To overcome the shortcomings of 1 D and 2 D Otsu’s thresholding techniques, the 3 D Otsu method has been developed.Among all Otsu’s methods, 3 D Otsu technique provides the best threshold values for the multi-level ...To overcome the shortcomings of 1 D and 2 D Otsu’s thresholding techniques, the 3 D Otsu method has been developed.Among all Otsu’s methods, 3 D Otsu technique provides the best threshold values for the multi-level thresholding processes. In this paper, to improve the quality of segmented images, a simple and effective multilevel thresholding method is introduced. The proposed approach focuses on preserving edge detail by computing the 3 D Otsu along the fusion phenomena. The advantages of the presented scheme include higher quality outcomes, better preservation of tiny details and boundaries and reduced execution time with rising threshold levels. The fusion approach depends upon the differences between pixel intensity values within a small local space of an image;it aims to improve localized information after the thresholding process. The fusion of images based on local contrast can improve image segmentation performance by minimizing the loss of local contrast, loss of details and gray-level distributions. Results show that the proposed method yields more promising segmentation results when compared to conventional1 D Otsu, 2 D Otsu and 3 D Otsu methods, as evident from the objective and subjective evaluations.展开更多
Infrared(IR) small target detection is one of the key technologies of infrared search and track(IRST)systems. Existing methods have some limitations in detection performance, especially when the target size is irregul...Infrared(IR) small target detection is one of the key technologies of infrared search and track(IRST)systems. Existing methods have some limitations in detection performance, especially when the target size is irregular or the background is complex. In this paper, we propose a pixel-level local contrast measure(PLLCM), which can subdivide small targets and backgrounds at pixel level simultaneously.With pixel-level segmentation, the difference between the target and the background becomes more obvious, which helps to improve the detection performance. First, we design a multiscale sliding window to quickly extract candidate target pixels. Then, a local window based on random walker(RW) is designed for pixel-level target segmentation. After that, PLLCM incorporating probability weights and scale constraints is proposed to accurately measure local contrast and suppress various types of background interference. Finally, an adaptive threshold operation is applied to separate the target from the PLLCM enhanced map. Experimental results show that the proposed method has a higher detection rate and a lower false alarm rate than the baseline algorithms, while achieving a high speed.展开更多
In order to address the problem of high false alarm rate and low probabilities of infrared small target detection in complex low-altitude background,an infrared small target detection method based on improved weighted...In order to address the problem of high false alarm rate and low probabilities of infrared small target detection in complex low-altitude background,an infrared small target detection method based on improved weighted local contrast is proposed in this paper.First,the ratio information between the target and local background is utilized as an enhancement factor.The local contrast is calculated by incorporating the heterogeneity between the target and local background.Then,a local product weighted method is designed based on the spatial dissimilarity between target and background to further enhance target while suppressing background.Finally,the location of target is obtained by adaptive threshold segmentation.As experimental results demonstrate,the method shows superior performance in several evaluation metrics compared with six existing algorithms on different datasets containing targets such as unmanned aerial vehicles(UAV).展开更多
