CN112560907A - Limited pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention - Google Patents
Limited pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention Download PDFInfo
- Publication number
- CN112560907A CN112560907A CN202011392343.9A CN202011392343A CN112560907A CN 112560907 A CN112560907 A CN 112560907A CN 202011392343 A CN202011392343 A CN 202011392343A CN 112560907 A CN112560907 A CN 112560907A
- Authority
- CN
- China
- Prior art keywords
- feature
- layer
- feature map
- attention
- size
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 39
- 238000010586 diagram Methods 0.000 claims abstract description 28
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims abstract description 18
- 230000004927 fusion Effects 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 230000007246 mechanism Effects 0.000 claims abstract description 10
- 238000005070 sampling Methods 0.000 claims abstract description 5
- 238000011176 pooling Methods 0.000 claims description 24
- 230000004913 activation Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
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)
- 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)
- Image Analysis (AREA)
- Aiming, Guidance, Guns With A Light Source, Armor, Camouflage, And Targets (AREA)
Abstract
The invention discloses a finite pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention, which is characterized in that an infrared image is subjected to characteristic extraction through a single-step cascade neural network based on an attention mechanism to obtain a characteristic diagram; and performing up-sampling on the characteristic diagram to obtain a multi-scale characteristic diagram, and performing characteristic fusion and prediction on the multi-scale characteristic diagram respectively to obtain a final target detection result. And an attention mechanism and a multi-scale feature fusion prediction method are adopted, so that the feature extraction capability and the detection accuracy of the limited pixel target are improved.
Description
Technical Field
The invention belongs to the field of infrared small target detection, and particularly relates to a limited pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention.
Background
Along with the continuous development of unmanned platform technology, the development and application of unmanned aerial vehicles in various countries are also active, and the unmanned aerial vehicle is widely applied to the military and civil fields by virtue of the advantages of high efficiency, low cost, high output, strong maneuverability and the like. Meanwhile, the method also brings a serious challenge to the safety guarantee of the airspace and even causes great social harm. The infrared imaging detection technology can be used for detecting and tracking the unmanned aerial vehicle, and the unmanned aerial vehicle can be effectively monitored. However, in an actual scene, due to the influence caused by long imaging distance and long-distance atmospheric radiation interference, the target signal-to-noise ratio is low, the number of pixel points is small, the shape texture is absent, and the interference of complex background noise and random noise can be received, so that the detection precision and the detection efficiency can not be considered simultaneously by a conventional target detection and identification algorithm.
To solve the problem of finite pixel target detection, there are two main methods at present: single frame detection and multiple frame detection. Since multi-frame detection algorithms generally take more time than single-frame detection algorithms, and multi-frame detection algorithms generally assume that the background is static, making multi-frame detection algorithms unsuitable for drone applications. This report mainly studies single frame detection algorithms.
The single frame-based detection method is largely classified into a filtering-based method, a Human Visual System (HVS) -based method, an image data structure-based method, and a neural network-based method. The filtering-based method achieves the purpose of detecting the infrared limited pixel target mainly by inhibiting the background, is easily influenced by clutter and noise in the background and influences the robustness of detection; constructing a saliency map capable of highlighting the target through local difference of the target and the background based on an HVS method, and further realizing detection of the target, but the requirement on stability of the background is higher; the method based on the image data structure separates the target from the single-frame image by using the characteristic difference between the target matrix and the background matrix, and mainly focuses on the establishment of a target model and lacks the description of a complex background.
Due to the great success of the deep neural network in the aspect of natural image processing, the convolutional neural network is introduced into the field of infrared finite pixel target detection. Fan et al trained multi-scale CNNs using MNIST datasets, extracting convolution kernels to enhance small target images. Wang et al transfer a CNN pre-trained on ILSVRC 2013 into a small target dataset to learn small target features, but objects in the real world usually contain a large amount of shape, color and structure information and have limited migration effects on limited pixel targets. Lin et al predict on different characteristic graphs by fusing semantic information of a high layer and position information of a low layer, and the method obviously improves small target detection. Cai et al use a series of IoU progressively larger detectors, each trained with positive and negative samples bounded by different IoU thresholds, with the input to the next detector coming from the output of the previous detector, to train a high quality detector. Li et al use a perceptually-generative countermeasure network to improve the characterization of small targets, improving target detection rates.
