用SRKDA算法可以确定最大准确度和最优对象标签集合,实现演化过程的分解;实例分析得到在20000次迭代后最大准确度为0.85,3个子演化过程分别包含41,33,26个对象。研究结果可为系统故障过程的特征分析提供参考方法。

Multi-Object Tracking with Micro Aerial Vehicle

A simple yet efficient tracking framework is proposed for real-time multi-object tracking with micro aerial vehicles(MAVs). It's basic missions for MAVs to detect specific targets and then track them automatically. In our method, candidate regions are generated using the salient detection in each frame and then classified by an eural network. A kernelized correlation filter(KCF) is employed to track each target until it disappears or the peak-sidelobe ratio is lower than a threshold. Besides, we define the birth and death of each tracker for the targets. The tracker is recycled if its target disappears and can be assigned to a new target. The algorithm is evaluated on the PAFISS and UAV123 datasets. The results show a good performance on both the tracking accuracy and speed.

A NOVEL FAST MOVING OBJECT CONTOUR TRACKING ALGORITHM

If a somewhat fast moving object exists in a complicated tracking environment, snake's nodes may fall into the inaccurate local minima. We propose a mean shift snake algorithm to solve this problem. However, if the object goes beyond the limits of mean shift snake module operation in suc- cessive sequences, mean shift snake's nodes may also fall into the local minima in their moving to the new object position. This paper presents a motion compensation strategy by using particle filter; therefore a new Particle Filter Mean Shift Snake (PFMSS) algorithm is proposed which combines particle filter with mean shift snake to fulfill the estimation of the fast moving object contour. Firstly, the fast moving object is tracked by particle filter to create a coarse position which is used to initialize the mean shift algorithm. Secondly, the whole relevant motion information is used to compensate the snake's node positions. Finally, snake algorithm is used to extract the exact object contour and the useful information of the object is fed back. Some real world sequences are tested and the results show that the novel tracking method have a good performance with high accuracy in solving the fast moving problems in cluttered background.

基于改进KCF的多目标人员检测与动态跟踪方法

针对煤矿巷道光照不足、目标尺度变化剧烈、目标容易被遮挡和矿灯干扰等因素,导致对于井下的目标检测和跟踪存在成功率和准确度低的问题,提出一种基于改进核相关滤波(KCF)算法的多目标人员检测与动态跟踪方法,为避免井下复杂环境中由于光照不均引起检测失败,在改进的KCF算法中引入SSD检测算法,以提升对多目标人员检测能力。(1)读取待跟踪视频序列,使用经过井下数据集训练后的SSD算法检测图像中的目标,若没有发现目标则继续读取下一帧。(2)将检测到的目标放入跟踪器中,对图像进行预处理,通过比较将所有的检测框按照设定的阈值进行打分,并根据分值从高到低依次排列,高分的检测结果直接输出,低分的检测结果用于滤除不良信息,以提升检测速度。(3)通过KCF跟踪预测目标M帧后清空跟踪器,再重新进行目标检测。通过检测算法和跟踪算法的叠加,保证对目标的持续跟踪能力。实验结果表明:(1)该方法最后的损失值稳定在1.675附近,检测结果较为稳定。(2)经过训练后的SSD算法识别精度较训练前的SSD算法识别精度提高了52.7%。(3)该方法对矿井人员检测成功率、跟踪准确率分别为87.9%,88.9%,均高于其他4种算法(KCF、CSRT、TLD及MIL)的检测成功率、跟踪准确率。(4)该方法在重叠阈值较低时具有较高成功率,直至重叠阈值大于0.8时,成功率大幅下降,这是因为矿井中环境多样,想要完全符合标注的框有一定难度。实际应用结果表明:在井下煤矿巷道光照不足、目标尺度变化剧烈、容易被遮挡和受矿灯干扰等复杂环境中,该方法具有较高的适用性。

