CN108052940A - SAR remote sensing images waterborne target detection methods based on deep learning - Google Patents
SAR remote sensing images waterborne target detection methods based on deep learning Download PDFInfo
- Publication number
- CN108052940A CN108052940A CN201711358737.0A CN201711358737A CN108052940A CN 108052940 A CN108052940 A CN 108052940A CN 201711358737 A CN201711358737 A CN 201711358737A CN 108052940 A CN108052940 A CN 108052940A
- Authority
- CN
- China
- Prior art keywords
- network
- target detection
- model
- training
- sar image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 119
- 238000013135 deep learning Methods 0.000 title claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 49
- 238000013145 classification model Methods 0.000 claims abstract description 25
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 35
- 238000013527 convolutional neural network Methods 0.000 claims description 34
- 238000000034 method Methods 0.000 claims description 28
- 238000000605 extraction Methods 0.000 claims description 22
- 238000002372 labelling Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 6
- 238000013526 transfer learning Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 6
- 230000008901 benefit Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 16
- 238000005516 engineering process Methods 0.000 description 9
- 238000002474 experimental method Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000004913 activation Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 229910010293 ceramic material Inorganic materials 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013100 final test Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of SAR remote sensing images waterborne target detection methods based on deep learning, mainly solve the problems, such as that detection speed is slow in existing SAR image waterborne target detection method and locating accuracy is low.Specific implementation is：Design and one image classification model of pre-training first are then based on the modelling and training objective detection model, suggest network and Fast R CNN object detectors including RPN regions, the model finally obtained using training is detected SAR image waterborne target.The present invention has many advantages, such as that detection speed is fast, Detection accuracy is high, and the waterborne target available for large format SAR image detects.
Description
Technical Field
The invention belongs to the technical field of computer vision identification, and particularly relates to a SAR image water surface target detection method based on deep learning in the technical field of SAR image target detection.
Background
The synthetic aperture radar SAR has the characteristics of all weather, all time, high resolution, strong penetrating power and the like, and is widely applied to the fields of military reconnaissance and remote sensing.
In recent years, with the rapid development of satellite remote sensing technology, sensor technology, computer technology and communication technology, the application of space remote sensing technology enters a brand new development stage, and a large amount of data is provided for SAR image target identification and detection due to the appearance of a batch of high-resolution and short-access-period imaging satellites. The ship is used as a main transport carrier and a military target on the sea, and the automatic detection and identification technology of the ship has wide application prospect in the military and civil fields. In the near future, with the rapid development of space-based and space-based carrying platforms, earth observation satellites with higher resolution, better performance and shorter revisit time continuously appear in the future, and the detection of sea surface ship targets by applying a space remote sensing technology becomes a key research direction of an SAR image target identification and detection technology.
The conventional target detection method generally comprises three stages: firstly, area selection is carried out, namely, candidate areas are selected on a given image, then, the areas are subjected to artificial feature extraction, and finally, a trained classifier is used for classification. However, there are two major problems with this detection method: one is that the region selection uses a sliding window strategy to traverse the whole image, and different scales and length-width ratios need to be set, and the region selection strategy has no pertinence and high time complexity; another is that the target features are designed manually, which is not very robust to variations in feature diversity.
Convolutional Neural Networks (CNN) are one of the most popular deep Neural Networks today, and combine artificial Neural Networks and deep learning techniques, and have global training features combining local receptive fields, structural hierarchies, feature extraction and classification. The weight sharing network structure reduces the complexity of a network model, reduces the number of weights, enables an image to be directly used as the input of a network, and avoids the complex processes of feature extraction and data reconstruction in the traditional recognition algorithm. Therefore, the target detection technology based on the deep learning convolutional neural network is a popular research field in recent years, and a convolutional neural network target detection method based on regional suggestion (RegionProposal) and a convolutional neural network target detection method based on regression are also taken as main representatives.
CN103400156A discloses a high-resolution SAR ship detection method based on CFAR and sparse representation. On the basis of normal constant false alarm rate CFAR detection, feature vectors are extracted from slices, and identification is carried out through a sparse representation classifier, so that a final ship detection result is obtained. The method has the disadvantages that the detection process needs two steps of detection and identification, end-to-end detection cannot be realized, the detection speed is low, and the detection performance is poor in a complex scene.
Disclosure of Invention
The invention aims to provide a SAR remote sensing image water surface target detection method based on deep learning, which overcomes the problems of low SAR image target detection speed and low accuracy in the prior art, and realizes accurate end-to-end detection of the SAR image water surface target.
In order to achieve the purpose, the invention adopts the following technical scheme:
the SAR image water surface target detection method based on deep learning comprises the following steps:
s1, collecting SAR images and expanding a data set;
s2, labeling and labeling the SAR image data set to construct a training sample set, wherein the labeling refers to recording coordinates of an upper left corner point and a lower right corner point of the SAR image water surface target in a whole image, and the labels refer to category labels of the labeled water surface target;
s3, designing a convolutional neural network classification model C0For C, the method of "transfer learning" is adopted0Pre-training, designing an RPN region suggestion network model and a Fast R-CNN target detection network model based on the model;
s4, training the RPN region suggestion network and the Fast R-CNN target detection network by adopting a cross training method to obtain a final target detection model;
and S5, detecting the SAR image water surface target by using the target detection model.
Compared with the prior art, the invention has the following remarkable advantages: the method solves the problems of low detection speed and low positioning accuracy in the existing SAR image water surface target detection method, has the advantages of high detection speed and high detection accuracy, and can be used for water surface target detection of large-format SAR images.
Drawings
FIG. 1 is a flow chart of the present invention.
Fig. 2 is a diagram of a target detection network model architecture designed by the present invention.
Fig. 3 is an original SAR image.
Fig. 4 is a segmented SAR image.
Fig. 5 is a diagram of the detection result of the target detection model on the water surface target of the SAR image.