Disclosure of Invention
In view of this, the main objective of the present invention is to provide a method for detecting a target of a limited-pixel infrared drone based on attention of a mixed domain.
In order to achieve the above purpose, the technical solution of the embodiment of the present invention is realized as follows:
the embodiment of the invention provides a finite pixel infrared unmanned aerial vehicle target detection method based on multi-scale feature fusion, which comprises the following steps:
performing feature extraction on the infrared image through a single-step cascade neural network based on an attention mechanism to obtain a feature map;
and performing up-sampling on the characteristic diagram to obtain a multi-scale characteristic diagram, and performing characteristic fusion and prediction on the multi-scale characteristic diagram respectively to obtain a final target detection result.
In the above scheme, the feature extraction is performed on the infrared image by the attention-based single-step cascade neural network to obtain a feature map, specifically: and taking the single-step cascade neural network as a feature extraction model, acquiring multiple layers of feature maps with different depths, and adding an attention module based on a mixed domain mechanism to improve the attention score of the target region.
In the above scheme, the feature map size of the input residual module in the feature extraction process is 208 × 208 × 32, where 208 × 208 represents the spatial size of the feature map, and 32 represents the number of channels, and a channel attention module and a spatial attention module are introduced into the feature extraction module and are respectively used to extract effective features in multiple channels and in a single-channel spatial feature.
In the above scheme, first, global average pooling and global maximum pooling operations are performed on the infrared image in channel dimension, and then weights are obtained through a network including a hidden layer and added to obtain a final attention weight vector, as shown in formula (1):
wherein,representing an input feature map; sigma represents a sigmoid activation function used for learning nonlinear relations among the features; h denotes a network with only one hidden layer for learning parameters r represents a reduction rate; AvgPool and MaxPool respectively represent average pooling and maximum pooling, wherein the average pooling is to add and average all characteristic values of each channel, and the maximum pooling is to select the maximum value of the characteristic values in each channel; the addition represents element-by-element multiplication, and the feature diagram F' is obtained by splicing according to the channel dimensions;
secondly, performing average pooling and maximum pooling operation based on channel dimension on the feature map F, splicing to obtain a feature map with the channel number of 2, and then performing image segmentation on the feature mapPassing through a 7 × 7 convolution layer, the activation function is Sigmoid, and the weight coefficient M is obtainedSAs shown in formula (2):
wherein, Conv7×7Represents performing convolution operation on the data, and the size of a convolution kernel is 7 multiplied by 7;
finally, a weighting factor M is usedSAnd multiplying the feature layer F' to obtain a new feature map after scaling.
In the above scheme, the upsampling the feature map to obtain a multi-scale feature map, and respectively performing feature fusion and prediction on the multi-scale feature map to obtain a result map specifically includes: according to the pixel value occupied by the unmanned aerial vehicle target, four feature maps with different levels are selected, wherein the feature maps are respectively 13 multiplied by 13, 26 multiplied by 26, 52 multiplied by 52 and 104 multiplied by 104, and multi-level feature fusion and prediction are carried out on the feature maps.
