Transmission line detection in aerial images: An instance segmentation approach based on multitask neural networks

Camera-based transmission line detection (TLD) is a fundamental and crucial task for automatically patrolling powerlines by aircraft. Motivated by instance segmentation, a TLD algorithm is proposed in this paper with a novel deep neural network, i.e., CableNet. The network structure is designed based on fully convolutional networks (FCNs) with two major improvements, considering the specific appearance characteristics of transmission lines. First, overlaying dilated convolutional layers and spatial convolutional layers are configured to better represent continuous long and thin cable shapes. Second, two branches of outputs are arranged to generate multidimensional feature maps for instance segmentation. Thus, cable pixels can be detected and assigned cable IDs simultaneously. Multiple experiments are conducted on aerial images, and the results show that the proposed algorithm obtains reliable detection performance and is superior to traditional TLD methods. Meanwhile, segmented pixels can be accurately identified as cable instances, contributing to line fitting for further applications.     

SOINN-Based Abnormal Trajectory Detection for Efficient Video Condensation

With the evolution of video surveillance systems, the requirement of video storage grows rapidly; in addition, safe guards and forensic officers spend a great deal of time observing surveillance videos to find abnormal events. As most of the scene in the surveillance video are redundant and contains no information needs attention, we propose a video condensation method to summarize the abnormal events in the video by rearranging the moving trajectory and sort them by the degree of anomaly. Our goal is to improve the condensation rate to reduce more storage size, and increase the accuracy in abnormal detection. As the trajectory feature is the key to both goals, in this paper, a new method for feature extraction of moving object trajectory is proposed, and we use the SOINN (Self-Organizing Incremental Neural Network) method to accomplish a high accuracy abnormal detection. In the results, our method is able to shirk the video size to 10% storage size of the original video, and achieves 95% accuracy of abnormal event detection, which shows our method is useful and applicable to the surveillance industry.     

A rapid detection method for the surface defects of mosaic ceramic tiles

Due to its unique artistic value, mosaic ceramics are widely used in construction-related fields. To meet the artist's demand for high-quality mosaic ceramic to create artistic works, it is necessary to meet the needs for efficient screening of mosaic ceramic tiles. Different from the ordinary large-target ceramics, mosaic ceramics exhibit characteristics of small tile sizes, a variety of colors, large demand for quantities, and easy reflection on the surface. Common manual detection methods show problems of low efficiency or accuracy, easy to fatigue, and many others. To solve these problems, this paper proposes a new detection method to identify surface defects of mosaic ceramic tiles and designs a detection system platform to achieve rapid detection. The experiment proves that the detection system has a detection rate of 93.99% for small defects on the surface of mosaic ceramic tiles, and the detection time of a single mosaic ceramic tile is less than 0.06 s. The detection method can quickly and accurately screen out high-quality, defect-free mosaic ceramic tiles, which can effectively improve the quality and artistic value of mosaic ceramic art creation.     

基于改进相干增强扩散与纹理能量测度和高斯混合模型的导光板表面缺陷检测方法

针对液晶屏(LCD)导光板表面缺陷检测方法存在漏检率和误检率较高,对产品表面复杂渐变的纹理结构适应性差的问题,提出一种基于改进相干增强扩散(ICED)与纹理能量测度和高斯混合模型(TEM-GMM)的LCD导光板表面缺陷检测方法。首先,构建ICED模型,基于结构张量引入平均曲率流扩散(MCF)滤波,使得相干增强扩散(CED)模型对缺陷的细线状纹理有良好的边缘保持效果,并利用相干性得到缺陷纹理增强和背景纹理抑制的滤波后图像;然后,根据Laws纹理能量测度(TEM)提取图像纹理特征,将图像的背景纹理特征作为离线阶段高斯混合模型(GMM)的训练数据,使用期望最大化(EM)算法估计GMM参数;最后,计算待检测图像各像素的后验概率,并将其作为在线检测阶段缺陷像素的判断依据。实验结果表明,该检测方法在导光颗粒随机、规则两种分布的缺陷图像测试数据组上的漏检率和误检率分别为3.27%、4.32%和3.59%、4.87%。所提检测方法适用范围广,可有效检测出LCD导光板表面划痕、异物、脏污和压伤等类型的缺陷。     

High-roubustness keystroke recognition method based on acoustic spatial gradient

For the fluctuation of CFCC caused by environmental noise is the main reason for the low accuracy of keystroke detection,the spatial characteristics of adjacent between CFCC were studied,and the spatial gradient structure of CFCC based on points was established.On this basis,the effect of CFCC spatial gradient on keystroke content recognition and the selection of precise neighborhood points were studied on training and testing.Finally,a high-robustness keystroke recognition algorithm based on acoustic signals was constructed.Extensive experiments in different environments demonstrate that the proposed CFCC spatial gradient sound feature achieves great performance and the recognition accuracy is 96.15%.     

