基于兴趣区域掩码卷积神经网络的红外-可见光图像融合与目标识别算法研究

建立权重独立的双通道残差卷积神经网络,对可见光与红外频段下的目标图像进行特征提取,生成多尺度复合频段特征图组.基于像点间的欧式距离计算双频段特征图显著性,根据目标在不同成像频段下的特征贡献值进行自适应融合.通过热源能量池化核与视觉注意力机制,分别生成目标在双频段下的兴趣区域逻辑掩码并叠加在融合图像上,凸显目标特征并抑制...     

Infrared image enhancement through saliency feature analysis based on multi-scale decomposition

To improve contrast between dim target region and background in infrared (IR) long-range surveillance, this paper proposes a fast image enhancement approach using saliency feature extraction based on multi-scale decomposition. Firstly, a smooth based multi-scale decomposition is designed and applied to original infrared image, generating sub-images with various frequency components at different decomposition levels. The dim target regions of sub-images are extracted by a local frequency-tuned based saliency feature detection method, secondly. With saliency maps created by saliency extraction using multi-scale local windows with different sizes, the sub-images are enhanced at different decomposition scales. Finally, the enhanced result is reconstructed by synthesizing the all sub-images with adjustable synthetic weights. Since salient areas are analyzed based on fast multi-scale image decomposition, IR image can be s enhanced with good contrast successfully and rapidly. Compared with other algorithms, the experimental results prove that the proposed method is robust and efficient for IR image enhancement.     

基于纹理显著性的微光图像目标检测

微光图像对比度较低,目标显著性不明显,目标自动探测难度大.针对此问题,本文提出一种噪声鲁棒性较好的图像局部纹理粗糙度算法,并给出一种适用于微光图像显著分析的纹理显著性算法.首先,提出一种新的局部纹理粗糙度算法,该算法利用最佳尺寸计算局部纹理粗糙度,对纹理图像进行加噪实验,与基于局部分形维的粗糙度方法相比,本文局部纹理粗糙度算法表现出较好的噪声鲁棒性;其次,在提取图像粗糙度特征图的基础上,给出一种针对纹理的显著性度量算法;最后,将纹理显著性算法应用于微光图像目标检测,实验结果证明了该算法的有效性.     

多尺度融合超分辨率算法在无人机探测中的应用

基于光电传感器的低慢小无人机探测系统能够快速准确地发现并识别无人机目标,但远距离非合作无人机目标在图像中像素比重过小,特征退化较明显,使识别率大大降低.图像超分辨技术能够从低分辨率目标图像区域中获得高分辨率图像并恢复更多的细节特征,现有超分辨技术很难在保证推理速度的前提下兼容图像的高低频特征,因此为了满足探测系统的需求...     

基于双边滤波的弱小目标背景抑制

 为解决红外弱小目标检测技术中结构化背景抑制的难题,利用双边滤波集成了图像几何、光度和局部结构相似性等信息并以非迭代、局部操作的优点,提出了一种基于双边滤波的红外弱小目标背景抑制算法,并引入了局部梯度的统计特性来抑制背景细节、增强目标信息,从而达到更好抑制图像中的背景,突出目标图像,提高图像整体对比度、信噪比的目的。实验结果显示,与小波滤波算法比较,该算法对含有弱小目标的复杂背景从主观视觉和数值指标都具有良好抑制效果。     

Moving target detection approach based on spatio-temporal salient perception

The differences in texture and motion between man-made object and natural scene are the key features for human biological visual system to detect moving object in scenery. The paper proposed a moving target detection approach based on spatio-temporal perception, which is a crucial function of the visual attention mechanism. The spatial feature including edge, orientation, texture and contrast of the image are extracted, and then the corresponding spatial salient map are constructed by fusing the features through difference of Gaussian (DOG) function, which can suppress the common and enhance the difference of local region. Then, the global motion, local motion and relative motion between continuous images are extracted by means of pyramid multi-resolution, and the moving salient map is constructed after the motion difference between moving target and background is confirmed. Finally, the spatio-temporal salient map is constructed by fusing the spatial salient map and the moving salient map through competition strategy, and the moving target could be detected by searching the maximum in the spatio-temporal salient map. Some experiments are included and the results show that the method can accurately detect the moving target in complex background.     

Detection of infrared small target based on fusion of multi-filters北大核心CSCD

Aiming at solving accuracy problem of infrared small target detection in sky and ocean background scenarios of infrared image sequences, a novel infrared small target detection based on multi-filters algorithm fusion method is presented in this paper. Firstly infrared small target and imaging, time and space characteristics of the corresponding background noise are analyzed. Tophat algorithm with improved Robinson guard filter are then integrated to highlight target and suppress clutter background by using infrared small target imaging features. Adaptive threshold segmentation is used to extract candidate targets, while Unger smoothing filter and multi-objects association filter are used to eliminate random noise and false targets in the candidate targets. Multiple experiments of infrared small target image sequences are implemented, and experimental results show that proposed method can detect infrared small targets at 99% detection rate with high reliability and good real-time performance. © 2017, Editorial Board, Journal of Applied Optics. All right reserved.     

Infrared/laser multi-sensor fusion and tracking based on the multi-scale model

The state estimation problem of targets detected by infrared/laser composite detection system with different sampling rates was studied in this paper. An effective state estimation algorithm based on data fusion is presented. Because sampling rate of infrared detection system is much higher than that of the laser detection system, the theory of multi-scale analysis is used to establish multi-scale model in this algorithm. At the fine scale, angle information provided by infrared detection system is used to estimate the target state through the unscented Kalman filter. It makes full use of the high frequency characteristic of infrared detection system to improve target state estimation accuracy. At the coarse scale, due to the sampling ratio of infrared and laser detection systems is an integer multiple, the angle information can be fused directly with the distance information of laser detection system to determine the target location. The fused information is served as observation, while the converted measurement Kalman filter (CMKF) is used to estimate the target state, which greatly reduces the complexity of filtering process and gets the optimal fusion estimation. The simulation results of tracking a target in 3-D space by infrared and laser detection systems demonstrate that the proposed algorithm in this paper is efficient and can obtain better performance than traditional algorithm.     

基于局部纹理分析的红外小目标检测及其DSP实现

在红外小目标图像中,目标具有与其邻域背景明显不同的纹理和频率特征,在不同尺度和不同频率通道上有不同的表现,利用小波的多尺度分析理论,可将小目标与其邻域背景区分开。采用适合在低信噪比下小目标检测的局部纹理分析方法实现了小目标检测。为了满足红外小目标检测的实时性要求,采用TI公司的高性能数字多媒体DSP芯片实现了小目标检测系统。通过软件程序的优化设计来进一步提高程序运行速度与流水效率,具有良好的软硬件体系结构。通过对实测红外序列图像进行实验表明,所设计的系统能实时地、稳定地检测复杂背景下的1~3个像素的运动小目标。     

改进的非下采样Contourlet变形红外弱小目标检测方法

针对存在复杂背景干扰和噪声的红外图像弱小目标检测问题,提出了一种非下采样Contourlet变换（NSCT）的改进的红外弱小目标检测方法。首先对含弱小目标的红外图像进行预处理,然后利用NSCT进行变换,并利用改进的非线性映射函数和能量交叉融合相结合的方法实现了背景杂波的抑制,最后引入Otsu算法进行阈值分割分离出红外弱小目标。通过与同类弱小目标检测算法的对比实验,验证了该方法的有效性。     

基于波原子变换的红外复杂背景杂波抑制算法

针对红外图像弱小目标检测技术中复杂背景杂波干扰问题,提出了一种基于波原子变换的红外图像背景抑制算法。首先,采用波原子变换对图像进行多尺度和多方向分解,获得原始图像的多尺度和多方向细节特征;然后,根据目标和背景杂波信号的差异,通过频域变换设计的系数调整函数修正经波原子变换后各子带系数,再经波原子逆变换重构得到估计的背景图像;最后,将其与原始图像相减获得背景杂波抑制后的图像。用真实的红外图像序列进行实验,结果显示,与最大中值和小波变换两种算法相比,该算法能有效地抑制红外弱小目标复杂背景杂波,突出目标信号,提高信杂比,具有良好的背景抑制性能。     

基于形状特征的水声图像小目标识别方法

为了抑制背景噪声,提高目标识别准确率,该文提出一种基于形状特征的水声图像小目标识别方法.对含有目标的水声图像进行非局部均值去噪处理后,使用OTSU算法自适应选取阈值对去噪图像进行二值化分割,结合形态学处理获得分割后的目标区域;提取目标区域的矩形度、圆形度、几何不变矩等各项形状参数,将目标的特征向量输入随机森林分类器实现...     

基于深度学习的铸件X射线图像分割研究

针对当前图像分割算法在实现工业铸件内部缺陷分割上精度低且算法不够轻量化的问题,提出一种基于改进DeepLabv3+的工业铸件内部缺陷检测算法Effi-DeepLab。该方法采用EfficientNet中的MBConv来代替原有的Xception模块进行特征提取,使特征提取网络更加高效与轻量化;针对工业铸件内部缺陷尺寸小的问题,重新设计空洞空间金字塔池化（ASPP）层中空洞卷积的扩张率,使得卷积块对小目标具有更高的鲁棒性;在解码端充分利用特征提取阶段的低阶语义信息进行多尺度特征融合,以提高小目标缺陷分割的精度。实验结果表明,在本文使用的汽车轮毂内部缺陷图像数据集中,Effi-DeepLab模型对缺陷的分割准确率和平均交并比（mIoU）分别为93.58%和89.39%,相比DeepLabv3+分别提升了2.65%和2.24%,具有更好的分割效果;此外,还通过实验验证了本文提出算法具有良好的泛化性。     

基于局部峰值的红外弱小目标快速检测

针对红外图像的小目标检测问题,提出了一种基于局部尖峰特性的检测方法.首先分析红外小目标的局部灰度特性,提出了一种红外目标的峰值特性判据;然后依据目标的峰值特性判据和时域特性,设计了一种目标检测的快速算法,算法先基于子块预选出局部极大值点,把后续运算限于各极大值点处以减少运算量,再根据极大点值在各方向上的灰度下降判断其尖峰特性;最后利用帧间的连续性滤去噪音引起的伪目标.实验表明本文的算法具有很快的处理速度,且能有效滤去图像中的随机噪音.     

An improved infrared dim and small target detection algorithm based on the contrast mechanism of human visual system

To achieve higher detection rate and lower false alarm rate in dim and small target detection, this paper proposed an improved algorithm based on the contrast mechanism of human visual system (HVS) for infrared small target detection in an image with complicated background. According to the contrast mechanism of HVS, Laplacian of Gaussian (LoG) filter is exploited to deal with the input image, which can not only suppress the background noise and clutter but also enhances the target intensity significantly. As a result it increases the contrast ratio between target and background. To further eliminate residual clutter, we process the filtered image with morphological method in all directions. True target is finally obtained by applying local thresholding segmentation to the pre-processed image. Experimental results demonstrate its superior and reliable detection performance by high detection rate and low false alarm rate.     