In the scheme, a feature map with the size of 13 × 13 at the 82 th layer of the single-step cascade neural network is detected, and a first result map with the size of 13 × 13 is obtained; secondly, performing 2 times of upsampling on a 79 th layer of feature map with the size of 13 × 13 to obtain a feature map with the size of 26 × 26, performing feature fusion on the upsampled feature map and a 61 st layer of feature map in a splicing manner, performing depth superposition during splicing, splicing the 26 × 26 × 255 feature map and the 26 × 26 × 255 feature map to obtain a 26 × 26 × 768 feature map, and then performing detection on a 94 th layer of a second detection layer to obtain a second result map with the size of 26 × 26; secondly, performing the same processing on the characteristic diagram obtained from the 91 st layer, splicing the characteristic diagram with the characteristic diagram obtained from the 36 th layer, and detecting the characteristic diagram obtained from the 106 th layer to obtain a third result diagram with the size of 52 multiplied by 255; and finally, performing the same operation on the 103 th layer and splicing the layer with the 11 th characteristic layer to obtain a fourth result graph with the size of 104 multiplied by 104.
Compared with the prior art, the invention has the beneficial effects that:
(1) according to the method, the attention mechanism is utilized to improve the attention degree score of the target area;
(2) the invention adopts a multi-scale feature fusion and prediction method to improve the feature extraction capability and the detection accuracy of the limited pixel target.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The embodiment of the invention provides a finite pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention, which is specifically realized by the following steps as shown in figure 1:
step 101: and (4) performing feature extraction by adopting a single-step cascade neural network based on an attention mechanism to obtain a feature map.
Specifically, a single-step cascade neural network is used as a feature extraction model to obtain a plurality of layers of feature maps with different depths, and an attention module based on a mixed domain mechanism is added to improve the attention score of a target region.
The feature extraction module of the single-step cascade neural network consists of 5 residual modules, and each residual module consists of 1 convolution layer of 1 multiplied by 1 and 1 multiplied by 3 and a jump connection structure.
In order to improve the characteristic extraction capability of the convolutional layer to the target, a channel attention and space attention module is added between each residual error module.
The feature map size of the input residual module in the feature extraction process is 208 × 208 × 32, wherein 208 × 208 represents the spatial size of the feature map, 32 represents the number of channels, and a channel attention module and a spatial attention module are introduced into the feature extraction module and are respectively used for extracting effective features in a plurality of channels and in a single-channel spatial feature to optimize the feature quality.
Firstly, respectively performing global average pooling and global maximum pooling on the infrared image in channel dimension, obtaining weights through a network containing a hidden layer, and adding the weights to obtain a final attention weight vector, wherein the formula (1) is as follows:
wherein,representing an input feature map; sigma represents a sigmoid activation function used for learning nonlinear relations among the features; h denotes a network with only one hidden layer for learning parameters r represents a reduction rate; AvgPool and MaxPool respectively represent average pooling and maximum pooling, wherein the average pooling is to add and average all characteristic values of each channel, and the maximum pooling is to select the maximum value of the characteristic values in each channel; the addition represents element-by-element multiplication, and the feature diagram F' is obtained by splicing according to the channel dimensions;
secondly, performing average pooling and maximum pooling operations based on channel dimensions on the feature graph F, splicing to obtain a feature graph with the channel number of 2, and obtaining a weight coefficient M by using a convolution layer with the size of 7 multiplied by 7 and an activation function of SigmoidSAs shown in formula (2):
wherein, Conv7×7Represents performing convolution operation on the data, and the size of a convolution kernel is 7 multiplied by 7;
finally, a weighting factor M is usedSAnd multiplying the feature layer F' to obtain a new feature map after scaling.
Step 102: and acquiring a multi-scale feature map through an up-sampling operation, and respectively carrying out feature fusion and prediction on the multi-scale feature map.
Specifically, according to the pixel value occupied by the unmanned aerial vehicle target, four feature maps with different levels are selected, wherein the feature maps are respectively 13 × 13, 26 × 26, 52 × 52 and 104 × 104, and multi-level feature fusion and prediction are performed on the feature maps.
