WO2023273337A1 - Representative feature-based method for detecting dense targets in remote sensing image - Google Patents
Representative feature-based method for detecting dense targets in remote sensing image Download PDFInfo
- Publication number
- WO2023273337A1 WO2023273337A1 PCT/CN2022/074542 CN2022074542W WO2023273337A1 WO 2023273337 A1 WO2023273337 A1 WO 2023273337A1 CN 2022074542 W CN2022074542 W CN 2022074542W WO 2023273337 A1 WO2023273337 A1 WO 2023273337A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- feature
- similarity
- network
- feature map
- category
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 17
- 239000013598 vector Substances 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 238000001514 detection method Methods 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 238000000691 measurement method Methods 0.000 claims description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000004880 explosion Methods 0.000 claims description 3
- 239000000523 sample Substances 0.000 description 8
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 240000004050 Pentaglottis sempervirens Species 0.000 description 1
- 235000004522 Pentaglottis sempervirens Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Definitions
- the invention relates to target detection, in particular to a dense target detection method in remote sensing images based on representative features.
- Remote sensing technology is a rapidly developing high-tech.
- the information network it forms provides people with a large amount of scientific data and dynamic information.
- Remote sensing image detection is the benchmark problem of target detection. It is used in many fields such as agriculture, meteorological surveying and mapping, and environmental protection. It has great application value.
- the target detection method based on deep learning, the target category corresponding to the ground truth label, the The sample is a positive sample, and the positive sample with a large error between the predicted value of the target classification confidence and the true value label is a difficult sample.
- Existing excellent detection models can detect most objects in the image, but often miss some difficult positive samples that are more difficult to detect.
- the difficult positive samples When detecting difficult samples, when the classification confidence of the target detection model for the positive sample prediction is lower than the set confidence threshold, the difficult positive samples will be filtered out in the post-processing stage, resulting in a decrease in the detection performance of the detection model; or artificially Lowering the confidence threshold in the post-processing stage of the network makes the detection model lose the ability to suppress low-confidence negative samples. Therefore, it is more challenging to accurately detect densely arranged multiple objects in remote sensing images.
- the purpose of the present invention is to provide a dense target detection method in remote sensing images based on representative features, by adaptively increasing the classification confidence of difficult positive samples, and then accurately detecting densely arranged multiple targets in remote sensing images. similar objects.
- a dense target detection method in remote sensing images based on representative features of the present invention comprising the following steps:
- Construct four network modules including a feature extraction network, a feature pyramid network, a preliminary prediction network and a final prediction network, input the remote sensing image to be detected into the feature extraction network and the feature pyramid network in turn, and output a preliminary feature map;
- step 3 Take the similarity obtained in step 3 as the weight, and adaptively increase the classification confidence based on the classification confidence of difficult positive samples, as the final classification confidence of difficult positive samples.
- step 2 the process of obtaining the highest classification confidence and representative features in step 2 is:
- the similarity in the step 3 includes feature semantic similarity and feature space similarity
- the feature semantic similarity calculation process includes:
- the described embedded Gaussian similarity measurement function is:
- RF k represents the representative feature RepFeature k of the k-th category
- F hw represents the feature vector of the h-th row and w-th column in the feature map FM ODM output by the final prediction network Feature vector RepFeature k
- i represents the eigenvalue of the i-th dimension in n dimensions
- W ⁇ , W ⁇ are learning weight matrices; ⁇ (RF k ) i , ⁇ (F hw ) i respectively represent the eigenvalues of the two eigenvectors in each dimension;
- N( ⁇ (RF)) is the normalization factor, by calculating the sum of the similarities between the feature vector F hw of the hth row and wth column in the final prediction network and K effective representative features RF k , K is the data The number of categories in the set, and the embedded Gaussian similarity is normalized to a range of 0 to 1 to avoid the gradient explosion problem caused by excessive similarity.
- the formula for calculating the normalization factor is as follows;
- the feature space similarity calculation process includes the following steps:
- Spatial_i means that RF k and F hw are taken from the i-th layer feature map from the bottom up of the feature pyramid network, and ⁇ is the scale parameter;
- the final classification confidence of the difficult positive sample in step 4 is added by weight to the confidence of the position at (h, w) of the final prediction network feature map with respect to category k by representing the confidence of category k Realized above, the calculation formula is:
- the method for measuring the feature semantic similarity includes using any one of Euclidean similarity, cosine similarity or Gaussian similarity.
- the feature extraction network uses a convolutional layer to reduce the size of the original image, and the extracted effective features are input to the feature pyramid network; the feature extraction network selects a ResNet or HRNet convolutional neural network.
- the preliminary prediction network selects the feature alignment module in the S 2 A-NET model to preliminarily predict the category information and location information of the object.
- the final prediction network selects the rotation detection module in the S 2 A-NET model to predict the final category information and position information of the object.
- the present invention uses representative features and representative confidence to adaptively improve the classification confidence of difficult positive samples, and improves the classification ability of difficult positive samples in dense remote sensing image scenes;
- Fig. 1 is a schematic diagram of the representative feature acquisition process of the present invention
- Fig. 2 is the flow chart of computing similarity of the present invention
- Fig. 3 is a schematic diagram of improving classification confidence for difficult positive samples in the present invention.
- (1) Construct four network modules, including feature extraction network, feature pyramid network, preliminary prediction network and final prediction network, input the remote sensing image to be detected into the feature extraction network, use the convolutional layer to reduce the size of the original image, feature extraction The network inputs the extracted effective features into the feature pyramid network, and then outputs the preliminary feature map FM FAM .
- the feature extraction network selects ResNet or HRNet convolutional neural network; the preliminary prediction network selects the feature alignment module FAM in the S 2 A-NET model; the final prediction network selects the rotation detection module ODM in the S 2 A-NET model.
- the classification confidence threshold is set to 0.6. Only when the representative confidence of category k is greater than 0.6, The representative features of category k are effective representative features. When RepConfidence k is low, such as 0.3, 0.4, the probability of RepFeature k itself belonging to category k is also low, and cannot be an effective representative feature.