Detailed Description
The invention relates to a SAR image water surface target detection method based on deep learning, which comprises the following steps:
s1, collecting SAR images and expanding a data set;
s2, labeling and labeling the SAR image data set to construct a training sample set, wherein the labeling refers to recording coordinates of an upper left corner point and a lower right corner point of the SAR image water surface target in a whole image, and the labels refer to category labels of the labeled water surface target;
s3, designing a convolutional neural network classification model C0For C, the method of "transfer learning" is adopted0Pre-training, designing an RPN region suggestion network model and a Fast R-CNN target detection network model based on the model;
s4, training the RPN region suggestion network and the Fast R-CNN target detection network by adopting a cross training method to obtain a final target detection model;
and S5, detecting the SAR image water surface target by using the target detection model.
Acquiring the SAR image and expanding the data set as described in S1 includes the following sub-steps:
s11, downloading an open-source Sentinel SAR image from an European office official network (http:// www.esa.int), covering a world large port, and cutting each image into a plurality of sub-images with the long sides of 1000 and the short sides of 600 to serve as a training set;
and S12, adopting an incremental learning method, and expanding the data set by performing geometric transformation (including horizontal turning, translation and random scaling) on the SAR image and using a generative confrontation network.
The step of generating the augmented data set against the network as described in S12 is as follows:
s121, constructing a generative model G, wherein the model comprises two fully-connected layers and two convolutional layers, and adding Up Sampling operation in front of the first convolutional layer and the second convolutional layer from top to bottom;
s122, constructing a discriminant model D, wherein the model comprises two convolution layers and two full-connection layers, and Max Pooling operations are added behind the first convolution layer and the second convolution layer from top to bottom;
s123, respectively fixing one of G and D in the training process, updating the weight of the other until D cannot judge whether the input data is from a real image or an image generated by G, namely the judgment accuracy is stabilized at about 50%, terminating the training, and obtaining an SAR image generated by G;
and S124, adding the SAR image generated by the generated countermeasure network into a data set to achieve the purpose of expansion.
S3, classifying the convolutional neural network model C by adopting a transfer learning method0The pre-training steps are as follows:
C0is an 8-layer network junctionThe structure comprises 5 convolutional layers and 3 fully-connected layers, Max Pooling Pooling operations are added behind the first convolutional layer, the second convolutional layer and the fifth convolutional layer from top to bottom, an MSTAR data set is used as network input to train a model to obtain a pre-trained classification model, and the model is used for carrying out parameter initialization on an RPN region suggestion network and a Fast R-CNN target detection network.
Classification model C based on S30The designed RPN region proposed network is as follows:
first five convolutional layers and classification model C of RPN regional advice network0The first five convolutional layers are consistent, an RPN convolutional layer is connected behind the fifth convolutional layer, and the seventh convolutional layer and the eighth convolutional layer are directly connected behind the RPN convolutional layer, wherein the first five convolutional layers are feature extraction layers, the RPN convolutional layer is a feature mapping layer, the seventh convolutional layer outputs a regression frame containing the confidence coefficient of the water surface target, and the eighth convolutional layer outputs the position parameter of the regression frame.
Classification model C based on S30The Fast R-CNN target detection network is designed as follows:
five convolutional layers, an ROI posing layer, a first full-connection layer and a second full-connection layer of the Fast R-CNN target detection network are directly and sequentially cascaded, a third full-connection layer and a fourth full-connection layer are directly connected to the second full-connection layer, the first five convolutional layers of the Fast R-CNN target detection network and the first five convolutional layers of the RPN convolutional neural network share parameters, the first full-connection layer and the second full-connection layer of the Fast R-CNN target detection network perform nonlinear transformation on characteristics, the third full-connection layer of the Fast R-CNN target detection network outputs confidence degrees of discriminants, and the fourth full-connection layer of the Fast R-CNN target detection network outputs position correction parameters of a rough selection frame.
S4, the steps of obtaining the final target detection model by adopting the cross training method are as follows:
s41 classification model C using pre-training0Parameterizing candidate area extraction networksInitializing, and finely adjusting network parameters end to obtain a candidate region extraction model.
And S42, extracting a region suggestion frame from the training data set by using the candidate region extraction model obtained by training in the S41, using the region suggestion frame as the input of the target detection network, initializing parameters of the target detection network by using the classification model, and finely adjusting network parameters to obtain the target detector.
S43, initializing parameters of the candidate area extraction network by using the target detection network obtained in S42, fixing the parameters of the shared convolutional layer unchanged, and only finely adjusting the parameters of the unique convolutional layer in the candidate area extraction network, wherein at present, two networks share the convolutional layer.
And S44, extracting candidate regions from the data set by using the candidate region extraction network obtained by training in the S43, using the candidate regions as input on the target detection network, fixing the parameters of the shared convolutional layer, and finely adjusting the parameters of the full connection layer in the target detection network. Thus, the two networks share the same convolutional layer, forming a unified detection network.
S5, the steps of using the target detection model to detect the SAR image water surface target are as follows:
s51, inputting the SAR water surface target image test set into an RPN region suggestion network model, extracting a candidate region set P ', and inhibiting NMS (network management system) to remove repeated candidate regions according to a non-maximum value of the candidate region set P' to obtain a final candidate region set P;
s52, inputting the candidate region set P into a target detection network model, and outputting a category judgment probability pr for each candidate region in the candidate region set P;
s53, setting the threshold t to 0.8, and reserving the region with the probability value pr greater than 0.8, which is the final detection result.
The invention will be further explained with reference to the drawings.
Referring to fig. 1, the method for detecting the water surface target of the SAR image mainly includes three stages of model design, training and testing, and the final test is an end-to-end process, that is, the final target detection result can be directly obtained by inputting the SAR image, so the embodiment mainly explains the two stages of model design and training.