Detecting a feature map with the size of 13 x 13 at the 82 th layer of the single-step cascade neural network to obtain a first result map with the size of 13 x 13; secondly, performing 2 times of upsampling on a 79 th layer of feature map with the size of 13 × 13 to obtain a feature map with the size of 26 × 26, performing feature fusion on the upsampled feature map and a 61 st layer of feature map in a splicing manner, performing depth superposition during splicing, splicing the 26 × 26 × 255 feature map and the 26 × 26 × 255 feature map to obtain a 26 × 26 × 768 feature map, and then performing detection on a 94 th layer of a second detection layer to obtain a second result map with the size of 26 × 26; secondly, performing the same processing on the characteristic diagram obtained from the 91 st layer, splicing the characteristic diagram with the characteristic diagram obtained from the 36 th layer, and detecting the characteristic diagram obtained from the 106 th layer to obtain a third result diagram with the size of 52 multiplied by 255; and finally, performing the same operation on the 103 th layer and splicing the layer with the 11 th characteristic layer to obtain a fourth result graph with the size of 104 multiplied by 104.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention.
Claims (6)
1. A limited pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention is characterized by comprising the following steps:
performing feature extraction on the infrared image through a single-step cascade neural network based on an attention mechanism to obtain a feature map;
and performing up-sampling on the characteristic diagram to obtain a multi-scale characteristic diagram, and performing characteristic fusion and prediction on the multi-scale characteristic diagram respectively to obtain a final target detection result.
2. The method for detecting the target of the limited-pixel infrared unmanned aerial vehicle based on the attention of the mixed domain as claimed in claim 1, wherein the method for extracting the features of the infrared image by the single-step cascade neural network based on the attention mechanism is used for obtaining a feature map, and specifically comprises the following steps: and taking the single-step cascade neural network as a feature extraction model, acquiring multiple layers of feature maps with different depths, and adding an attention module based on a mixed domain mechanism to improve the attention score of the target region.
3. The mixed-domain attention-based limited-pixel infrared unmanned aerial vehicle target detection method as claimed in claim 2, wherein the feature map size of the input residual module in the feature extraction process is 208 x 32, wherein 208 x 208 represents the spatial size of the feature map, 32 represents the number of channels, and a channel attention module and a spatial attention module are introduced into the feature extraction module and are respectively used for extracting effective features in a plurality of channels and in a single-channel spatial feature.
4. The method for detecting the target of the infrared unmanned aerial vehicle with the limited pixels based on the attention of the mixed domain according to claim 3, wherein the method comprises the following steps of firstly, respectively performing global average pooling and global maximum pooling on the infrared image in a channel dimension, obtaining weights through a network containing a hidden layer, and adding the weights to obtain a final attention weight vector, wherein the final attention weight vector is shown in formula (1):
wherein,representing an input feature map; sigma represents a sigmoid activation function used for learning nonlinear relations among the features; h denotes a network with only one hidden layer for learning parameters r represents a reduction rate; AvgPool and MaxPool respectively represent average pooling and maximum pooling, wherein the average pooling is to add and average all characteristic values of each channel, and the maximum pooling is to select the maximum value of the characteristic values in each channel; addition means element by element multiplication and put throughSplicing the lane dimensions to obtain a characteristic diagram F';
secondly, performing average pooling and maximum pooling operations based on channel dimensions on the feature graph F, splicing to obtain a feature graph with the channel number of 2, and obtaining a weight coefficient M by using a convolution layer with the size of 7 multiplied by 7 and an activation function of SigmoidSAs shown in formula (2):
wherein, Conv7×7Represents performing convolution operation on the data, and the size of a convolution kernel is 7 multiplied by 7;
finally, a weighting factor M is usedSAnd multiplying the feature layer F' to obtain a new feature map after scaling.
5. The method for detecting the target of the limited-pixel infrared unmanned aerial vehicle based on mixed domain attention according to any one of claims 1 to 4, wherein the up-sampling is performed on the feature map to obtain a multi-scale feature map, and feature fusion and prediction are respectively performed on the multi-scale feature map to obtain a final target detection result, specifically: according to the pixel value occupied by the unmanned aerial vehicle target, four feature maps with different levels are selected, wherein the feature maps are respectively 13 multiplied by 13, 26 multiplied by 26, 52 multiplied by 52 and 104 multiplied by 104, and multi-level feature fusion and prediction are carried out on the feature maps.