- the measurement method of feature semantic similarity can adopt any one of Euclidean similarity, cosine similarity or Gaussian similarity.
- Gaussian similarity is used to calculate feature semantic similarity. The process is as follows:
- the embedded Gaussian similarity measurement function is used to calculate the similarity of feature semantic information, and the measurement method used is normalized.
- the embedded Gaussian similarity measurement function is:
- RF k represents the representative feature RepFeature k of the k-th category
- F hw represents the feature vector of the h-th row and w-th column in the feature map FM ODM output by the final prediction network Feature vector RepFeature k
- n of the feature vector in this embodiment is 256
- i represents the feature value of the i-th dimension in the n dimensions
- W ⁇ , W ⁇ are learning weight matrices; ⁇ (RF k ) i , ⁇ (F hw ) i respectively represent the eigenvalues of the two eigenvectors in each dimension;
- N( ⁇ (RF)) is the normalization factor.
- the feature space similarity calculation process includes the following steps:
- Spatial_i means that RF k and F hw are taken from the i-th layer feature map from the bottom up of the feature pyramid network, and ⁇ is a scale parameter.
- ⁇ is set to 1/64 so that two features with closer distances can be There is a high spatial location correlation.
- step 3 Take the similarity obtained in step 3 as the weight, and adaptively increase the classification confidence based on the classification confidence of difficult positive samples, as the final classification confidence of difficult positive samples, as shown in Figure 3.
- the final classification confidence of the difficult positive sample is weighted by adding the representative confidence of category k to the confidence of category k at the (h, w)th position of the final prediction network feature map Realized above, the calculation formula is:
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
A representative feature-based method for detecting dense targets in a remote sensing image, comprising: constructing a feature extraction network, a feature pyramid network, a preliminary prediction network, and a final prediction network, and sequentially inputting a remote sensing image to be detected into the feature extraction network and the feature pyramid network to output a preliminary feature map; inputting the preliminary feature map into the preliminary prediction network, and selecting a representative feature of semantic information of each category from all categories in a data set and representative confidence of each category in the whole feature map; inputting a feature map outputted by the preliminary prediction network into the final prediction network to obtain a final feature map, and calculating a similarity between the representative feature of a same category and a feature vector at the same position of the final feature map; and by taking the similarity as a weight, adaptively improving classification confidence on the basis of classification confidence of a hard positive sample.
Description
本发明涉及目标检测,特别涉及一种基于代表特征的遥感图像中的密集目标检测方法。The invention relates to target detection, in particular to a dense target detection method in remote sensing images based on representative features.
遥感技术是一种发展迅速的高新技术,它所形成的信息网络为人们提供大量的科学数据和动态信息,遥感图像检测是目标检测的基准问题,在农业、气象测绘、环境保护等多个领域有着很大的应用价值。Remote sensing technology is a rapidly developing high-tech. The information network it forms provides people with a large amount of scientific data and dynamic information. Remote sensing image detection is the benchmark problem of target detection. It is used in many fields such as agriculture, meteorological surveying and mapping, and environmental protection. It has great application value.
随着深度学习算法在计算机视觉领域取得的极大成功,已被认为是遥感图像处理的首选方法。由于俯瞰视角拍摄和更大的空间视野,遥感图像中存在更多的密集场景并包含大量密集排列的物体,在基于深度学习的目标检测方法中,与真值标签对应的目标类别来说,该样本为正样本,对于目标分类置信度的预测值与真值标签误差较大的正样本为困难样本。现有的优良检测模型可以检测出图像中的多数物体,却往往会漏掉其中检测难度较大的部分困难正样本。当检测困难样本时,目标检测模型对于正样本预测的分类置信度低于设定的置信度阈值时,困难正样本会在后处理阶段被滤除,从而导致检测模型检测性能的降低;或者人为在网络的后处理阶段降低置信度阈值,则使得检测模型失去抑制低置信度负样本的能力。所以,在遥感图像中准确地检测出密集排列的多个物体具有更多挑战性。With the great success of deep learning algorithm in the field of computer vision, it has been considered as the preferred method for remote sensing image processing. Due to the bird's-eye view and larger spatial field of view, there are more dense scenes in remote sensing images and contain a large number of densely arranged objects. In the target detection method based on deep learning, the target category corresponding to the ground truth label, the The sample is a positive sample, and the positive sample with a large error between the predicted value of the target classification confidence and the true value label is a difficult sample. Existing excellent detection models can detect most objects in the image, but often miss some difficult positive samples that are more difficult to detect. When detecting difficult samples, when the classification confidence of the target detection model for the positive sample prediction is lower than the set confidence threshold, the difficult positive samples will be filtered out in the post-processing stage, resulting in a decrease in the detection performance of the detection model; or artificially Lowering the confidence threshold in the post-processing stage of the network makes the detection model lose the ability to suppress low-confidence negative samples. Therefore, it is more challenging to accurately detect densely arranged multiple objects in remote sensing images.
发明内容Contents of the invention
发明目的:针对以上问题,本发明目的是提供一种基于代表特征的遥感图像中的密集目标检测方法,通过自适应增加困难正样本的分类置信度,进而准确检测出遥感图像中密集排列的多个同类物体。Purpose of the invention: In view of the above problems, the purpose of the present invention is to provide a dense target detection method in remote sensing images based on representative features, by adaptively increasing the classification confidence of difficult positive samples, and then accurately detecting densely arranged multiple targets in remote sensing images. similar objects.