A conventional global contrast enhancement is difficult to apply in various images because image quality and contrast enhancement are dependent on image characteristics largely. And a local contrast enhancement not on...A conventional global contrast enhancement is difficult to apply in various images because image quality and contrast enhancement are dependent on image characteristics largely. And a local contrast enhancement not only causes a washed-out effect, but also blocks. To solve these drawbacks, this paper derives an optimal global equalization function with variable size block based local contrast enhancement. The optimal equalization function makes it possible to get a good quality image through the global contrast enhancement. The variable size block segmentation is firstly exeoated using intensity differences as a measure of similarity. In the second step, the optimal global equalization function is obtained from the enhanced contrast image having variable size blocks. Conformed experiments have showed that the proposed algorithm produces a visually comfortable result image.展开更多
The goal of salient object detection is to estimate the regions which are most likely to attract human's visual attention. As an important image preprocessing procedure to reduce the computational complexity, sali...The goal of salient object detection is to estimate the regions which are most likely to attract human's visual attention. As an important image preprocessing procedure to reduce the computational complexity, salient object detection is still a challenging problem in computer vision. In this paper, we proposed a salient object detection model by integrating local and global superpixel contrast at multiple scales. Three features are computed to estimate the saliency of superpixel. Two optimization measures are utilized to refine the resulting saliency map. Extensive experiments with the state-of-the-art saliency models on four public datasets demonstrate the effectiveness of the proposed model.展开更多
In recent years, the accuracy of edge detection on several benchmarks has been significantly improved by deep learning based methods. However, the prediction of deep neural networks is usually blurry and needs further...In recent years, the accuracy of edge detection on several benchmarks has been significantly improved by deep learning based methods. However, the prediction of deep neural networks is usually blurry and needs further post-processing including non-maximum suppression and morphological thinning. In this paper, we demonstrate that the blurry effect arises from the binary cross-entropy loss, and crisp edges could be obtained directly from deep convolutional neural networks. We propose to learn edge maps as the representation of local contrast with a novel local contrast loss. The local contrast is optimized in a stochastic way to focus on specific edge directions. Experiments show that the edge detection network trained with local contrast loss achieves a high accuracy comparable to previous methods and dramatically improves the crispness. We also present several applications of the crisp edges, including image completion, image retrieval, sketch generation, and video stylization.展开更多
The important effects of local land-sea thermodynamic contrast between the South China Sea (SCS) and Indochina Peninsula on SCS summer monsoon onset are preliminarily studied by using two sets of SSTA tests and two id...The important effects of local land-sea thermodynamic contrast between the South China Sea (SCS) and Indochina Peninsula on SCS summer monsoon onset are preliminarily studied by using two sets of SSTA tests and two ideal tests in s-p regional climate model. The result shows that warm SST in the SCS in winter and spring is favorable for the formation of monsoon circulation throughout all levels of the atmosphere over the sea, which hastens the onset of SCS summer monsoon. The effects of cold SST are generally the opposite. The local land-sea contrast in the SCS is one of the possible reasons for SCS summer monsoon onset. Superposed upon large-scale land-sea thermodynamic differences, it facilitates the formation of out-breaking onset characteristics of SCS summer monsoon in the SCS area.展开更多
Visual saliency is an important cue in human visual system to identify salient region in the image;it can be useful in many applications including image retrieval,object recognition,image segmentation,etc.Image contra...Visual saliency is an important cue in human visual system to identify salient region in the image;it can be useful in many applications including image retrieval,object recognition,image segmentation,etc.Image contrast has been used as an effective feature to detect visual salient region.However,the conventional contrast measures either in spectral domain or in spatial domain fail to give sufficient consideration towards the local and global characteristics of the image.This paper presents a visual saliency detection algorithm based on a novel contrast measurement.This measurement extracts the spectral information of image block using the 2D discrete Fourier transform(DFT),and combines with the total variation(TV)of image block in spatial domain.The proposed algorithm is used to perform salient region detection in the image,and compared with state-of-the-art algorithms.The experimental results from the MSRA dataset validate the effectiveness of the proposed algorithm.展开更多