改进多尺度卷积结构与高斯核的E-CenterNet算法

针对ResNet和DLA(deep layer aggregation)为主干网络的CenterNet算法特征提取能力不充分、热力图与目标真实边界框吻合度不高、关键点损失函数未充分考虑预测值对难易样本训练比重的影响等不足,提出一种改进多尺度卷积结构与高斯核的E-CenterNet算法。引入轻量型EfficientNetV2-S作为主干网络,并结合基于金字塔分割注意力网络的多尺度卷积结构对其进行改进,提高特征提取能力;对高斯核进行改进,使CenterNet产生的热力图由固定的圆形改进为随边界框宽高变化的椭圆形,增强算法对边界框宽高差异大的目标的检测能力;提出一种基于关键点预测值的关键点损失函数,提高算法对难样本的训练比例。在Pascal VOC数据集上的实验结果表明:E-CenterNet算法的mAP达到83.3%,比原始算法提升了2.6个百分点,检测性能优于CenterNet算法。

复杂环境下基于改进DeepSORT的行人实时稳定跟踪方法

实时多目标跟踪算法取得了理想的跟踪性能,但大多数现有算法的跟踪速度较慢,且随着背景复杂度的增加,跟踪精度也随之降低。针对此问题,本文提出了基于在线数据关联的行人实时跟踪算法。首先,设计了核相关滤波和卡尔曼滤波双轨道预测机制,配合DeepSORT中的级联匹配形成了预测-跟踪-校准体系,使数据关联更加可靠。此外,在目标检测部分引入了注意力机制,通过强化目标对象的位置信息增强特征表示能力,从而提升跟踪的精度。该模型在MOT16数据集上的MOTA达到了66.5%、IDF1达到了64.2%、IDSW达到了641。与DeepSORT算法对比,MOTA和IDF1分别提升了13%和13.2%,IDSW下降了410。本文算法有助于解决行人实时跟踪时出现的目标误检、漏检等问题,在跟踪中对严重遮挡情况仍保持了较高的跟踪精度,在复杂环境下可以实现行人实时稳定跟踪。

融合共现统计与fhog特征的目标跟踪方法

针对复杂场景目标跟踪过程中难以完全有效利用特征对目标信息进行完备表达的问题,提出了一种融合共现统计与fhog特征的目标跟踪算法。在fDSST跟踪算法的基础上,引入了一种基于图像共现统计的像素滤波模块,利用图像中经常出现的像素值在共现矩阵中的权重较高的特性,与原始跟踪算法中fhog特征进行融合,增加了基于纹理相似性的共现统计信息。将输入跟踪图像映射到嵌入图像中,使得嵌入图像空间中像素值之间的欧氏距离类似于原始空间中的共现统计量,从而得到更优的跟踪结果。在OTB50和OTB100两个跟踪数据集上的实验表明,本文算法相比原算法在跟踪精度和成功率上有所提升,性能较优。

基于核函数的目标跟踪系统建模与仿真

针对目标跟踪中的快速运动、旋转、遮挡等问题,采用核函数算法实现目标跟踪。利用卷积实现密集采样,使用高斯核函数岭回归估计目标位置,在线更新高斯模型和核相关滤波器。通过OTB数据集对算法进行实验验证,结果表明,算法Precision能达到69.7%,Success达到66.8%,跟踪速度达412fps。实验表明,算法在跟踪过程中可对旋转、模糊、遮挡图像中的目标进行有效跟踪,在多个指标上优于其它比较算法。

改进YOLOX-S模型的施工场景目标检测

现有YOLOX-S模型在施工环境干扰下目标检测平均精准率(AP)偏低,不能较好满足实际应用需要。针对上述问题,从引入结构重参数化模块、引入卷积注意力模块、引入AdamW优化算法三方面对YOLOX-S模型进行改进。首先,利用RepVGGBlock解耦训练阶段与测试阶段的模型结构,在训练阶段模型的Backbone与Neck中构建更多残差结构,提高模型的特征提取能力。其次,利用LKA模块提取局部特征信息与长距离依赖关系,为后续计算目标边界框位置与大小提供更加有效的注意力指引,提升检测平均精准率。然后,使用AdamW优化算法替代Adam优化算法更新模型参数,进一步改良模型收敛结果,提升模型泛化能力。最后,在建筑工地运动目标数据集(MOCS)上进行实验,结果表明,改进YOLOX-S模型检测所有目标的平均精准率提升3.3个百分点,检测大目标、中目标、小目标的平均精准率分别提升3.2、2.3、2.2个百分点。同时,改进YOLOX-S模型计算代价未明显增加,可在实时运行的同时更好满足施工场景下对目标检测平均精准率的需要。