技术方案:本发明的一种基于代表特征的遥感图像中的密集目标检测方法,包括如下步骤:Technical solution: A dense target detection method in remote sensing images based on representative features of the present invention, comprising the following steps:
(1)构建四个网络模块,包括特征提取网络、特征金字塔网络、初步预测网络和最终预测网络,将待检测遥感图像依次输入到特征提取网络和特征金子塔网络中,输出初步特征图;(1) Construct four network modules, including a feature extraction network, a feature pyramid network, a preliminary prediction network and a final prediction network, input the remote sensing image to be detected into the feature extraction network and the feature pyramid network in turn, and output a preliminary feature map;
(2)将初步特征图输入到初步预测网络中,在数据集所有类别中选取每个类别语义信息的代表特征和各个类别在整张特征图中代表置信度;(2) Input the preliminary feature map into the preliminary prediction network, select the representative features of each category semantic information in all categories of the data set and each category represents the confidence in the entire feature map;
(3)将初步预测网络输出的特征图输入到最终预测网络中,得到最终特征图,计算同类别的代表特征与最终特征图相同位置特征向量之间的相似度;(3) Input the feature map output by the preliminary prediction network into the final prediction network to obtain the final feature map, and calculate the similarity between the representative features of the same category and the same position feature vector of the final feature map;
(4)以步骤3得到的相似度为权重,在困难正样本分类置信度基础上自适应提升分类置信度,作为困难正样本的最终分类置信度。(4) Take the similarity obtained in step 3 as the weight, and adaptively increase the classification confidence based on the classification confidence of difficult positive samples, as the final classification confidence of difficult positive samples.
进一步,所述步骤2得到最高分类置信度和代表特征的过程为:Further, the process of obtaining the highest classification confidence and representative features in step 2 is:
(201)在初步预测网络的分类支路中,计算整张特征图H×W位置处各类别的分类置信度
其中H为特征图的长度,W为宽度,k为数据集的类别;
(201) In the classification branch of the preliminary prediction network, calculate the classification confidence of each category at the H×W position of the entire feature map Where H is the length of the feature map, W is the width, and k is the category of the data set;
(202)在
中找出最高的分类置信度作为类别k的代表置信度RepConfidences,找出取得最高分类置信度的位置(h,w),其中h为长度,w为宽度;
(202) at Find the highest classification confidence as the representative confidence RepConfidences of category k, and find the position (h, w) where the highest classification confidence is obtained, where h is the length and w is the width;
(203)在初步特征图FM
FAM中提取第h行、第w列的特征信息
用以表示类别k的代表特征RepFeature
k,其中FM
FAM是初步预测网络的分类支路和回归支路所共享的前一层特征图;
(203) Extract the feature information of the hth row and wth column in the preliminary feature map FM FAM It is used to represent the representative feature RepFeature k of category k, where FM FAM is the previous layer feature map shared by the classification branch and the regression branch of the preliminary prediction network;
(204)设置分类置信度阈值,只有当类别k的代表置信度大于分类置信度阈值时,类别k的代表特征才为有效的代表特征。(204) Set a classification confidence threshold, and only when the representative confidence of category k is greater than the classification confidence threshold, the representative feature of category k is an effective representative feature.
进一步,所述步骤3中相似度包括特征语义相似度和特征空间相似度,所述特征语义相似度计算过程包括:Further, the similarity in the step 3 includes feature semantic similarity and feature space similarity, and the feature semantic similarity calculation process includes:
采用嵌入高斯相似性度量函数计算特征语义信息相似性,并对采用的度量方法进行归一化,所述的嵌入高斯相似性度量函数为:Using the embedded Gaussian similarity measurement function to calculate the feature semantic information similarity, and normalize the measurement method adopted, the described embedded Gaussian similarity measurement function is:
其中RF
k表示第k种类别的代表特征RepFeature
k,F
hw表示最终预测网络输出的特征图FM
ODM中第h行、第w列的特征向量
特征向量RepFeature
k、
均为1×1×n维,i表示n个维度中第i个维度的特征值;
Where RF k represents the representative feature RepFeature k of the k-th category, and F hw represents the feature vector of the h-th row and w-th column in the feature map FM ODM output by the final prediction network Feature vector RepFeature k , Both are 1×1×n-dimensional, and i represents the eigenvalue of the i-th dimension in n dimensions;
采用线性嵌入空间的形式:Take the form of a linear embedding space:
φ(RF
k)=W
φRF
k
φ(RF k )=W φ RF k
θ(F
hw)=W
θF
hw
θ(F hw )=W θ F hw
其中W
φ、W
θ是学习权重矩阵;φ(RF
k)
i、θ(F
hw)
i分别表示两个特征向量在每个维度中的特征值;
Where W φ , W θ are learning weight matrices; φ(RF k ) i , θ(F hw ) i respectively represent the eigenvalues of the two eigenvectors in each dimension;
N(φ(RF))为归一化因子,通过计算最终预测网络中第h行、第w列的特征向量F
hw分别与K个有效代表特征RF
k的相似度的加和,K为数据集的类别数目,将 嵌入高斯相似度归一化为0到1的范围内,以避免相似度过高而产生的梯度***问题,归一化因子计算公式如式为;
N(φ(RF)) is the normalization factor, by calculating the sum of the similarities between the feature vector F hw of the hth row and wth column in the final prediction network and K effective representative features RF k , K is the data The number of categories in the set, and the embedded Gaussian similarity is normalized to a range of 0 to 1 to avoid the gradient explosion problem caused by excessive similarity. The formula for calculating the normalization factor is as follows;
进一步,所述特征空间相似度计算过程包括如下步骤:Further, the feature space similarity calculation process includes the following steps:
(301)计算特征向量RepFeature
k和
在特征图维度上的空间距离dis(RF
k,F
hw),计算公式为:
(301) Calculate feature vector RepFeature k and The spatial distance dis(RF k ,F hw ) in the dimension of the feature map is calculated as:
其中
为特征向量RepFeature
k在特征图中的横、纵坐标,
为特征向量
在特征图中的横、纵坐标;
in is the abscissa and ordinate of the feature vector RepFeature k in the feature map, is the eigenvector The horizontal and vertical coordinates in the feature map;
(302)利用dis(RF
k,F
hw)乘以各个特征图的步长stride
i得到两个特征向量在原始图像上的空间距离Corr
Spatial_i(RF
k,F
hw),计算公式为:
(302) Multiply dis(RF k , F hw ) by the stride i of each feature map to obtain the spatial distance Corr Spatial_i (RF k , F hw ) of the two feature vectors on the original image, the calculation formula is:
其中,Spatial_i表示RF
k、F
hw取自特征金字塔网络自底向上的第i层特征图,α是尺度参数;
Among them, Spatial_i means that RF k and F hw are taken from the i-th layer feature map from the bottom up of the feature pyramid network, and α is the scale parameter;
所以,所述步骤3中的相似度表达式为:Therefore, the similarity expression in the step 3 is:
Similarity(RF
k,F
hw)
Similarity(RF k ,F hw )
=Sim
Embedded_Gaussian(RF
k,F
hw)+Corr
Spatial_i(RF
k,F
hw)
=Sim Embedded_Gaussian (RF k ,F hw )+Corr Spatial_i (RF k ,F hw )
进一步,所述步骤4困难正样本最终的分类置信度通过将类别k的代表置信度按权重加到最终预测网络特征图的第(h,w)处位置关于类别k的置信度
上实现的,计算公式为:
Further, the final classification confidence of the difficult positive sample in step 4 is added by weight to the confidence of the position at (h, w) of the final prediction network feature map with respect to category k by representing the confidence of category k Realized above, the calculation formula is:
进一步,所述特征语义相似度的度量方法包括采用欧式相似度、余弦相似度或高斯相似度中的任一种。Further, the method for measuring the feature semantic similarity includes using any one of Euclidean similarity, cosine similarity or Gaussian similarity.