An adaptive contrast enhancement (ACE) algorithm is presented in this paper, in which the contrast gain is determined by mapping the local standard deviation (LSD) histogram of an image to a Gaussian distribution func...An adaptive contrast enhancement (ACE) algorithm is presented in this paper, in which the contrast gain is determined by mapping the local standard deviation (LSD) histogram of an image to a Gaussian distribution function. The contrast gain is nonlinearly adjusted to avoid noise overenhancement and ringing artifacts while improving the detail contrast with less computational burden. The effectiveness of our method is demonstrated with radiological images and compared with other algorithms.展开更多
To address the problems of missing inside and incomplete edge contours in camouflaged target detection results,we propose a camouflaged moving target detection algorithm based on local minimum difference constraints(L...To address the problems of missing inside and incomplete edge contours in camouflaged target detection results,we propose a camouflaged moving target detection algorithm based on local minimum difference constraints(LMDC).The algorithm first uses the mean to optimize the initial background model,removes the stable background region by global comparison,and extracts the edge point set in the potential target region so that each boundary point(seed)grows along the center of the target.Finally,we define the minor difference constraints term,combine the seed path and the target space consistency,and calculate the attributes of each pixel in the potential target area to realize camouflaged moving target detection.The algorithm of this paper is verified based on a public data sofa video and test videos and compared with the five classic algorithms.The experimental results show that the proposed algorithm yields good results based on integrity,accuracy,and a number of objective evaluation indexes,and its overall performance is better than that of the compared algorithms.展开更多
This paper reports our study of a novel motion estimation algorithm based on global and local compensability analysis. The spatial correlation of motion field is used to reduce the burden of estimation computation and...This paper reports our study of a novel motion estimation algorithm based on global and local compensability analysis. The spatial correlation of motion field is used to reduce the burden of estimation computation and extra bit rate for motion vectors. Experimental results show that this algorithm is more efficient than the conventional methods, especially for temporal activity regions.展开更多
In the past few years,there have been multiple advances in magnetic resonance (MR) instrumentation,in vivo devices,real-time imaging sequences and interventional procedures with new therapies.More recently,interventio...In the past few years,there have been multiple advances in magnetic resonance (MR) instrumentation,in vivo devices,real-time imaging sequences and interventional procedures with new therapies.More recently,interventionists have started to use minimally invasive image-guided procedures and local therapies,which reduce the pain from conventional surgery and increase drug effectiveness,respectively.Local therapy also reduces the systemic dose and eliminates the toxic side effects of some drugs to other organs.The success of MR-guided procedures depends on visualization of the targets in 3D and precise deployment of ablation catheters,local therapies and devices.MR contrast media provide a wealth of tissue contrast and allows 3D and 4D image acquisitions.After the development of fast imaging sequences,the clinical applications of MR contrast media have been substantially expanded to include pre-during-and post-interventions.Prior to intervention,MR contrast media have the potential to localize and delineate pathologic tissues of vital organs,such as the brain,heart,breast,kidney,prostate,liver and uterus.They also offer other options such as labeling therapeutic agents or cells.During intervention,these agents have the capability to map blood vessels and enhance the contrast between the endovascular guidewire/catheters/devices,blood and tissues as well as direct therapies to the target.Furthermore,labeling therapeutic agents or cells aids in visualizing their delivery sites and tracking their tissue distribution.After intervention,MR contrast media have been used for assessing the efficacy of ablation and therapies.It should be noted that most image-guided procedures are under preclinical research and development.It can be concluded that MR contrast media have great value in preclinical and some clinical interventional procedures.Future applications of MR contrast media in image-guided procedures depend on their safety,tolerability,tissue specificity and effectiveness in demonstrating success of the interventions and therapies.