The identification of objects in binary images is a fundamental task in image analysis and pattern recognition tasks. The Euler number of a binary image is an important topological measure which is used as a feature i...The identification of objects in binary images is a fundamental task in image analysis and pattern recognition tasks. The Euler number of a binary image is an important topological measure which is used as a feature in image analysis. In this paper, a very fast algorithm for the detection and localization of the objects and the computation of the Euler number of a binary image is proposed. The proposed algorithm operates in one scan of the image and is based on the Image Block Representation (IBR) scheme. The proposed algorithm is more efficient than conventional pixel based algorithms in terms of execution speed and representation of the extracted information.展开更多
针对坐标注意力(CA)在水平和垂直方向特征的平均池化可能丢失目标显著特征,以及使用二维普通卷积对小目标特征学习不足的情况,提出了CARFB(coordinate attention and receptive field block)模块。该模块将CA的平均池化修改为平均+最大...针对坐标注意力(CA)在水平和垂直方向特征的平均池化可能丢失目标显著特征,以及使用二维普通卷积对小目标特征学习不足的情况,提出了CARFB(coordinate attention and receptive field block)模块。该模块将CA的平均池化修改为平均+最大池化,以保留输入特征在水平和垂直方向的显著和细节信息;利用RFB具有不同大小感受野的优势,在水平和垂直方向分别使用RFB模块代替CA的融合特征统一卷积,以同时提取不同大小目标的特征;引入包含不同大小卷积核和步长的CBS模块,替换CA的二维普通卷积,进一步提取水平和垂直方向的特征,得到重新加权的输出特征。CARFB模块在水平和垂直方向保存目标位置信息,利用不同感受野提取不同大小目标的强辨别性特征,从而具有更强的特征学习能力。为了验证提出的即插即用模块CARFB的性能,将其嵌入ObjectBox目标检测框架,得到ObjectBox-CARFB模型;用CARFB模块替换RFBnet中的RFB模块,得到CARFBnet目标检测模型。MSCOCO数据集的实验测试表明,ObjectBox-CARFB模型的性能得到全面提升,尤其对小目标的检测性能提升突出;PASCALVOC和MSCOCO数据集的实验结果表明,CARFBnet300和CARFBnet512的目标检测能力分别优于原始RFBnet300和RFBnet512模型,并优于其他同系列对比模型。提出的CARFB模块具有更强的特征学习能力,对不同尺度目标均能取得较好的检测效果,特别是在小目标检测方面,效果提升显著。提出的CARFB模块可以嵌入到任何一个卷积神经网络,能保存更多的目标信息,具有更强的特征学习能力和更高的网络性能,对不同尺度目标均能取得较好的检测效果,尤其对小目标的检测效果提升显著。展开更多
In order to solve the tracking problem occurred during occlusions, an adaptive hierarchical block tracking method is proposed after analyzing the changes of the target characteristics under partial occlusions. Firstly...In order to solve the tracking problem occurred during occlusions, an adaptive hierarchical block tracking method is proposed after analyzing the changes of the target characteristics under partial occlusions. Firstly, color histogram features are selected to describe the target. The similarity between the target model and the candidates is measured by the Bhattacharyya coefficient, which can also be used to evaluate the degree of occlusions. The object is divided into four blocks when it is occluded, and the mean shift procedure is used to track each block separately. Then, according to the value of the Bhattacharyya coefficient, the partially occluded block is found and divided into four sub-blocks, which are tracked by block matching algorithm separately. Finally, the information of all the blocks is used to determine the displacement vector of the target. Experimental results show that compared to the traditional mean shift tracking method, this method can make full use of the features of the unoccluded sub-blocks, improve the tracking accuracy and solve the target tracking problem in case of partial occlusions.展开更多
文摘The identification of objects in binary images is a fundamental task in image analysis and pattern recognition tasks. The Euler number of a binary image is an important topological measure which is used as a feature in image analysis. In this paper, a very fast algorithm for the detection and localization of the objects and the computation of the Euler number of a binary image is proposed. The proposed algorithm operates in one scan of the image and is based on the Image Block Representation (IBR) scheme. The proposed algorithm is more efficient than conventional pixel based algorithms in terms of execution speed and representation of the extracted information.
文摘针对坐标注意力(CA)在水平和垂直方向特征的平均池化可能丢失目标显著特征,以及使用二维普通卷积对小目标特征学习不足的情况,提出了CARFB(coordinate attention and receptive field block)模块。该模块将CA的平均池化修改为平均+最大池化,以保留输入特征在水平和垂直方向的显著和细节信息;利用RFB具有不同大小感受野的优势,在水平和垂直方向分别使用RFB模块代替CA的融合特征统一卷积,以同时提取不同大小目标的特征;引入包含不同大小卷积核和步长的CBS模块,替换CA的二维普通卷积,进一步提取水平和垂直方向的特征,得到重新加权的输出特征。CARFB模块在水平和垂直方向保存目标位置信息,利用不同感受野提取不同大小目标的强辨别性特征,从而具有更强的特征学习能力。为了验证提出的即插即用模块CARFB的性能,将其嵌入ObjectBox目标检测框架,得到ObjectBox-CARFB模型;用CARFB模块替换RFBnet中的RFB模块,得到CARFBnet目标检测模型。MSCOCO数据集的实验测试表明,ObjectBox-CARFB模型的性能得到全面提升,尤其对小目标的检测性能提升突出;PASCALVOC和MSCOCO数据集的实验结果表明,CARFBnet300和CARFBnet512的目标检测能力分别优于原始RFBnet300和RFBnet512模型,并优于其他同系列对比模型。提出的CARFB模块具有更强的特征学习能力,对不同尺度目标均能取得较好的检测效果,特别是在小目标检测方面,效果提升显著。提出的CARFB模块可以嵌入到任何一个卷积神经网络,能保存更多的目标信息,具有更强的特征学习能力和更高的网络性能,对不同尺度目标均能取得较好的检测效果,尤其对小目标的检测效果提升显著。
基金Supported by State Key Laboratory of Explosion Science and Technology Foundation(ZDKT08-05)
文摘In order to solve the tracking problem occurred during occlusions, an adaptive hierarchical block tracking method is proposed after analyzing the changes of the target characteristics under partial occlusions. Firstly, color histogram features are selected to describe the target. The similarity between the target model and the candidates is measured by the Bhattacharyya coefficient, which can also be used to evaluate the degree of occlusions. The object is divided into four blocks when it is occluded, and the mean shift procedure is used to track each block separately. Then, according to the value of the Bhattacharyya coefficient, the partially occluded block is found and divided into four sub-blocks, which are tracked by block matching algorithm separately. Finally, the information of all the blocks is used to determine the displacement vector of the target. Experimental results show that compared to the traditional mean shift tracking method, this method can make full use of the features of the unoccluded sub-blocks, improve the tracking accuracy and solve the target tracking problem in case of partial occlusions.