进一步,所述特征提取网络采用卷积层缩小原始图像的尺寸,并将提取的有效特征输入到特征金字塔网络;所述特征提取网络选取ResNet或HRNet卷积神经网络。Further, the feature extraction network uses a convolutional layer to reduce the size of the original image, and the extracted effective features are input to the feature pyramid network; the feature extraction network selects a ResNet or HRNet convolutional neural network.
进一步,所述初步预测网络选取S
2A-NET模型中的特征对齐模块,初步预测 物体的类别信息和位置信息。
Further, the preliminary prediction network selects the feature alignment module in the S 2 A-NET model to preliminarily predict the category information and location information of the object.
进一步,所述最终预测网络选取S
2A-NET模型中的旋转检测模块,预测物体最终的类别信息和位置信息。
Further, the final prediction network selects the rotation detection module in the S 2 A-NET model to predict the final category information and position information of the object.
有益效果:本发明与现有技术相比,其显著优点是:Beneficial effect: the present invention compares with prior art, and its remarkable advantage is:
1、本发明利用代表特征和代表置信度自适应提升了困难正样本分类置信度,提升了遥感图像密集场景下的困难正样本的分类能力;1. The present invention uses representative features and representative confidence to adaptively improve the classification confidence of difficult positive samples, and improves the classification ability of difficult positive samples in dense remote sensing image scenes;
2、利用两个阶段分类支路参数确保相似度计算过程的一致性,减少检测模型的复杂度以及网络参数量。2. Use the two-stage classification branch parameters to ensure the consistency of the similarity calculation process and reduce the complexity of the detection model and the amount of network parameters.
图1为本发明代表特征获取过程示意图;Fig. 1 is a schematic diagram of the representative feature acquisition process of the present invention;
图2为本发明计算相似度流程图;Fig. 2 is the flow chart of computing similarity of the present invention;
图3为本发明困难正样本提升分类置信度示意图。Fig. 3 is a schematic diagram of improving classification confidence for difficult positive samples in the present invention.
本实施例所述的本一种基于代表特征的遥感图像中的密集目标检测方法,包括如下步骤:The dense target detection method in a remote sensing image based on representative features described in this embodiment includes the following steps:
(1)构建四个网络模块,包括特征提取网络、特征金字塔网络、初步预测网络和最终预测网络,将待检测遥感图像输入到特征提取网络中,采用卷积层缩小原始图像的尺寸,特征提取网络将提取的有效特征输入到特征金字塔网络中,然后输出初步特征图FM
FAM。
(1) Construct four network modules, including feature extraction network, feature pyramid network, preliminary prediction network and final prediction network, input the remote sensing image to be detected into the feature extraction network, use the convolutional layer to reduce the size of the original image, feature extraction The network inputs the extracted effective features into the feature pyramid network, and then outputs the preliminary feature map FM FAM .
其中特征提取网络选取ResNet或HRNet卷积神经网络;初步预测网络选取S
2A-NET模型中的特征对齐模块FAM;最终预测网络选取S
2A-NET模型中的旋转检测模块ODM。
The feature extraction network selects ResNet or HRNet convolutional neural network; the preliminary prediction network selects the feature alignment module FAM in the S 2 A-NET model; the final prediction network selects the rotation detection module ODM in the S 2 A-NET model.
(2)将初步特征图FM
FAM输入到特征对齐模块FAM中,在数据集所有类别中选取每个类别语义信息的代表特征和各个类别在整张特征图中代表置信度,如图1所示,过程为:
(2) Input the preliminary feature map FM FAM into the feature alignment module FAM, select the representative features of each category semantic information in all categories of the data set and each category represents the confidence in the entire feature map, as shown in Figure 1 , the process is:
(201)在初步预测网络的分类支路中,计算整张特征图H×W位置处各类别的分类置信度
其中H为特征图的长度,W为宽度,k为数据集的类别,本实施例数据集包含15种物体类别,对15种类别依次进行计算;
(201) In the classification branch of the preliminary prediction network, calculate the classification confidence of each category at the H×W position of the entire feature map Wherein H is the length of the feature map, W is the width, and k is the category of the data set. The data set of this embodiment contains 15 object categories, and the 15 categories are calculated sequentially;
(202)在
中找出最高的分类置信度作为类别k的代表置信度RepConfidences,找出取得最高分类置信度的位置(h,w),其中h为长度,w为宽度;
(202) at Find the highest classification confidence as the representative confidence RepConfidences of category k, and find the position (h, w) where the highest classification confidence is obtained, where h is the length and w is the width;
(203)在初步特征图FM
FAM中提取第h行、第w列的特征信息
用以表 示类别k的代表特征RepFeature
k,其中FM
FAM是初步预测网络的分类支路和回归支路所共享的前一层特征图;FM
FAM同时包含物体类别和位置信息的相关特征,用于后续特征之间相似度的计算;特征图FM
FAM为H×W×C维,其中H、W为特征图的长、宽,C为特征图的通道数,本实施例中C为256;
(203) Extract the feature information of the hth row and wth column in the preliminary feature map FM FAM It is used to represent the representative feature RepFeature k of category k, where FM FAM is the previous layer feature map shared by the classification branch and regression branch of the preliminary prediction network; FM FAM contains related features of object category and location information at the same time, used Calculation of the similarity between subsequent features; the feature map FM FAM is H×W×C dimension, where H and W are the length and width of the feature map, and C is the number of channels of the feature map. In this embodiment, C is 256;
(204)设置分类置信度阈值以保证代表特征的可靠性和成为代表特征的难度之间的最佳平衡,本实施例中将阈值设置为0.6,只有当类别k的代表置信度大于0.6时,类别k的代表特征才为有效的代表特征。当RepConfidence
k较低如0.3、0.4时,RepFeature
k自身属于类别k的概率也较低,无法成为有效的代表特征。
(204) Set the classification confidence threshold to ensure the best balance between the reliability of the representative feature and the difficulty of becoming a representative feature. In this embodiment, the threshold is set to 0.6. Only when the representative confidence of category k is greater than 0.6, The representative features of category k are effective representative features. When RepConfidence k is low, such as 0.3, 0.4, the probability of RepFeature k itself belonging to category k is also low, and cannot be an effective representative feature.
(3)将初步预测网络输出的特征图输入到最终预测网络中,得到最终特征图,计算同类别的代表特征与最终特征图相同位置特征向量之间的相似度,流程图如图2所示。相似度包括特征语义相似度和特征空间相似度,(3) Input the feature map output by the preliminary prediction network into the final prediction network to obtain the final feature map, and calculate the similarity between the representative features of the same category and the feature vectors at the same position in the final feature map. The flow chart is shown in Figure 2 . Similarity includes feature semantic similarity and feature space similarity,
特征语义相似度的度量方法可以采用欧式相似度、余弦相似度或高斯相似度中的任一种。本实施采用高斯相似度来进行特征语义相似度计算,过程如下:The measurement method of feature semantic similarity can adopt any one of Euclidean similarity, cosine similarity or Gaussian similarity. In this implementation, Gaussian similarity is used to calculate feature semantic similarity. The process is as follows:
采用嵌入高斯相似性度量函数计算特征语义信息相似性,并对采用的度量方法进行归一化,嵌入高斯相似性度量函数为:The embedded Gaussian similarity measurement function is used to calculate the similarity of feature semantic information, and the measurement method used is normalized. The embedded Gaussian similarity measurement function is:
其中RF
k表示第k种类别的代表特征RepFeature
k,F
hw表示最终预测网络输出的特征图FM
ODM中第h行、第w列的特征向量
特征向量RepFeature
k、
均为1×1×n维,本实施例中特征向量的维度n取256,i表示n个维度中第i个维度的特征值;
Where RF k represents the representative feature RepFeature k of the k-th category, and F hw represents the feature vector of the h-th row and w-th column in the feature map FM ODM output by the final prediction network Feature vector RepFeature k , Both are 1×1×n dimensions, and the dimension n of the feature vector in this embodiment is 256, and i represents the feature value of the i-th dimension in the n dimensions;
采用线性嵌入空间的形式:Take the form of a linear embedding space:
φ(RF
k)=W
φRF
k
φ(RF k )=W φ RF k
θ(F
hw)=W
θF
hw
θ(F hw )=W θ F hw
其中W
φ、W
θ是学习权重矩阵;φ(RF
k)
i、θ(F
hw)
i分别表示两个特征向量在每个维度中的特征值;
Where W φ , W θ are learning weight matrices; φ(RF k ) i , θ(F hw ) i respectively represent the eigenvalues of the two eigenvectors in each dimension;
N(φ(RF))为归一化因子,通过计算最终预测网络中第h行、第w列的特征向量F
hw分别与15个有效代表特征RF
k的相似度的加和,将嵌入高斯相似度归一化为0到1的范围内,以避免相似度过高而产生的梯度***问题,归一化因子计算公式如式为;
N(φ(RF)) is the normalization factor. By calculating the sum of the similarities between the feature vector F hw of the hth row and wth column in the final prediction network and the 15 effective representative features RF k , the embedded Gaussian The similarity is normalized to a range of 0 to 1 to avoid the gradient explosion problem caused by excessive similarity. The formula for calculating the normalization factor is as follows:
特征空间相似度计算过程包括如下步骤:The feature space similarity calculation process includes the following steps:
(301)计算特征向量RepFeature
k和
在特征图维度上的空间距离dis(RF
k,F
hw),计算公式为:
(301) Calculate feature vector RepFeature k and The spatial distance dis(RF k ,F hw ) in the dimension of the feature map is calculated as:
其中
为特征向量RepFeature
k在特征图中的横、纵坐标,
为特征向量
在特征图中的横、纵坐标;训练模型时,使用特征金子塔网络自底向上感受5层特征图进行预测,5层特征图的步长stride
i取值分别是8,16,32,64,128;
in is the abscissa and ordinate of the feature vector RepFeature k in the feature map, is the eigenvector The horizontal and vertical coordinates in the feature map; when training the model, use the feature pyramid network to experience the 5-layer feature map from the bottom up for prediction, and the stride i values of the 5-layer feature map are 8, 16, 32, 64, 128 respectively ;
(302)利用dis(RF
k,F
hw)乘以各个特征图的步长stride
i得到两个特征向量在原始图像上的空间距离Corr
Spatial_i(RF
k,F
hw),计算公式为:
(302) Multiply dis(RF k , F hw ) by the stride i of each feature map to obtain the spatial distance Corr Spatial_i (RF k , F hw ) of the two feature vectors on the original image, the calculation formula is:
其中,Spatial_i表示RF
k、F
hw取自特征金字塔网络自底向上的第i层特征图,α是尺度参数,本实施例中将α设为1/64以使得两个距离较近的特征可以有较高的空间位置相关性。
Among them, Spatial_i means that RF k and F hw are taken from the i-th layer feature map from the bottom up of the feature pyramid network, and α is a scale parameter. In this embodiment, α is set to 1/64 so that two features with closer distances can be There is a high spatial location correlation.
相似度表达式为:The similarity expression is:
Similarity(RF
k,F
hw)
Similarity(RF k ,F hw )
=Sim
Embedded_Gaussian(RF
k,F
hw)+Corr
Spatial_i(RF
k,F
hw)
=Sim Embedded_Gaussian (RF k ,F hw )+Corr Spatial_i (RF k ,F hw )
(4)以步骤3得到的相似度为权重,在困难正样本分类置信度基础上自适应提升分类置信度,作为困难正样本的最终分类置信度,示意图如图3所示。(4) Take the similarity obtained in step 3 as the weight, and adaptively increase the classification confidence based on the classification confidence of difficult positive samples, as the final classification confidence of difficult positive samples, as shown in Figure 3.
困难正样本最终的分类置信度通过将类别k的代表置信度按权重加到最终预测网络特征图的第(h,w)处位置关于类别k的置信度
上实现的,计算公式为:
The final classification confidence of the difficult positive sample is weighted by adding the representative confidence of category k to the confidence of category k at the (h, w)th position of the final prediction network feature map Realized above, the calculation formula is:
Claims (9)
- 一种基于代表特征的遥感图像中的密集目标检测方法,其特征在于,包括如下步骤:A dense target detection method in remote sensing images based on representative features, characterized in that it comprises the following steps:(1)构建四个网络模块,包括特征提取网络、特征金字塔网络、初步预测网络和最终预测网络,将待检测遥感图像依次输入到特征提取网络和特征金子塔网络中,输出初步特征图;(1) Construct four network modules, including a feature extraction network, a feature pyramid network, a preliminary prediction network and a final prediction network, input the remote sensing image to be detected into the feature extraction network and the feature pyramid network in turn, and output a preliminary feature map;(2)将初步特征图输入到初步预测网络中,在数据集所有类别中选取每个类别语义信息的代表特征和各个类别在整张特征图中代表置信度;(2) Input the preliminary feature map into the preliminary prediction network, select the representative features of each category semantic information in all categories of the data set and each category represents the confidence in the entire feature map;(3)将初步预测网络输出的特征图输入到最终预测网络中,得到最终特征图,计算同类别的代表特征与最终特征图相同位置特征向量之间的相似度;(3) Input the feature map output by the preliminary prediction network into the final prediction network to obtain the final feature map, and calculate the similarity between the representative features of the same category and the same position feature vector of the final feature map;(4)以步骤3得到的相似度为权重,在困难正样本分类置信度基础上自适应提升分类置信度,作为困难正样本的最终分类置信度。(4) Take the similarity obtained in step 3 as the weight, and adaptively increase the classification confidence based on the classification confidence of difficult positive samples, as the final classification confidence of difficult positive samples.
- 根据权利要求1所述的密集目标检测方法,其特征在于,所述步骤2得到最高分类置信度和代表特征的过程为:The dense target detection method according to claim 1, wherein the process of obtaining the highest classification confidence and representative features in the step 2 is:(201)在初步预测网络的分类支路中,计算整张特征图H×W位置处各类别的分类置信度 其中H为特征图的长度,W为宽度,k为数据集的类别; (201) In the classification branch of the preliminary prediction network, calculate the classification confidence of each category at the H×W position of the entire feature map Where H is the length of the feature map, W is the width, and k is the category of the data set;(202)在 中找出最高的分类置信度作为类别k的代表置信度RepConfidences,找出取得最高分类置信度的位置(h,w),其中h为长度,w为宽度; (202) at Find the highest classification confidence as the representative confidence RepConfidences of category k, and find the position (h, w) where the highest classification confidence is obtained, where h is the length and w is the width;(203)在初步特征图FM FAM中提取第h行、第w列的特征信息 用以表示类别k的代表特征RepFeature k,其中FM FAM是初步预测网络的分类支路和回归支路所共享的前一层特征图; (203) Extract the feature information of the hth row and wth column in the preliminary feature map FM FAM It is used to represent the representative feature RepFeature k of category k, where FM FAM is the previous layer feature map shared by the classification branch and the regression branch of the preliminary prediction network;(204)设置分类置信度阈值,只有当类别k的代表置信度大于分类置信度阈值时,类别k的代表特征才为有效的代表特征。(204) Set a classification confidence threshold, and only when the representative confidence of category k is greater than the classification confidence threshold, the representative feature of category k is an effective representative feature.
- 根据权利要求2所述的密集目标检测方法,其特征在于,所述步骤3中相似度包括特征语义相似度和特征空间相似度,所述特征语义相似度计算过程包括:The dense target detection method according to claim 2, wherein the similarity in the step 3 includes a feature semantic similarity and a feature space similarity, and the feature semantic similarity calculation process includes:采用嵌入高斯相似性度量函数计算特征语义信息相似性,并对采用的度量方法进行归一化,所述的嵌入高斯相似性度量函数为:The embedding Gaussian similarity measurement function is used to calculate the similarity of feature semantic information, and the measurement method adopted is normalized. The embedded Gaussian similarity measurement function is:其中RF k表示第k种类别的代表特征RepFeature k,F hw表示最终预测网络输出的 特征图FM ODM中第h行、第w列的特征向量 特征向量RepFeature k、 均为1×1×n维,i表示n个维度中第i个维度的特征值; Where RF k represents the representative feature RepFeature k of the k-th category, and F hw represents the feature vector of the h-th row and w-th column in the feature map FM ODM output by the final prediction network Feature vector RepFeature k , Both are 1×1×n-dimensional, and i represents the eigenvalue of the i-th dimension in n dimensions;采用线性嵌入空间的形式:Take the form of a linear embedding space:φ(RF k)=W φRF k φ(RF k )=W φ RF kθ(F hw)=W θF hw θ(F hw )=W θ F hw其中W φ、W θ是学习权重矩阵;φ(RF k) i、θ(F hw) i分别表示两个特征向量在每个维度中的特征值; Where W φ , W θ are learning weight matrices; φ(RF k ) i , θ(F hw ) i respectively represent the eigenvalues of the two eigenvectors in each dimension;N(φ(RF))为归一化因子,通过计算最终预测网络中第h行、第w列的特征向量F hw分别与K个有效代表特征RF k的相似度的加和,将嵌入高斯相似度归一化为0到1的范围内,以避免相似度过高而产生的梯度***问题,归一化因子计算公式如式为; N(φ(RF)) is the normalization factor. By calculating the sum of the similarities between the feature vector F hw of the hth row and wth column in the final prediction network and K effective representative features RF k , the embedded Gaussian The similarity is normalized to a range of 0 to 1 to avoid the gradient explosion problem caused by excessive similarity. The formula for calculating the normalization factor is as follows:
- 根据权利要求3所述的密集目标检测方法,其特征在于,所述特征空间相似度计算过程包括如下步骤:The dense target detection method according to claim 3, wherein the feature space similarity calculation process comprises the following steps:(301)计算特征向量RepFeature k和 在特征图维度上的空间距离dis(RF k,F hw),计算公式为: (301) Calculate feature vector RepFeature k and The spatial distance dis(RF k ,F hw ) in the dimension of the feature map is calculated as:其中 为特征向量RepFeature k在特征图中的横、纵坐标, 为特征向量 在特征图中的横、纵坐标; in is the abscissa and ordinate of the feature vector RepFeature k in the feature map, is the eigenvector The horizontal and vertical coordinates in the feature map;(302)利用dis(RF k,F hw)乘以各个特征图的步长stride i得到两个特征向量在原始图像上的空间距离Corr Spatial_i(RF k,F hw),计算公式为: (302) Multiply dis(RF k , F hw ) by the stride i of each feature map to obtain the spatial distance Corr Spatial_i (RF k , F hw ) of the two feature vectors on the original image, the calculation formula is:其中,Spatial_i表示RF k、F hw取自特征金字塔网络自底向上的第i层特征图,α是尺度参数; Among them, Spatial_i means that RF k and F hw are taken from the i-th layer feature map from the bottom up of the feature pyramid network, and α is the scale parameter;所以,所述步骤3中的相似度表达式为:Therefore, the similarity expression in the step 3 is:Similarity(RF k,F hw) Similarity(RF k ,F hw )=Sim Embedded_Gaussian(RF k,F hw)+Corr Spatial_i(RF k,F hw) =Sim Embedded_Gaussian (RF k ,F hw )+Corr Spatial_i (RF k ,F hw )
- 根据权利要求4所述的密集目标检测方法,其特征在于,所述步骤4困难正样本最终的分类置信度通过将类别k的代表置信度按权重加到最终预测网络特征图的第(h,w)处位置关于类别k的置信度 上实现的,计算公式为: The dense target detection method according to claim 4, wherein the final classification confidence of the difficult positive sample in step 4 is added to the (h, The confidence of the position at w) about the category k Realized above, the calculation formula is:
- 根据权利要求3所述的密集目标检测方法,其特征在于,所述特征语义相似度的度量方法包括采用欧式相似度、余弦相似度或高斯相似度中的任一种。The dense target detection method according to claim 3, wherein the method for measuring feature semantic similarity includes any one of Euclidean similarity, cosine similarity or Gaussian similarity.
- 根据权利要求1所述的密集目标检测方法,其特征在于,所述特征提取网络采用卷积层缩小原始图像的尺寸,并将提取的有效特征输入到特征金字塔网络;所述特征提取网络选取ResNet或HRNet卷积神经网络。The dense target detection method according to claim 1, wherein the feature extraction network uses a convolutional layer to reduce the size of the original image, and the effective features extracted are input to the feature pyramid network; the feature extraction network selects ResNet or HRNet Convolutional Neural Network.
- 根据权利要求1所述的密集目标检测方法,其特征在于,所述初步预测网络选取S 2A-NET模型中的特征对齐模块。 The dense target detection method according to claim 1, wherein the feature alignment module in the S 2 A-NET model is selected for the preliminary prediction network.
- 根据权利要求1所述的密集目标检测方法,其特征在于,所述最终预测网络选取S 2A-NET模型中的旋转检测模块。 The dense target detection method according to claim 1, wherein the final prediction network selects the rotation detection module in the S 2 A-NET model.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110725564.1A CN113536986B (en) | 2021-06-29 | 2021-06-29 | Dense target detection method in remote sensing image based on representative features |
CN202110725564.1 | 2021-06-29 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023273337A1 true WO2023273337A1 (en) | 2023-01-05 |
Family
ID=78097103
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/074542 WO2023273337A1 (en) | 2021-06-29 | 2022-01-28 | Representative feature-based method for detecting dense targets in remote sensing image |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN113536986B (en) |
WO (1) | WO2023273337A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116935226A (en) * | 2023-08-01 | 2023-10-24 | 西安电子科技大学 | HRNet-based improved remote sensing image road extraction method, system, equipment and medium |
CN117746314A (en) * | 2023-11-20 | 2024-03-22 | 江苏星图智能科技有限公司 | Method, equipment and medium for judging OOD object based on multistage combination |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113536986B (en) * | 2021-06-29 | 2024-06-14 | 南京逸智网络空间技术创新研究院有限公司 | Dense target detection method in remote sensing image based on representative features |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200097771A1 (en) * | 2018-09-25 | 2020-03-26 | Nec Laboratories America, Inc. | Deep group disentangled embedding and network weight generation for visual inspection |
CN111259758A (en) * | 2020-01-13 | 2020-06-09 | 中国矿业大学 | Two-stage remote sensing image target detection method for dense area |