展开更多
To overcome the obstacles of poor feature extraction and little prior information on the appearance of infrared dim small targets,we propose a multi-domain attention-guided pyramid network(MAGPNet).Specifically,we des...To overcome the obstacles of poor feature extraction and little prior information on the appearance of infrared dim small targets,we propose a multi-domain attention-guided pyramid network(MAGPNet).Specifically,we design three modules to ensure that salient features of small targets can be acquired and retained in the multi-scale feature maps.To improve the adaptability of the network for targets of different sizes,we design a kernel aggregation attention block with a receptive field attention branch and weight the feature maps under different perceptual fields with attention mechanism.Based on the research on human vision system,we further propose an adaptive local contrast measure module to enhance the local features of infrared small targets.With this parameterized component,we can implement the information aggregation of multi-scale contrast saliency maps.Finally,to fully utilize the information within spatial and channel domains in feature maps of different scales,we propose the mixed spatial-channel attention-guided fusion module to achieve high-quality fusion effects while ensuring that the small target features can be preserved at deep layers.Experiments on public datasets demonstrate that our MAGPNet can achieve a better performance over other state-of-the-art methods in terms of the intersection of union,Precision,Recall,and F-measure.In addition,we conduct detailed ablation studies to verify the effectiveness of each component in our network.展开更多
【目的】跨视角对象级地理定位(CVOGL)旨在卫星影像上精确定位地面街景或无人机影像所观测目标的地理位置。现有方法多聚焦于图像级匹配,通过对整张影像全局处理实现跨视角关联,缺乏对特定目标的位置编码研究,导致无法将模型的注意力引...【目的】跨视角对象级地理定位(CVOGL)旨在卫星影像上精确定位地面街景或无人机影像所观测目标的地理位置。现有方法多聚焦于图像级匹配,通过对整张影像全局处理实现跨视角关联,缺乏对特定目标的位置编码研究,导致无法将模型的注意力引导到感兴趣目标。并且由于参考图像覆盖范围的变化,查询目标在对应卫星图像中的像素占比极低,精确定位较为困难。【方法】针对以上问题,本文提出了一种基于高斯核函数与异构空间对比损失的跨视角对象级地理定位方法(Cross-View Object-Level Geo-Localization Method with Gaussian Kernel Function and Heterogeneous Spatial Contrastive Loss,GHGeo),用于精确定位感兴趣目标位置。该方法首先通过高斯核函数对查询目标进行精确位置编码,实现了对目标中心点及其分布特征的精细化建模;此外还提出了动态注意力精细化融合模块来动态加权交叉感知全局上下文与局部几何特征的空间相似性,以概率密度预测查询目标在卫星影像中的精确位置;最后通过异构空间对比损失函数来约束其训练过程,缓解跨视角特征差异。【结果】本文在CVOGL数据集进行了实验,实验结果显示:GHGeo在该数据集的“无人机-卫星”任务中,当交并比(IoU)≥25%和≥50%时定位准确率分别达到67.73%和63.00%,相较于基准方法DetGeo分别提升了5.76%和5.34%;在“街景-卫星”定位任务中,对应IoU阈值下的定位准确率分别为48.41%和45.43%的定位准确率,相较于基准方法DetGeo分别提升了2.98%和3.19%。同时与TransGeo,SAFA和VAGeo等方法在CVOGL数据集上进行对比,GHGeo则展现出了更高的定位准确性。【结论】本文方法有效提升了跨视角对象级地理定位方法的精度,为城市规划监测,应急救援调度等应用领域提供关键技术支持和精确位置信息支撑。展开更多
文摘To overcome the shortcomings of 1 D and 2 D Otsu’s thresholding techniques, the 3 D Otsu method has been developed.Among all Otsu’s methods, 3 D Otsu technique provides the best threshold values for the multi-level thresholding processes. In this paper, to improve the quality of segmented images, a simple and effective multilevel thresholding method is introduced. The proposed approach focuses on preserving edge detail by computing the 3 D Otsu along the fusion phenomena. The advantages of the presented scheme include higher quality outcomes, better preservation of tiny details and boundaries and reduced execution time with rising threshold levels. The fusion approach depends upon the differences between pixel intensity values within a small local space of an image;it aims to improve localized information after the thresholding process. The fusion of images based on local contrast can improve image segmentation performance by minimizing the loss of local contrast, loss of details and gray-level distributions. Results show that the proposed method yields more promising segmentation results when compared to conventional1 D Otsu, 2 D Otsu and 3 D Otsu methods, as evident from the objective and subjective evaluations.