First, model design phase
As shown in fig. 2, the target detection model adopted by the present invention includes a classification model, an RPN region suggestion network, and a FastR-CNN target detection network, and the design steps are as follows:
step 1, designing a classification model C based on a convolutional neural network0。
Classification model C0The multilayer ceramic material mainly comprises five convolution layers and two full-connection layers, wherein the specific connection mode among the layers is as follows:
the first layer uses 96 convolution kernels with the size of 7 multiplied by 7, the sliding step length is 2, the activation function is ReLU, the layer is followed by a Max Pooling layer, the characteristic diagram output by the layer is subjected to down-sampling for dimensionality reduction, the kernel window size is 3 multiplied by 3, and the sliding step length is 2;
the second layer uses 256 convolution kernels with the size of 5 multiplied by 5, the sliding step length is 2, the activation function is ReLU, the second layer is followed by a Max Pooling layer, the characteristic diagram output by the second layer is subjected to down-sampling for dimensionality reduction, the kernel window size is 3 multiplied by 3, and the sliding step length is 2;
the third layer uses 384 convolution kernels with the size of 3 multiplied by 3, the sliding step length is 1, and the activation function is ReLU;
the fourth layer uses 384 convolution kernels with the size of 3 x 3, the sliding step size is 1, and the activation function is ReLU;
the fifth layer uses 256 convolution kernels with the size of 3 × 3, the sliding step is 1, and the activation function is ReLU;
the sixth layer and the seventh layer are all full connection layers and output one-dimensional vectors with the size of 4096.
Step 2,Design based on classification model C0The RPN region of (2) suggests a network.
First five convolutional layers and classification model C of RPN regional advice network0The first five convolutional layers are consistent;
connecting an RPN convolutional layer after the fifth convolutional layer, wherein the convolutional layer uses 256 convolutional kernels with the size of 3 multiplied by 3, and the sliding step length is 1;
two convolution layers, namely a frame regression layer and a classification layer, are connected in parallel after the RPN convolution layer, the sizes of the convolution cores are 1 multiplied by 1, and the sliding step length is 1.
Step 3, designing a classification-based model C0The Fast R-CNN target detection network.
First five convolutional layers of Fast R-CNN target detection network and classification model C0The first five convolutional layers are consistent;
connecting an ROI Pooling layer behind the fifth convolution layer, wherein the size of a kernel window is 6 multiplied by 6;
the ROI Pooling layer is connected with two full connection layers, and the two full connection layers output one-dimensional vectors with the size of 4096;
and connecting two full connection layers, namely a classification layer and a frame regression layer in parallel after the last full connection layer, wherein the classification layer is used for judging the category of the target, and the frame regression layer is used for accurately positioning the position of the target.
Second, model training phase
As the target detection model mainly comprises the RPN region suggestion network and the Fast R-CNN target detection network, the training of the model is mainly the training of the two networks. The loss function adopted by the method is a multi-task loss function and comprises a classification loss function and a frame regression loss function, and the training of the model mainly optimizes the loss function value to the minimum. The loss function is defined as:
wherein λ represents an equilibrium coefficient for controlling the specific gravity of the two loss functions; n is a radical ofclsRepresenting the number of input samples in the mini-batch; n is a radical ofregRepresents the number of anchors positions; p is a radical ofiRepresenting the probability that the anchor prediction is the target;indicating the groudtruth (gt) tag, if anchor is positive,is 1, otherwise is 0; t is ti＝{tx,ty,tw,thIs a vector representing the 4 parameterized coordinates of the predicted bounding box;a coordinate vector representing a GT bounding box corresponding to the positive anchor; l isclsRepresenting the classification loss function, LregRepresenting the bounding box regression loss function.
Classification loss function L in formulaclsIn a specific form of
Wherein,indicating that the ith reference region is predicted as a true tag yiThe probability of (c).
Border regression loss function L in formularegIn a specific form of
Wherein L is1(x) Representing a smooth 1-norm.
The two sub-networks are trained by a back propagation and random gradient descent method, because the two sub-networks are the shared convolutional layers, if the two sub-networks are independently trained, the additional calculation cost is increased, and the detection efficiency is reduced, therefore, the cross training method is adopted to train the model, so that the two sub-networks share the convolutional layers, and the method comprises the following steps:
a) classification model C Using Pre-training0And initializing parameters of the candidate region extraction network, and performing end-to-end fine adjustment on network parameters to obtain a candidate region extraction model.
b) Extracting a region suggestion frame from the training data set by using the candidate region extraction model obtained by training in the step a) as the input of the target detection network, performing parameter initialization on the target detection network by using the classification model, and finely adjusting network parameters to obtain the target detector.
c) Initializing the parameters of the candidate area extraction network by using the target detection network obtained in the step b), fixing the parameters of the shared convolution layer unchanged, and only finely adjusting the parameters of the unique convolution layer in the candidate area extraction network, wherein the two networks share the convolution layer.
d) Extracting candidate regions from the data set by using the candidate region extraction network obtained by training in the step c) as input on the target detection network, fixing the parameters of the shared convolutional layer, and finely adjusting the parameters of the full connection layer in the target detection network. Thus, the two networks share the same convolutional layer, forming a unified detection network.
The effect of the process of the invention can be further illustrated by the following experiments:
the SAR image has the characteristics of high resolution and small target, and the shooting angle of the image has randomness and diversity. In the experiment, the water surface ship targets in the SAR image are divided into a large ship and a small ship, and the large ship and the small ship are detected and identified.
1. Conditions of the experiment
The original data set used in the experiment of the present invention is Sentinel data of multiple ports downloaded from the official network of the european space, as shown in fig. 3, the size of the image is 26019 × 16706, the image contains ship targets of multiple attitudes and sizes, and contains multiple interference information, such as buildings, trees, lawns, etc. Since the original image size is too large, the target detection network cannot process such high resolution images, and therefore, each original SAR image is divided into a plurality of sub-images with the size of 1022 × 761 as shown in fig. 4 as training data, and more satisfactory SAR images are generated by using the generative countermeasure network as an extension of the training set. The purpose of the experiment is to detect all water surface targets in the SAR image.
The experimental environment of the invention is as follows: the operating system is Ubuntu 16.04LTS 64 bit version, the processor is Intel Xeon (to strong) E5-1620 v2, the video card is Nvidia Quadro K2200/4GB, and the software platform is Caffe.
2. Analysis of experimental content and results
The experiment of the invention is to divide the water surface target in the SAR image into a large ship and a small ship for detection, firstly, a target detection model is trained by utilizing a manufactured SAR image training set to obtain a trained model.
Then, inputting the test sample into a trained target detection model, setting the probability threshold value to be 0.8 for screening detection, and quantifying the detection result by the statistical accuracy after the detection is finished, wherein,
t is the number of the targets which are correctly detected and correctly classified, and NP is the number of the targets.
The data set used in the experiment was 712 ship targets, 395 for the marked large ship targets and 317 for the small ship targets. Wherein, the test set has 388 ship targets, 206 large ship targets and 182 small ship targets. The detection results of the trained model on the large and small ship targets on the test set are shown in table 1.
TABLE 1
As can be seen from table 1, the model obtained by training after the classification of the ship targets can correctly detect 187 targets in 206 ships, the detection rate is 90.8%, 166 targets in 182 ships can be correctly detected, the accuracy is 91.2%, the overall average detection accuracy is 91.0%, it needs to be pointed out that the project uses a target detection network based on deep learning, the detection time of a single picture reaches 0.14s, the detection speed is hundreds of times faster than that of the traditional target detection algorithm, and the advantage of fast deep learning speed is embodied.
Fig. 5 is a diagram of a detection result of a water surface target of the SAR image by using a target detection model.
Claims (8)
1. The SAR image water surface target detection method based on deep learning is characterized by comprising the following steps:
step 1: acquiring an SAR image and expanding a data set;
step 2: labeling and labeling an SAR image data set, and constructing a training sample set;
and step 3: designing a convolutional neural network classification model C0For C, the method of "transfer learning" is adopted0Pre-training, designing an RPN region suggestion network model and a Fast R-CNN target detection network model based on the model;
and 4, step 4: training the RPN region proposed network and the Fast R-CNN target detection network by adopting a cross training method to obtain a final target detection model;
and 5: and detecting the water surface target of the SAR image by using a target detection model.
2. The SAR image water surface target detection method according to claim 1, characterized in that: step 1, acquiring an SAR image, and expanding a data set comprises:
collecting SAR images of all ports in the world, and cutting each image into a plurality of subimages with the long sides of 1000 and the short sides of 600 to serve as a training set;
the method adopts an incremental learning method, and expands the data set by performing geometric transformation on the SAR image and using a generative countermeasure network, wherein the geometric transformation comprises horizontal turning, translation and random scaling.
3. The SAR image water surface target detection method according to claim 1, characterized in that: in the step 2, the labeling refers to recording coordinates of an upper left corner point and a lower right corner point of the water surface target of the SAR image in the whole image, and the label refers to a category mark of the labeled water surface target.
4. The SAR image water surface target detection method according to claim 1, characterized in that: step 3 adopts a transfer learning method, uses MSTAR data set to classify the convolutional neural network classification model C0And pre-training to obtain a trained classification model, and designing an RPN region suggestion network model and a Fast R-CNN target detection network model based on the model.
5. The SAR image water surface target detection method according to claim 4, characterized in that: classification model C for convolutional neural network using MSTAR dataset0The pre-training steps are as follows:
C0in the form of an 8-tier network structure,the method comprises the steps of adding Max Pooling Pooling operation on the first layer, the second layer and the fifth layer from top to bottom, training a model by using an MSTAR data set as network input to obtain a pre-trained classification model, and performing parameter initialization on an RPN region suggestion network and a Fast R-CNN target detection network by using the model.
6. The SAR image water surface target detection method according to claim 4, characterized in that: the classification-based model C0The designed RPN region proposed network is as follows:
first five convolutional layers and classification model C of RPN regional advice network0The first five convolutional layers of (a) coincide, one RPN convolutional layer is connected after the fifth convolutional layer, and the seventh and eighth convolutional layers are directly connected after the RPN convolutional layer.
7. The SAR image water surface target detection method according to claim 4, characterized in that: the classification-based model C0The Fast R-CNN target detection network is designed as follows:
five convolutional layers, an ROI posing layer, a first full-connection layer and a second full-connection layer of the Fast R-CNN target detection network are directly and sequentially cascaded, and a third full-connection layer and a fourth full-connection layer are directly connected to the second full-connection layer.
8. The SAR image water surface target detection method according to claim 1, characterized in that: the specific method for obtaining the final target detection model by adopting the cross training method in the step 4 is as follows:
a) classification model C Using Pre-training0Initializing parameters of the candidate region extraction network, and performing end-to-end fine adjustment on network parameters to obtain a candidate region extraction model;