CN112818777A (en) * | 2021-01-21 | 2021-05-18 | 上海电力大学 | Remote sensing image target detection method based on dense connection and feature enhancement |
CN113536986A (en) * | 2021-06-29 | 2021-10-22 | 南京逸智网络空间技术创新研究院有限公司 | Representative feature-based dense target detection method in remote sensing image |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108829826B (en) * | 2018-06-14 | 2020-08-07 | 清华大学深圳研究生院 | Image retrieval method based on deep learning and semantic segmentation |
CN109145713B (en) * | 2018-07-02 | 2021-09-28 | 南京师范大学 | Small target semantic segmentation method combined with target detection |
CN109063594A (en) * | 2018-07-13 | 2018-12-21 | 吉林大学 | Remote sensing images fast target detection method based on YOLOv2 |
CN109961089B (en) * | 2019-02-26 | 2023-04-07 | 中山大学 | Small sample and zero sample image classification method based on metric learning and meta learning |
CN110298298B (en) * | 2019-06-26 | 2022-03-08 | 北京市商汤科技开发有限公司 | Target detection and target detection network training method, device and equipment |
CN110287927B (en) * | 2019-07-01 | 2021-07-27 | 西安电子科技大学 | Remote sensing image target detection method based on depth multi-scale and context learning |
CN110569879B (en) * | 2019-08-09 | 2024-03-15 | 平安科技(深圳)有限公司 | Tongue image extraction method, tongue image extraction device and computer readable storage medium |
CN111126205A (en) * | 2019-12-12 | 2020-05-08 | 南京邮电大学 | Optical remote sensing image airplane target detection method based on rotary positioning network |
CN111753677B (en) * | 2020-06-10 | 2023-10-31 | 杭州电子科技大学 | Multi-angle remote sensing ship image target detection method based on characteristic pyramid structure |
CN111723748B (en) * | 2020-06-22 | 2022-04-29 | 电子科技大学 | Infrared remote sensing image ship detection method |
CN112766361A (en) * | 2021-01-18 | 2021-05-07 | 山东师范大学 | Target fruit detection method and detection system under homochromatic background |
-
2021
- 2021-06-29 CN CN202110725564.1A patent/CN113536986B/en active Active
-
2022
- 2022-01-28 WO PCT/CN2022/074542 patent/WO2023273337A1/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200097771A1 (en) * | 2018-09-25 | 2020-03-26 | Nec Laboratories America, Inc. | Deep group disentangled embedding and network weight generation for visual inspection |
CN111259758A (en) * | 2020-01-13 | 2020-06-09 | 中国矿业大学 | Two-stage remote sensing image target detection method for dense area |
CN112818777A (en) * | 2021-01-21 | 2021-05-18 | 上海电力大学 | Remote sensing image target detection method based on dense connection and feature enhancement |
CN113536986A (en) * | 2021-06-29 | 2021-10-22 | 南京逸智网络空间技术创新研究院有限公司 | Representative feature-based dense target detection method in remote sensing image |
Non-Patent Citations (2)
Title |
---|
LI PENG, REN PENG, ZHANG XIAOYU, WANG QIAN, ZHU XIAOBIN, WANG LEI: "Region-Wise Deep Feature Representation for Remote Sensing Images", REMOTE SENSING, vol. 10, no. 6, 1 January 2018 (2018-01-01), pages 1 - 14, XP093020230, DOI: 10.3390/rs10060871 * |
XIN-YI TONG; GUI-SONG XIA; FAN HU; YANFEI ZHONG; MIHAI DATCU; LIANGPEI ZHANG: "Exploiting Deep Features for Remote Sensing Image Retrieval: A Systematic Investigation", ARXIV.ORG, 23 July 2017 (2017-07-23), pages 1 - 26, XP081560840 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116935226A (en) * | 2023-08-01 | 2023-10-24 | 西安电子科技大学 | HRNet-based improved remote sensing image road extraction method, system, equipment and medium |
CN117746314A (en) * | 2023-11-20 | 2024-03-22 | 江苏星图智能科技有限公司 | Method, equipment and medium for judging OOD object based on multistage combination |
Also Published As
Publication number | Publication date |
---|---|
CN113536986A (en) | 2021-10-22 |
CN113536986B (en) | 2024-06-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110110642B (en) | Pedestrian re-identification method based on multi-channel attention features | |
Xie et al. | Multilevel cloud detection in remote sensing images based on deep learning | |
WO2023273337A1 (en) | Representative feature-based method for detecting dense targets in remote sensing image | |
Li et al. | SAR image change detection using PCANet guided by saliency detection | |
CN110569738B (en) | Natural scene text detection method, equipment and medium based on densely connected network | |
Liu et al. | Bipartite differential neural network for unsupervised image change detection | |
Xia et al. | Loop closure detection for visual SLAM using PCANet features | |
Wang et al. | SSRNet: In-field counting wheat ears using multi-stage convolutional neural network | |
CN112085772B (en) | Remote sensing image registration method and device | |
CN106096658B (en) | Aerial Images classification method based on unsupervised deep space feature coding | |
CN107862680B (en) | Target tracking optimization method based on correlation filter | |
CN112580480B (en) | Hyperspectral remote sensing image classification method and device | |
CN112883850A (en) | Multi-view aerospace remote sensing image matching method based on convolutional neural network | |
CN109766752B (en) | Target matching and positioning method and system based on deep learning and computer | |
CN111967537A (en) | SAR target classification method based on two-way capsule network | |
Su et al. | FSRDD: An efficient few-shot detector for rare city road damage detection | |
CN112489089A (en) | Airborne ground moving target identification and tracking method for micro fixed wing unmanned aerial vehicle | |
CN112270285B (en) | SAR image change detection method based on sparse representation and capsule network | |
Liu et al. | A novel deep framework for change detection of multi-source heterogeneous images | |
Guo et al. | Multitemporal images change detection based on AMMF and spectral constraint strategy | |
Ge et al. | MMSRC: A multidirection multiscale spectral–spatial residual network for hyperspectral multiclass change detection | |
CN105844299B (en) | A kind of image classification method based on bag of words | |
CN104851090B (en) | Image change detection method and device | |
Ma et al. | Multi-spectral image change detection based on single-band iterative weighting and fuzzy C-means clustering | |
CN115272153A (en) | Image matching enhancement method based on feature sparse area detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 22831171 Country of ref document: EP Kind code of ref document: A1 |