一种新的空谱联合探测高光谱影像目标探测算法

高光谱遥感影像不但具有高分辨率的空间信息还包含连续的光谱信息,因此在目标探测领域具有独特的应用优势。传统的高光谱遥感影像目标探测侧重于光谱信息的应用,形成了确定性算法和统计学算法。确定性算法通过计算目标光谱与待检测光谱之间的距离来查找目标,不能检测亚像素目标,而且容易受到噪声的影响;统计学目标检测计算背景统计特性,通过探测异常点来检测目标,可以检测亚像素目标和小目标,但容易受到目标尺寸的影响,不能很好的检测大目标。随着高光谱遥感影像的空间分辨率的增加,探测目标已有亚像素目标逐步转换为单像素及多像素目标,此时,在高光谱图像中,相同类别的地物在空间分布上呈现聚类特性, 因此,在利用高光谱遥感影像进行目标探测时,需要将其空间信息融入算法中。将空间特征引入传统目标探测算法。提出了一种新的空谱结合的高光谱目标探测算法,将传统的基于统计的目标探测算子与空域邻域聚类算法相结合,首先利用目标探测算子将影像划分为潜在目标区域与背景区域;通过计算潜在目标区域的质心,以质心为中心进行邻域聚类,剔除潜在目标区域中的背景区域,通过迭代计算获取最终目标探测结果。传统的基于统计的目标探测算子,将整个探测区域定义为背景区域,实现对背景区域的统计特征提取,而该方法将背景区域与潜在目标区域分离,剔除了目标区域对背景区域的统计干扰。将本算子与传统的约束能量最小化算子和自适应余弦探测算子进行分析比较可知,该算子的大目标探测性能优于传统的统计算子。     

多尺度条件卷积的OCT视网膜图像降噪研究

散斑噪声存在于光学相干层析成像(OCT)中,影响OCT图像质量.在使用OCT设备诊断各种常见眼科疾病时,高质量的OCT图像是极为重要的.利用深度神经网络对OCT图像进行降噪处理,使图像在保留空间结构细节的基础上能展示更多的信息.提出了一种基于残差学习网络的新型OCT图像降噪网络-CMCNN,其具有多尺度、多权重和多层次...     

特征增强SSD算法及其在遥感目标检测中的应用

为了提高对复杂场景下多尺度遥感目标的检测精度,提出了基于多尺度单发射击检测(SSD)的特征增强目标检测算法.首先对SSD的金字塔特征层中的浅层网络设计浅层特征增强模块,以提高浅层网络对小目标物体的特征提取能力;然后设计深层特征融合模块,替换SSD金字塔特征层中的深层网络,提高深层网络的特征提取能力;最后将提取的图像特征与不同纵横比的候选框进行匹配以执行不同尺度遥感图像目标检测与定位.在光学遥感图像数据集上的实验结果表明,该算法能够适应不同背景下的遥感目标检测,有效地提高了复杂场景下的遥感目标的检测精度.此外,在拓展实验中,文中算法对图像中的模糊目标的检测效果也优于SSD.     

自适应红外目标特征增强算法

利用直方图均衡化和灰度变换增强算法,不能有效增强红外图像目标。鉴于此,在研究红外图像特点的基础上,提出了一种自适应红外目标特征增强算法。该算法先对红外图像进行中值滤波,滤除掉图像中的随机噪声,然后利用直方图分割将红外图像分为目标和背景2部分,通过线性加权叠加抑制背景和增强目标。实验表明,该算法不仅能够根据红外图像中目标的灰度特性自适应地选取直方图分割阈值,而且在去除噪声和增加对比度的同时还抑制了背景,达到了预期的效果。该算法尤其适用于目标和背景像素比例相近时直方图具有局域双峰特征的红外图像中目标的增强。     

多尺度形态算子融合图像滤波技术及滤波质量评价

针对舰载红外警戒系统的红外和电视图像,提出了一种新的海空背景下受强杂波、噪声污染的图像目标滤波算法和滤波效果的定量评价算子。算法采用多尺度的形态算子对输入的图像并行滤波,大尺度形态算子抑制图像噪声,小尺度形态算子提取目标边缘细节信息。处理后的图像进行基于树状小波帧变换的图像信息融合,融合图像可完备提取不同尺度滤波后的图像信息。针对目标检测跟踪的图像滤波算法的评价,提出了目标与背景的交叉分辨力评价算子及评价准则。仿真实验表明。该滤波算法要优于中值滤波、自适应滤波、小波变换滤波算法,滤波质量的定量评价算法是合理的、有效的。算法适用于舰载红外警戒系统。