b) extracting a region suggestion frame from the training data set by using the candidate region extraction model obtained by training in the step a) as the input of a target detection network, performing parameter initialization on the target detection network by using a classification model, and finely adjusting network parameters to obtain a target detector;
c) initializing parameters of the candidate area extraction network by using the target detection network obtained in the step b), fixing the shared convolutional layer parameters unchanged, and only finely adjusting the unique convolutional layer parameters in the candidate area extraction network;
d) extracting candidate regions from the data set by using the candidate region extraction network obtained by training in the step c), using the candidate regions as input on the target detection network, fixing the parameters of the shared convolutional layer, and finely adjusting the parameters of the full connection layer in the target detection network; thus, the two networks share the same convolution layer to form a unified detection network;
and (3) taking the test image as the input of a target detection model, and marking the water surface ship target detected in the image by using a rectangular frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711358737.0A CN108052940A (en) | 2017-12-17 | 2017-12-17 | SAR remote sensing images waterborne target detection methods based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711358737.0A CN108052940A (en) | 2017-12-17 | 2017-12-17 | SAR remote sensing images waterborne target detection methods based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108052940A true CN108052940A (en) | 2018-05-18 |
Family
ID=62133414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711358737.0A Pending CN108052940A (en) | 2017-12-17 | 2017-12-17 | SAR remote sensing images waterborne target detection methods based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108052940A (en) |
Cited By (47)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830213A (en) * | 2018-06-12 | 2018-11-16 | 北京理工大学 | Car plate detection and recognition methods and device based on deep learning |
| CN108830209A (en) * | 2018-06-08 | 2018-11-16 | 西安电子科技大学 | Based on the remote sensing images method for extracting roads for generating confrontation network |
| CN108898066A (en) * | 2018-06-06 | 2018-11-27 | 天津大学 | A kind of body movement detection method based on production confrontation network |
| CN108960143A (en) * | 2018-07-04 | 2018-12-07 | 北京航空航天大学 | Detect deep learning method in a kind of naval vessel in High Resolution Visible Light remote sensing images |
| CN109190750A (en) * | 2018-07-06 | 2019-01-11 | 国家计算机网络与信息安全管理中心 | The small sample generation method and device of network are generated based on confrontation |
| CN109255390A (en) * | 2018-09-30 | 2019-01-22 | 京东方科技集团股份有限公司 | Preprocess method and module, discriminator, the readable storage medium storing program for executing of training image |
| CN109299688A (en) * | 2018-09-19 | 2019-02-01 | 厦门大学 | Ship Detection based on deformable fast convolution neural network |
| CN109325527A (en) * | 2018-09-05 | 2019-02-12 | 哈尔滨工程大学 | A kind of sidescan-sonar image target automatic classification method based on transfer learning and deep learning |
| CN109341689A (en) * | 2018-09-12 | 2019-02-15 | 北京工业大学 | Vision navigation method of mobile robot based on deep learning |
| CN109544518A (en) * | 2018-11-07 | 2019-03-29 | 中国科学院深圳先进技术研究院 | A kind of method and its system applied to the assessment of skeletal maturation degree |
| CN109583425A (en) * | 2018-12-21 | 2019-04-05 | 西安电子科技大学 | A kind of integrated recognition methods of the remote sensing images ship based on deep learning |
| CN109583424A (en) * | 2018-12-21 | 2019-04-05 | 杭州电子科技大学 | A kind of overlapping Ship Target candidate frame screening technique based on decaying confidence level |
| CN109583369A (en) * | 2018-11-29 | 2019-04-05 | 北京邮电大学 | A kind of target identification method and device based on target area segmentation network |
| CN109858481A (en) * | 2019-01-09 | 2019-06-07 | 杭州电子科技大学 | A kind of Ship Target Detection method based on the detection of cascade position sensitivity |
| CN109859103A (en) * | 2019-01-09 | 2019-06-07 | 杭州电子科技大学 | A kind of Ship Target Detection method based on the accurate pondization strategy of bilinear interpolation method |
| CN109886073A (en) * | 2018-12-26 | 2019-06-14 | 深圳云天励飞技术有限公司 | A kind of image detecting method and device |
| CN109934095A (en) * | 2019-01-22 | 2019-06-25 | 中交信息技术国家工程实验室有限公司 | A kind of remote sensing images Clean water withdraw method and system based on deep learning |
| CN109934122A (en) * | 2019-02-21 | 2019-06-25 | 北京以萨技术股份有限公司 | A kind of remote sensing image ship detecting method based on deep learning |
| CN109934088A (en) * | 2019-01-10 | 2019-06-25 | 海南大学 | Sea ship discrimination method based on deep learning |
| CN109961105A (en) * | 2019-04-08 | 2019-07-02 | 上海市测绘院 | A kind of Classification of High Resolution Satellite Images method based on multitask deep learning |
| CN110070124A (en) * | 2019-04-15 | 2019-07-30 | 广州小鹏汽车科技有限公司 | A kind of image amplification method and system based on production confrontation network |
| CN110135522A (en) * | 2019-05-28 | 2019-08-16 | 金陵科技学院 | It is a kind of to detect and the mark integrated intelligent method of remote sensing images Small object |
| CN110210463A (en) * | 2019-07-03 | 2019-09-06 | 中国人民解放军海军航空大学 | Radar target image detecting method based on Precise ROI-Faster R-CNN |
| CN110222632A (en) * | 2019-06-04 | 2019-09-10 | 哈尔滨工程大学 | A kind of waterborne target detection method of gray prediction auxiliary area suggestion |
| CN110232687A (en) * | 2019-06-21 | 2019-09-13 | 华北电力大学（保定） | The detection method of cotter bolt defect in a kind of electric inspection process image |
| CN110427981A (en) * | 2019-07-11 | 2019-11-08 | 四川大学 | SAR ship detecting system and method based on deep neural network |
| CN110458043A (en) * | 2019-07-20 | 2019-11-15 | 中国船舶重工集团公司第七二四研究所 | A kind of SAR target identification method based on transfer learning and the output of full articulamentum |