6. The method for detecting the target of the limited-pixel infrared unmanned aerial vehicle based on the multi-scale feature fusion as claimed in claim 5, wherein a feature map with the size of 13 x 13 at the 82 th layer of the single-step cascade neural network is detected to obtain a first result map with the size of 13 x 13; secondly, performing 2 times of upsampling on a 79 th layer of feature map with the size of 13 × 13 to obtain a feature map with the size of 26 × 26, performing feature fusion on the upsampled feature map and a 61 st layer of feature map in a splicing manner, performing depth superposition during splicing, splicing the 26 × 26 × 255 feature map and the 26 × 26 × 255 feature map to obtain a 26 × 26 × 768 feature map, and then performing detection on a 94 th layer of a second detection layer to obtain a second result map with the size of 26 × 26; secondly, performing the same processing on the characteristic diagram obtained from the 91 st layer, splicing the characteristic diagram with the characteristic diagram obtained from the 36 th layer, and detecting the characteristic diagram obtained from the 106 th layer to obtain a third result diagram with the size of 52 multiplied by 255; and finally, performing the same operation on the 103 th layer and splicing the layer with the 11 th characteristic layer to obtain a fourth result graph with the size of 104 multiplied by 104.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011392343.9A CN112560907B (en) | 2020-12-02 | 2020-12-02 | Finite pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011392343.9A CN112560907B (en) | 2020-12-02 | 2020-12-02 | Finite pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112560907A true CN112560907A (en) | 2021-03-26 |
CN112560907B CN112560907B (en) | 2024-05-28 |
Family
ID=75047298
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011392343.9A Active CN112560907B (en) | 2020-12-02 | 2020-12-02 | Finite pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112560907B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113255693A (en) * | 2021-05-19 | 2021-08-13 | 西华大学 | Unmanned aerial vehicle multi-scale detection and identification method based on imaging metadata assistance |
CN116935045A (en) * | 2023-06-27 | 2023-10-24 | 武汉大学 | Retina blood vessel segmentation method and system based on mixed attention and multi-scale cascade |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110084210A (en) * | 2019-04-30 | 2019-08-02 | 电子科技大学 | The multiple dimensioned Ship Detection of SAR image based on attention pyramid network |
WO2019153908A1 (en) * | 2018-02-11 | 2019-08-15 | 北京达佳互联信息技术有限公司 | Image recognition method and system based on attention model |
AU2020100274A4 (en) * | 2020-02-25 | 2020-03-26 | Huang, Shuying DR | A Multi-Scale Feature Fusion Network based on GANs for Haze Removal |
CN111179217A (en) * | 2019-12-04 | 2020-05-19 | 天津大学 | Attention mechanism-based remote sensing image multi-scale target detection method |
CN111738110A (en) * | 2020-06-10 | 2020-10-02 | 杭州电子科技大学 | Remote sensing image vehicle target detection method based on multi-scale attention mechanism |
CN111738344A (en) * | 2020-06-24 | 2020-10-02 | 上海应用技术大学 | Rapid target detection method based on multi-scale fusion |
CN111767882A (en) * | 2020-07-06 | 2020-10-13 | 江南大学 | Multi-mode pedestrian detection method based on improved YOLO model |
-
2020
- 2020-12-02 CN CN202011392343.9A patent/CN112560907B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019153908A1 (en) * | 2018-02-11 | 2019-08-15 | 北京达佳互联信息技术有限公司 | Image recognition method and system based on attention model |
CN110084210A (en) * | 2019-04-30 | 2019-08-02 | 电子科技大学 | The multiple dimensioned Ship Detection of SAR image based on attention pyramid network |
CN111179217A (en) * | 2019-12-04 | 2020-05-19 | 天津大学 | Attention mechanism-based remote sensing image multi-scale target detection method |