基金supported by the National Natural Science Foundation of China under Grant 62003247, Grant 62075169, and Grant 62061160370。
文摘Infrared(IR) small target detection is one of the key technologies of infrared search and track(IRST)systems. Existing methods have some limitations in detection performance, especially when the target size is irregular or the background is complex. In this paper, we propose a pixel-level local contrast measure(PLLCM), which can subdivide small targets and backgrounds at pixel level simultaneously.With pixel-level segmentation, the difference between the target and the background becomes more obvious, which helps to improve the detection performance. First, we design a multiscale sliding window to quickly extract candidate target pixels. Then, a local window based on random walker(RW) is designed for pixel-level target segmentation. After that, PLLCM incorporating probability weights and scale constraints is proposed to accurately measure local contrast and suppress various types of background interference. Finally, an adaptive threshold operation is applied to separate the target from the PLLCM enhanced map. Experimental results show that the proposed method has a higher detection rate and a lower false alarm rate than the baseline algorithms, while achieving a high speed.
基金supported by the National Natural Science Foundation of China (No.U1833203),the National Natural Science Foundation of China (No.62301036)the Aviation Science Foundation (No.2020Z019055001)China Postdoctoral Science Foundation Funded Project (No.2022M720446)。
文摘In order to address the problem of high false alarm rate and low probabilities of infrared small target detection in complex low-altitude background,an infrared small target detection method based on improved weighted local contrast is proposed in this paper.First,the ratio information between the target and local background is utilized as an enhancement factor.The local contrast is calculated by incorporating the heterogeneity between the target and local background.Then,a local product weighted method is designed based on the spatial dissimilarity between target and background to further enhance target while suppressing background.Finally,the location of target is obtained by adaptive threshold segmentation.As experimental results demonstrate,the method shows superior performance in several evaluation metrics compared with six existing algorithms on different datasets containing targets such as unmanned aerial vehicles(UAV).
文摘A conventional global contrast enhancement is difficult to apply in various images because image quality and contrast enhancement are dependent on image characteristics largely. And a local contrast enhancement not only causes a washed-out effect, but also blocks. To solve these drawbacks, this paper derives an optimal global equalization function with variable size block based local contrast enhancement. The optimal equalization function makes it possible to get a good quality image through the global contrast enhancement. The variable size block segmentation is firstly exeoated using intensity differences as a measure of similarity. In the second step, the optimal global equalization function is obtained from the enhanced contrast image having variable size blocks. Conformed experiments have showed that the proposed algorithm produces a visually comfortable result image.
基金the Natural Science Foundation of China(Nos.61602349,61375053,and 61273225)the China Scholarship Council(No.201508420248)Hubei Chengguang Talented Youth Development Foundation(No.2015B22)
文摘The goal of salient object detection is to estimate the regions which are most likely to attract human's visual attention. As an important image preprocessing procedure to reduce the computational complexity, salient object detection is still a challenging problem in computer vision. In this paper, we proposed a salient object detection model by integrating local and global superpixel contrast at multiple scales. Three features are computed to estimate the saliency of superpixel. Two optimization measures are utilized to refine the resulting saliency map. Extensive experiments with the state-of-the-art saliency models on four public datasets demonstrate the effectiveness of the proposed model.
基金supported by the National Natural Science Foundation of China under Grant No. 61832016.
文摘In recent years, the accuracy of edge detection on several benchmarks has been significantly improved by deep learning based methods. However, the prediction of deep neural networks is usually blurry and needs further post-processing including non-maximum suppression and morphological thinning. In this paper, we demonstrate that the blurry effect arises from the binary cross-entropy loss, and crisp edges could be obtained directly from deep convolutional neural networks. We propose to learn edge maps as the representation of local contrast with a novel local contrast loss. The local contrast is optimized in a stochastic way to focus on specific edge directions. Experiments show that the edge detection network trained with local contrast loss achieves a high accuracy comparable to previous methods and dramatically improves the crispness. We also present several applications of the crisp edges, including image completion, image retrieval, sketch generation, and video stylization.
基金National Natural Science Foundation of China (40175021 40233037)
文摘The important effects of local land-sea thermodynamic contrast between the South China Sea (SCS) and Indochina Peninsula on SCS summer monsoon onset are preliminarily studied by using two sets of SSTA tests and two ideal tests in s-p regional climate model. The result shows that warm SST in the SCS in winter and spring is favorable for the formation of monsoon circulation throughout all levels of the atmosphere over the sea, which hastens the onset of SCS summer monsoon. The effects of cold SST are generally the opposite. The local land-sea contrast in the SCS is one of the possible reasons for SCS summer monsoon onset. Superposed upon large-scale land-sea thermodynamic differences, it facilitates the formation of out-breaking onset characteristics of SCS summer monsoon in the SCS area.
基金the Natural Science Foundation of Hubei Province(No.2014CFB247)the National Natural Science Foundation of China(Nos.61440016,61273225,61273303 and 31201121)the Project of the Key Laboratory for Metallurgical Equipment and Control of Ministry of Education(No.2013B08)
文摘Visual saliency is an important cue in human visual system to identify salient region in the image;it can be useful in many applications including image retrieval,object recognition,image segmentation,etc.Image contrast has been used as an effective feature to detect visual salient region.However,the conventional contrast measures either in spectral domain or in spatial domain fail to give sufficient consideration towards the local and global characteristics of the image.This paper presents a visual saliency detection algorithm based on a novel contrast measurement.This measurement extracts the spectral information of image block using the 2D discrete Fourier transform(DFT),and combines with the total variation(TV)of image block in spatial domain.The proposed algorithm is used to perform salient region detection in the image,and compared with state-of-the-art algorithms.The experimental results from the MSRA dataset validate the effectiveness of the proposed algorithm.
基金the National Natural Science Foundation of China(No:3 963 0 1 1 0 ) the National Key Technologies R&D Programme under Con-tract96-92 0 -1 2 -0 1
文摘An adaptive contrast enhancement (ACE) algorithm is presented in this paper, in which the contrast gain is determined by mapping the local standard deviation (LSD) histogram of an image to a Gaussian distribution function. The contrast gain is nonlinearly adjusted to avoid noise overenhancement and ringing artifacts while improving the detail contrast with less computational burden. The effectiveness of our method is demonstrated with radiological images and compared with other algorithms.
文摘To address the problems of missing inside and incomplete edge contours in camouflaged target detection results,we propose a camouflaged moving target detection algorithm based on local minimum difference constraints(LMDC).The algorithm first uses the mean to optimize the initial background model,removes the stable background region by global comparison,and extracts the edge point set in the potential target region so that each boundary point(seed)grows along the center of the target.Finally,we define the minor difference constraints term,combine the seed path and the target space consistency,and calculate the attributes of each pixel in the potential target area to realize camouflaged moving target detection.The algorithm of this paper is verified based on a public data sofa video and test videos and compared with the five classic algorithms.The experimental results show that the proposed algorithm yields good results based on integrity,accuracy,and a number of objective evaluation indexes,and its overall performance is better than that of the compared algorithms.
文摘This paper reports our study of a novel motion estimation algorithm based on global and local compensability analysis. The spatial correlation of motion field is used to reduce the burden of estimation computation and extra bit rate for motion vectors. Experimental results show that this algorithm is more efficient than the conventional methods, especially for temporal activity regions.
文摘In the past few years,there have been multiple advances in magnetic resonance (MR) instrumentation,in vivo devices,real-time imaging sequences and interventional procedures with new therapies.More recently,interventionists have started to use minimally invasive image-guided procedures and local therapies,which reduce the pain from conventional surgery and increase drug effectiveness,respectively.Local therapy also reduces the systemic dose and eliminates the toxic side effects of some drugs to other organs.The success of MR-guided procedures depends on visualization of the targets in 3D and precise deployment of ablation catheters,local therapies and devices.MR contrast media provide a wealth of tissue contrast and allows 3D and 4D image acquisitions.After the development of fast imaging sequences,the clinical applications of MR contrast media have been substantially expanded to include pre-during-and post-interventions.Prior to intervention,MR contrast media have the potential to localize and delineate pathologic tissues of vital organs,such as the brain,heart,breast,kidney,prostate,liver and uterus.They also offer other options such as labeling therapeutic agents or cells.During intervention,these agents have the capability to map blood vessels and enhance the contrast between the endovascular guidewire/catheters/devices,blood and tissues as well as direct therapies to the target.Furthermore,labeling therapeutic agents or cells aids in visualizing their delivery sites and tracking their tissue distribution.After intervention,MR contrast media have been used for assessing the efficacy of ablation and therapies.It should be noted that most image-guided procedures are under preclinical research and development.It can be concluded that MR contrast media have great value in preclinical and some clinical interventional procedures.Future applications of MR contrast media in image-guided procedures depend on their safety,tolerability,tissue specificity and effectiveness in demonstrating success of the interventions and therapies.