| CN110502654A (en) * | 2019-08-26 | 2019-11-26 | 长光卫星技术有限公司 | A kind of object library generation system suitable for multi-source heterogeneous remotely-sensed data |
| CN110837856A (en) * | 2019-10-31 | 2020-02-25 | 深圳市商汤科技有限公司 | Neural network training and target detection method, device, equipment and storage medium |
| CN110969213A (en) * | 2019-12-10 | 2020-04-07 | 珠海大横琴科技发展有限公司 | Ship detection method and device based on fast RCNN and electronic equipment |
| CN111008603A (en) * | 2019-12-08 | 2020-04-14 | 中南大学 | Multi-class target rapid detection method for large-scale remote sensing image |
| CN111079642A (en) * | 2019-12-13 | 2020-04-28 | 国网浙江余姚市供电有限公司 | Line movable monitoring method and device and computer readable medium |
| CN111144208A (en) * | 2019-11-22 | 2020-05-12 | 北京航天控制仪器研究所 | Automatic detection and identification method for marine vessel target and target detector |
| CN111339864A (en) * | 2020-02-17 | 2020-06-26 | 北京百度网讯科技有限公司 | Abnormal behavior alarm method and device |
| CN111553212A (en) * | 2020-04-16 | 2020-08-18 | 中国科学院深圳先进技术研究院 | Remote sensing image target detection method based on smooth frame regression function |
| CN111611834A (en) * | 2019-12-23 | 2020-09-01 | 珠海大横琴科技发展有限公司 | Ship identification method and device based on SAR |
| CN111950465A (en) * | 2020-08-13 | 2020-11-17 | 金陵科技学院 | SAE-CNN-based SAR image flood detection and prevention method |
| CN111983619A (en) * | 2020-08-07 | 2020-11-24 | 西北工业大学 | Underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning |
| CN112114303A (en) * | 2020-08-17 | 2020-12-22 | 安徽捷纳森电子科技有限公司 | Method for identifying sand stealer through unidirectional passive array control scanning |
| CN112132042A (en) * | 2020-09-24 | 2020-12-25 | 西安电子科技大学 | SAR image target detection method based on anti-domain adaptation |
| CN112257627A (en) * | 2020-10-29 | 2021-01-22 | 上海交通大学 | Overwater image data set expansion method |
| CN112329559A (en) * | 2020-10-22 | 2021-02-05 | 空间信息产业发展股份有限公司 | Method for detecting homestead target based on deep convolutional neural network |
| CN113065446A (en) * | 2021-03-29 | 2021-07-02 | 青岛东坤蔚华数智能源科技有限公司 | Depth inspection method for automatically identifying ship corrosion area |
| CN113240015A (en) * | 2021-05-18 | 2021-08-10 | 哈尔滨工业大学（深圳） | Image matching method combined with deep learning |
| CN113569097A (en) * | 2021-07-23 | 2021-10-29 | 北京百度网讯科技有限公司 | Structured information extraction method, device, equipment and storage medium |
| US11348005B2 (en) | 2018-09-30 | 2022-05-31 | Boe Technology Group Co., Ltd. | Apparatus, method, and computer-readable medium for image processing, and system for training a neural network |
| EP4022503A4 (en) * | 2020-04-15 | 2022-09-21 | Aselsan Elektronik Sanayi ve Ticaret Anonim Sirketi | A method for training shallow convolutional neural networks for infrared target detection using a two-phase learning strategy |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001004398A (en) * | 1999-06-25 | 2001-01-12 | Mitsubishi Space Software Kk | Movement information detection method based on satellite sar image |
| CN106156744A (en) * | 2016-07-11 | 2016-11-23 | 西安电子科技大学 | SAR target detection method based on CFAR detection with degree of depth study |
| CN106203506A (en) * | 2016-07-11 | 2016-12-07 | 上海凌科智能科技有限公司 | A kind of pedestrian detection method based on degree of depth learning art |
| CN106228124A (en) * | 2016-07-17 | 2016-12-14 | 西安电子科技大学 | SAR image object detection method based on convolutional neural networks |
| CN106980895A (en) * | 2017-02-22 | 2017-07-25 | 中国科学院自动化研究所 | Convolutional neural networks Forecasting Methodology based on rotary area |
| CN107452023A (en) * | 2017-07-21 | 2017-12-08 | 上海交通大学 | A kind of monotrack method and system based on convolutional neural networks on-line study |
-
2017
- 2017-12-17 CN CN201711358737.0A patent/CN108052940A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001004398A (en) * | 1999-06-25 | 2001-01-12 | Mitsubishi Space Software Kk | Movement information detection method based on satellite sar image |
| CN106156744A (en) * | 2016-07-11 | 2016-11-23 | 西安电子科技大学 | SAR target detection method based on CFAR detection with degree of depth study |
| CN106203506A (en) * | 2016-07-11 | 2016-12-07 | 上海凌科智能科技有限公司 | A kind of pedestrian detection method based on degree of depth learning art |
| CN106228124A (en) * | 2016-07-17 | 2016-12-14 | 西安电子科技大学 | SAR image object detection method based on convolutional neural networks |
| CN106980895A (en) * | 2017-02-22 | 2017-07-25 | 中国科学院自动化研究所 | Convolutional neural networks Forecasting Methodology based on rotary area |
| CN107452023A (en) * | 2017-07-21 | 2017-12-08 | 上海交通大学 | A kind of monotrack method and system based on convolutional neural networks on-line study |
Non-Patent Citations (5)
| Title |
|---|
| JIANWEI LI ET AL: "Ship detection in SAR images based on an improved faster R-CNN", 《2017 SAR IN BIG DATA ERA: MODELS, METHODS AND APPLICATIONS (BIGSARDATA)》 * |
| MIAO KANG ET AL: "A Modified Faster R-CNN Based on CFAR Algorithm for SAR Ship Detection", 《2017 INTERNATIONAL WORKSHOP ON REMOTE SENSING WITH INTELLIGENT PROCESSING (RSIP)》 * |
| 袁秋壮等: "基于深度学习神经网络的SAR星上目标识别系统研究", 《上海航天》 * |
| 郑儒楠等: "应用虚拟样本对SAR图像目标识别的研究", 《机械与电子》 * |
| 黄孝平: "《当代机器深度学习方法与应用研究》", 30 November 2017 * |
Cited By (67)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108898066A (en) * | 2018-06-06 | 2018-11-27 | 天津大学 | A kind of body movement detection method based on production confrontation network |
| CN108898066B (en) * | 2018-06-06 | 2022-01-04 | 天津大学 | Human motion detection method based on generating type countermeasure network |
| CN108830209A (en) * | 2018-06-08 | 2018-11-16 | 西安电子科技大学 | Based on the remote sensing images method for extracting roads for generating confrontation network |
| CN108830209B (en) * | 2018-06-08 | 2021-12-17 | 西安电子科技大学 | Remote sensing image road extraction method based on generation countermeasure network |
| CN108830213A (en) * | 2018-06-12 | 2018-11-16 | 北京理工大学 | Car plate detection and recognition methods and device based on deep learning |
| CN108960143A (en) * | 2018-07-04 | 2018-12-07 | 北京航空航天大学 | Detect deep learning method in a kind of naval vessel in High Resolution Visible Light remote sensing images |
| CN109190750A (en) * | 2018-07-06 | 2019-01-11 | 国家计算机网络与信息安全管理中心 | The small sample generation method and device of network are generated based on confrontation |
| CN109190750B (en) * | 2018-07-06 | 2021-06-08 | 国家计算机网络与信息安全管理中心 | Small sample generation method and device based on countermeasure generation network |
| CN109325527A (en) * | 2018-09-05 | 2019-02-12 | 哈尔滨工程大学 | A kind of sidescan-sonar image target automatic classification method based on transfer learning and deep learning |
| CN109341689A (en) * | 2018-09-12 | 2019-02-15 | 北京工业大学 | Vision navigation method of mobile robot based on deep learning |
| CN109299688A (en) * | 2018-09-19 | 2019-02-01 | 厦门大学 | Ship Detection based on deformable fast convolution neural network |
| CN109299688B (en) * | 2018-09-19 | 2021-10-01 | 厦门大学 | Ship detection method based on deformable fast convolution neural network |
| US11449751B2 (en) | 2018-09-30 | 2022-09-20 | Boe Technology Group Co., Ltd. | Training method for generative adversarial network, image processing method, device and storage medium |
| US11348005B2 (en) | 2018-09-30 | 2022-05-31 | Boe Technology Group Co., Ltd. | Apparatus, method, and computer-readable medium for image processing, and system for training a neural network |
| US11615505B2 (en) | 2018-09-30 | 2023-03-28 | Boe Technology Group Co., Ltd. | Apparatus and method for image processing, and system for training neural network |
| US11361222B2 (en) | 2018-09-30 | 2022-06-14 | Boe Technology Group Co., Ltd. | System, method, and computer-readable medium for image classification |
| CN109255390A (en) * | 2018-09-30 | 2019-01-22 | 京东方科技集团股份有限公司 | Preprocess method and module, discriminator, the readable storage medium storing program for executing of training image |
| CN109255390B (en) * | 2018-09-30 | 2021-01-29 | 京东方科技集团股份有限公司 | Training image preprocessing method and module, discriminator and readable storage medium |
| CN109544518A (en) * | 2018-11-07 | 2019-03-29 | 中国科学院深圳先进技术研究院 | A kind of method and its system applied to the assessment of skeletal maturation degree |
| CN109544518B (en) * | 2018-11-07 | 2020-11-03 | 中国科学院深圳先进技术研究院 | Method and system applied to bone maturity assessment |
| CN109583369B (en) * | 2018-11-29 | 2020-11-13 | 北京邮电大学 | Target identification method and device based on target area segmentation network |
| CN109583369A (en) * | 2018-11-29 | 2019-04-05 | 北京邮电大学 | A kind of target identification method and device based on target area segmentation network |
| CN109583424A (en) * | 2018-12-21 | 2019-04-05 | 杭州电子科技大学 | A kind of overlapping Ship Target candidate frame screening technique based on decaying confidence level |
| CN109583425A (en) * | 2018-12-21 | 2019-04-05 | 西安电子科技大学 | A kind of integrated recognition methods of the remote sensing images ship based on deep learning |
| CN109583425B (en) * | 2018-12-21 | 2023-05-02 | 西安电子科技大学 | Remote sensing image ship integrated recognition method based on deep learning |
| CN109886073A (en) * | 2018-12-26 | 2019-06-14 | 深圳云天励飞技术有限公司 | A kind of image detecting method and device |
| CN109859103A (en) * | 2019-01-09 | 2019-06-07 | 杭州电子科技大学 | A kind of Ship Target Detection method based on the accurate pondization strategy of bilinear interpolation method |
| CN109858481A (en) * | 2019-01-09 | 2019-06-07 | 杭州电子科技大学 | A kind of Ship Target Detection method based on the detection of cascade position sensitivity |
| CN109934088A (en) * | 2019-01-10 | 2019-06-25 | 海南大学 | Sea ship discrimination method based on deep learning |
| CN109934095A (en) * | 2019-01-22 | 2019-06-25 | 中交信息技术国家工程实验室有限公司 | A kind of remote sensing images Clean water withdraw method and system based on deep learning |
| CN109934122A (en) * | 2019-02-21 | 2019-06-25 | 北京以萨技术股份有限公司 | A kind of remote sensing image ship detecting method based on deep learning |
| CN109961105B (en) * | 2019-04-08 | 2020-03-27 | 上海市测绘院 | High-resolution remote sensing image classification method based on multitask deep learning |
| CN109961105A (en) * | 2019-04-08 | 2019-07-02 | 上海市测绘院 | A kind of Classification of High Resolution Satellite Images method based on multitask deep learning |
| CN110070124A (en) * | 2019-04-15 | 2019-07-30 | 广州小鹏汽车科技有限公司 | A kind of image amplification method and system based on production confrontation network |
| CN110135522B (en) * | 2019-05-28 | 2023-03-28 | 金陵科技学院 | Intelligent method for detecting and marking small target integration of remote sensing image |
| CN110135522A (en) * | 2019-05-28 | 2019-08-16 | 金陵科技学院 | It is a kind of to detect and the mark integrated intelligent method of remote sensing images Small object |
| CN110222632A (en) * | 2019-06-04 | 2019-09-10 | 哈尔滨工程大学 | A kind of waterborne target detection method of gray prediction auxiliary area suggestion |
| CN110232687B (en) * | 2019-06-21 | 2022-04-12 | 华北电力大学（保定） | Method for detecting defects of bolts with pins in power inspection image |
| CN110232687A (en) * | 2019-06-21 | 2019-09-13 | 华北电力大学（保定） | The detection method of cotter bolt defect in a kind of electric inspection process image |
| CN110210463A (en) * | 2019-07-03 | 2019-09-06 | 中国人民解放军海军航空大学 | Radar target image detecting method based on Precise ROI-Faster R-CNN |
| CN110427981B (en) * | 2019-07-11 | 2023-01-31 | 四川大学 | SAR ship detection system and method based on deep neural network |
| CN110427981A (en) * | 2019-07-11 | 2019-11-08 | 四川大学 | SAR ship detecting system and method based on deep neural network |
| CN110458043A (en) * | 2019-07-20 | 2019-11-15 | 中国船舶重工集团公司第七二四研究所 | A kind of SAR target identification method based on transfer learning and the output of full articulamentum |
| CN110502654A (en) * | 2019-08-26 | 2019-11-26 | 长光卫星技术有限公司 | A kind of object library generation system suitable for multi-source heterogeneous remotely-sensed data |
| CN110837856A (en) * | 2019-10-31 | 2020-02-25 | 深圳市商汤科技有限公司 | Neural network training and target detection method, device, equipment and storage medium |
| CN111144208A (en) * | 2019-11-22 | 2020-05-12 | 北京航天控制仪器研究所 | Automatic detection and identification method for marine vessel target and target detector |
| CN111008603A (en) * | 2019-12-08 | 2020-04-14 | 中南大学 | Multi-class target rapid detection method for large-scale remote sensing image |
| CN111008603B (en) * | 2019-12-08 | 2023-04-18 | 中南大学 | Multi-class target rapid detection method for large-scale remote sensing image |
| CN110969213A (en) * | 2019-12-10 | 2020-04-07 | 珠海大横琴科技发展有限公司 | Ship detection method and device based on fast RCNN and electronic equipment |
| CN111079642A (en) * | 2019-12-13 | 2020-04-28 | 国网浙江余姚市供电有限公司 | Line movable monitoring method and device and computer readable medium |
| CN111079642B (en) * | 2019-12-13 | 2023-11-14 | 国网浙江余姚市供电有限公司 | Line movable monitoring method and device and computer readable medium |
| CN111611834A (en) * | 2019-12-23 | 2020-09-01 | 珠海大横琴科技发展有限公司 | Ship identification method and device based on SAR |
| CN111339864B (en) * | 2020-02-17 | 2023-11-10 | 北京百度网讯科技有限公司 | Abnormal behavior alarm method and device |
| CN111339864A (en) * | 2020-02-17 | 2020-06-26 | 北京百度网讯科技有限公司 | Abnormal behavior alarm method and device |
| EP4022503A4 (en) * | 2020-04-15 | 2022-09-21 | Aselsan Elektronik Sanayi ve Ticaret Anonim Sirketi | A method for training shallow convolutional neural networks for infrared target detection using a two-phase learning strategy |
| WO2021208502A1 (en) * | 2020-04-16 | 2021-10-21 | 中国科学院深圳先进技术研究院 | Remote-sensing image target detection method based on smooth bounding box regression function |
| CN111553212A (en) * | 2020-04-16 | 2020-08-18 | 中国科学院深圳先进技术研究院 | Remote sensing image target detection method based on smooth frame regression function |
| CN111983619A (en) * | 2020-08-07 | 2020-11-24 | 西北工业大学 | Underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning |
| CN111950465A (en) * | 2020-08-13 | 2020-11-17 | 金陵科技学院 | SAE-CNN-based SAR image flood detection and prevention method |
| CN112114303A (en) * | 2020-08-17 | 2020-12-22 | 安徽捷纳森电子科技有限公司 | Method for identifying sand stealer through unidirectional passive array control scanning |
| CN112132042B (en) * | 2020-09-24 | 2023-08-01 | 西安电子科技大学 | SAR image target detection method based on contrast domain adaptation |
| CN112132042A (en) * | 2020-09-24 | 2020-12-25 | 西安电子科技大学 | SAR image target detection method based on anti-domain adaptation |
| CN112329559A (en) * | 2020-10-22 | 2021-02-05 | 空间信息产业发展股份有限公司 | Method for detecting homestead target based on deep convolutional neural network |
| CN112257627A (en) * | 2020-10-29 | 2021-01-22 | 上海交通大学 | Overwater image data set expansion method |
| CN113065446A (en) * | 2021-03-29 | 2021-07-02 | 青岛东坤蔚华数智能源科技有限公司 | Depth inspection method for automatically identifying ship corrosion area |
| CN113240015A (en) * | 2021-05-18 | 2021-08-10 | 哈尔滨工业大学（深圳） | Image matching method combined with deep learning |
| CN113569097A (en) * | 2021-07-23 | 2021-10-29 | 北京百度网讯科技有限公司 | Structured information extraction method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108052940A (en) | SAR remote sensing images waterborne target detection methods based on deep learning | |
| CN111507271B (en) | Airborne photoelectric video target intelligent detection and identification method | |
| Wang et al. | SAR target detection based on SSD with data augmentation and transfer learning | |
| US10607362B2 (en) | Remote determination of containers in geographical region | |
| CN108460382B (en) | Optical remote sensing image ship detection method based on deep learning single-step detector | |
| CN109598241B (en) | Satellite image marine ship identification method based on Faster R-CNN | |
| CN111738112B (en) | Remote sensing ship image target detection method based on deep neural network and self-attention mechanism | |
| CN108460341B (en) | Optical remote sensing image target detection method based on integrated depth convolution network | |
| White | Change detection in SAR imagery | |
| CN110889324A (en) | Thermal infrared image target identification method based on YOLO V3 terminal-oriented guidance | |
| CN108898065B (en) | Deep network ship target detection method with candidate area rapid screening and scale self-adaption | |
| CN108596108B (en) | Aerial remote sensing image change detection method based on triple semantic relation learning | |
| US10755146B2 (en) | Network architecture for generating a labeled overhead image | |
| CN111476159A (en) | Method and device for training and detecting detection model based on double-angle regression | |
| Zhou et al. | Multi-resolution networks for ship detection in infrared remote sensing images | |
| Li et al. | Gaofen-3 sea ice detection based on deep learning | |
| Chini et al. | Comparing statistical and neural network methods applied to very high resolution satellite images showing changes in man-made structures at rocky flats | |
| Onishi et al. | Deep convolutional neural network for cloud coverage estimation from snapshot camera images | |
| EP3553700A2 (en) | Remote determination of containers in geographical region | |
| CN111553280A (en) | Target part identification method based on deep learning | |
| CN114565824B (en) | Single-stage rotating ship detection method based on full convolution network | |
| CN106709515A (en) | Downward-looking scene matching area selection criteria intervention method | |
| Yang et al. | Sar images target detection based on yolov5 | |
| CN114998749B (en) | SAR data amplification method for target detection | |
| Bupphawat et al. | Super-resolution land cover mapping based on deep learning and level set method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180518 |