基于全相位预处理的时域多重信号分类波达方向估计方法

低信噪比下,针对宽带短脉冲情况下频域多重信号分类(MUSIC)中噪声子空间估计不稳定问题,提出一种基于全相位预处理的时域多重信号分类波达方向(DOA)估计方法。①对线列阵接收数据进行分组处理;②按搜索角度对各组数据进行相移预处理,并对各组数据预处理结果进行相加,得到一组新数据;③对线列阵接收数据在时域构建相移后的协方差矩阵,在更短数据长度下,稳定实现噪声子空间估计,并依据估计出的噪声子空间含有的正交特性,通过单位矩阵加法器得到相应空间谱估计值,实现波达方向估计。数值仿真和实测数据处理结果表明,相比频域MUSIC方法,该方法有效提高了线列阵接收数据协方差矩阵中信号含有量和信噪比,能够在更短数据长度情况下实现对噪声子空间的稳定估计,具有较好的稳定性和检测性能,提高了MUSIC方法在实际波达方向估计中的鲁棒性。     

多机器人体间动态碰撞检测方法仿真研究

当前多结合包围求碰撞检测法、Average-Case法、K-DOPs法等实现多机器人体间动态碰撞的检测,均存在寻优性能较差、检测效率较低的问题。为此提出一种基于动态粒子群的多机器人体间动态碰撞检测方法。采用OBB层次包围盒方法,缩小多机器人之间需要动态碰撞检测的区域,同时把动态碰撞检测问题转换为物体特征对间距离机制的非线性优化问题,进而构建层次拓扑框架进行局部碰撞检测,将机器人体引入到粒子群算法中建立混合进化算法,找到动态碰撞检测的最优解,实现多机器人体间动态碰撞检测。仿真结果证明,所提方法的检测效率高达96%,且具有较高的寻优性能。     

基于多尺度的贝叶斯模型显著性检测

针对传统基于贝叶斯模型的显著性检测算法存在准确率不理想的问题,提出了一种基于多尺度的贝叶斯模型显著性检测算法。通过超像素分割算法(SLIC)将原图分割成不同尺度的超像素,根据超像素边界信息得到背景种子,进而通过距离计算和多尺度融合得到背景先验;对原图进行颜色增强,采用Harris算子对增强图进行检测角点求得凸包,融合不同尺度下的超像素得到凸包先验;融合背景先验和凸包先验得到最终先验;利用颜色直方图和凸包计算似然概率;将最终先验和似然概率通过贝叶斯模型计算显著图。在公开数据集MSRA1000、ECSSD上与多种传统算法进行准确率和召回率对比,该算法有更好的表现。     

改进RetinaFace的自然场景口罩佩戴检测算法

新型冠状病毒可以通过空气中的飞沫、气溶胶等载体进行传播，在公共场所下正确佩戴口罩可以有效地防止病毒的传播。提出了一种自然场景下人脸口罩佩戴检测方法，对RetinaFace算法进行了改进，增加了人脸口罩佩戴检测任务，优化了损失函数。在特征金字塔网络中引入了一种改进的自注意力机制，增强了特征图的表达能力。建立了包含3 000张图片的数据集，并进行手工标注，用于网络训练。实验结果表明该算法可以有效进行口罩佩戴检测，在自然场景视频中也取得了不错的检测效果。     

基于主动学习的离群点集成挖掘方法研究

离群点检测任务通常缺少可用的标注数据，且离群数据只占整个数据集的很小一部分，相较于其他的数据挖掘任务，离群点检测的难度较大，尚没有单一的算法适合于所有的场景。因此，结合多样性模型集成和主动学习思想，提出了一种基于主动学习的离群点集成检测方法OMAL（Outlier Mining based on Active Learning）。在主动学习框架指导下，根据各种基学习器的对比分析，选择了基于统计的、基于相似性的、基于子空间划分的三个无监督模型作为基学习器。将各基学习器评判的处于离群和正常边界的数据整合后呈现给人类专家进行标注，以最大化人类专家反馈的信息量；从标注的数据集和各基学习器投票产生的数据集中抽样，基于GBM（Gradient BoostingMachine）训练一个有监督二元分类模型，并将该模型应用于全数据集，得出最终的挖掘结果。实验表明，提出方法的AUC有了较为明显的提升，且具有良好的运行效率，具备较好的实用价值。