基金the Industry-University-Research Cooperation Fund Project of the Eighth Research Institute of China Aerospace Science and Technology Corporation(No.USCAST2021-5)。
文摘To overcome the obstacles of poor feature extraction and little prior information on the appearance of infrared dim small targets,we propose a multi-domain attention-guided pyramid network(MAGPNet).Specifically,we design three modules to ensure that salient features of small targets can be acquired and retained in the multi-scale feature maps.To improve the adaptability of the network for targets of different sizes,we design a kernel aggregation attention block with a receptive field attention branch and weight the feature maps under different perceptual fields with attention mechanism.Based on the research on human vision system,we further propose an adaptive local contrast measure module to enhance the local features of infrared small targets.With this parameterized component,we can implement the information aggregation of multi-scale contrast saliency maps.Finally,to fully utilize the information within spatial and channel domains in feature maps of different scales,we propose the mixed spatial-channel attention-guided fusion module to achieve high-quality fusion effects while ensuring that the small target features can be preserved at deep layers.Experiments on public datasets demonstrate that our MAGPNet can achieve a better performance over other state-of-the-art methods in terms of the intersection of union,Precision,Recall,and F-measure.In addition,we conduct detailed ablation studies to verify the effectiveness of each component in our network.
文摘【目的】跨视角对象级地理定位(CVOGL)旨在卫星影像上精确定位地面街景或无人机影像所观测目标的地理位置。现有方法多聚焦于图像级匹配,通过对整张影像全局处理实现跨视角关联,缺乏对特定目标的位置编码研究,导致无法将模型的注意力引导到感兴趣目标。并且由于参考图像覆盖范围的变化,查询目标在对应卫星图像中的像素占比极低,精确定位较为困难。【方法】针对以上问题,本文提出了一种基于高斯核函数与异构空间对比损失的跨视角对象级地理定位方法(Cross-View Object-Level Geo-Localization Method with Gaussian Kernel Function and Heterogeneous Spatial Contrastive Loss,GHGeo),用于精确定位感兴趣目标位置。该方法首先通过高斯核函数对查询目标进行精确位置编码,实现了对目标中心点及其分布特征的精细化建模;此外还提出了动态注意力精细化融合模块来动态加权交叉感知全局上下文与局部几何特征的空间相似性,以概率密度预测查询目标在卫星影像中的精确位置;最后通过异构空间对比损失函数来约束其训练过程,缓解跨视角特征差异。【结果】本文在CVOGL数据集进行了实验,实验结果显示:GHGeo在该数据集的“无人机-卫星”任务中,当交并比(IoU)≥25%和≥50%时定位准确率分别达到67.73%和63.00%,相较于基准方法DetGeo分别提升了5.76%和5.34%;在“街景-卫星”定位任务中,对应IoU阈值下的定位准确率分别为48.41%和45.43%的定位准确率,相较于基准方法DetGeo分别提升了2.98%和3.19%。同时与TransGeo,SAFA和VAGeo等方法在CVOGL数据集上进行对比,GHGeo则展现出了更高的定位准确性。【结论】本文方法有效提升了跨视角对象级地理定位方法的精度,为城市规划监测,应急救援调度等应用领域提供关键技术支持和精确位置信息支撑。