AU2020100274A4 (en) * | 2020-02-25 | 2020-03-26 | Huang, Shuying DR | A Multi-Scale Feature Fusion Network based on GANs for Haze Removal |
CN111738110A (en) * | 2020-06-10 | 2020-10-02 | 杭州电子科技大学 | Remote sensing image vehicle target detection method based on multi-scale attention mechanism |
CN111738344A (en) * | 2020-06-24 | 2020-10-02 | 上海应用技术大学 | Rapid target detection method based on multi-scale fusion |
CN111767882A (en) * | 2020-07-06 | 2020-10-13 | 江南大学 | Multi-mode pedestrian detection method based on improved YOLO model |
Non-Patent Citations (3)
Title |
---|
ZHUANGZHUANG HU等: "Infrared polymorphic target recognition based on single step cascade neural network", PROCEEDINGS OF SPIE * |
沈文祥;秦品乐;曾建潮;: "基于多级特征和混合注意力机制的室内人群检测网络", 计算机应用, no. 12 * |
麻森权;周克;: "基于注意力机制和特征融合改进的小目标检测算法", 计算机应用与软件, no. 05 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113255693A (en) * | 2021-05-19 | 2021-08-13 | 西华大学 | Unmanned aerial vehicle multi-scale detection and identification method based on imaging metadata assistance |
CN116935045A (en) * | 2023-06-27 | 2023-10-24 | 武汉大学 | Retina blood vessel segmentation method and system based on mixed attention and multi-scale cascade |
Also Published As
Publication number | Publication date |
---|---|
CN112560907B (en) | 2024-05-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110188705B (en) | Remote traffic sign detection and identification method suitable for vehicle-mounted system | |
CN110232394B (en) | Multi-scale image semantic segmentation method | |
CN110348376B (en) | Pedestrian real-time detection method based on neural network | |
CN110210551B (en) | Visual target tracking method based on adaptive subject sensitivity | |
CN111079739B (en) | Multi-scale attention feature detection method | |
CN109977774B (en) | Rapid target detection method based on adaptive convolution | |
CN104299006A (en) | Vehicle license plate recognition method based on deep neural network | |
CN113591968A (en) | Infrared weak and small target detection method based on asymmetric attention feature fusion | |
CN113822352B (en) | Infrared dim target detection method based on multi-feature fusion | |
CN111738114B (en) | Vehicle target detection method based on anchor-free accurate sampling remote sensing image | |
CN112529065B (en) | Target detection method based on feature alignment and key point auxiliary excitation | |
CN115631344B (en) | Target detection method based on feature self-adaptive aggregation | |
CN112560907B (en) | Finite pixel infrared unmanned aerial vehicle target detection method based on mixed domain attention | |
CN113408398B (en) | Remote sensing image cloud detection method based on channel attention and probability up-sampling | |
CN114612769A (en) | Integrated sensing infrared imaging ship detection method integrated with local structure information | |
CN112149526B (en) | Lane line detection method and system based on long-distance information fusion | |
CN114241296A (en) | Method for detecting meteorite crater obstacle during lunar landing, storage medium and electronic device | |
CN117095368A (en) | Traffic small target detection method based on YOLOV5 fusion multi-target feature enhanced network and attention mechanism | |
CN114202473A (en) | Image restoration method and device based on multi-scale features and attention mechanism | |
CN112132867B (en) | Remote sensing image change detection method and device | |
Qin et al. | Dense sampling and detail enhancement network: Improved small object detection based on dense sampling and detail enhancement | |
CN116953702A (en) | Rotary target detection method and device based on deduction paradigm | |
CN112926667A (en) | Method and device for detecting saliency target of depth fusion edge and high-level feature | |
CN115311550B (en) | Remote sensing image semantic change detection method and device, electronic equipment and storage medium | |
CN114821356B (en) | Optical remote sensing target detection method for accurate positioning |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |