CN112926383B - Automatic target identification system based on underwater laser image - Google Patents
Automatic target identification system based on underwater laser image Download PDFInfo
- Publication number
- CN112926383B CN112926383B CN202110025071.7A CN202110025071A CN112926383B CN 112926383 B CN112926383 B CN 112926383B CN 202110025071 A CN202110025071 A CN 202110025071A CN 112926383 B CN112926383 B CN 112926383B
- Authority
- CN
- China
- Prior art keywords
- target
- underwater
- image
- laser image
- underwater laser
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- 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
- G06N3/08—Learning methods
-
- 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
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an automatic target identification system based on an underwater laser image, which realizes automatic feature extraction and target identification of an underwater target based on the underwater laser image, adopts a laser image containing one or more underwater targets as input, improves the quality of the underwater laser image through image enhancement, carries out automatic feature extraction and identification model modeling of the underwater laser target based on a constructed underwater laser image database, and carries out new underwater laser target identification by using the constructed identification model. The invention realizes automatic feature extraction and target identification, can identify a plurality of targets in the same underwater laser image, has the advantages of high accuracy, high speed, strong reliability and the like, and solves the defects of complicated steps, low identification accuracy, low speed and the like of the conventional underwater laser target identification, only being suitable for a single target, and the like.
Description
Technical Field
The invention relates to the field of underwater laser target identification, in particular to an automatic target identification system based on an underwater laser image.
Background
The ocean area accounts for about 70% of the earth surface, abundant resources such as mineral products, organisms, energy sources and the like in the ocean are huge wealth naturally endowed to human beings, the human beings are required to continuously explore, and since researchers in 1963 find that the attenuation of seawater to blue-green light in a wave band of 470nm-580nm is weaker than that of the seawater in light attenuation of other wave bands, the detection application of blue-green laser in the seawater becomes a focus of numerous scholars. Meanwhile, underwater detection based on sound waves is seriously influenced by seawater, topography, marine organisms and the like, the quality of sonar underwater detection is seriously influenced, underwater laser detection is not easily influenced by seawater temperature, salinity and the like, two-dimensional imaging can be directly realized, better target distance measurement and imaging effects are achieved, the underwater laser detection can be applied to underwater detection and identification of mines, submarines and the like, the underwater navigation collision avoidance of sensitive water areas can be implemented, and the underwater detection device can also be applied to hydrological survey, underwater operation maintenance, underwater environment monitoring, marine fish school detection, submarine topography exploration, marine organism research and the like, so that the underwater laser detection technology has important significance for future marine exploration.
At present, most of traditional underwater laser target recognition is based on laser images which are acquired aiming at targets individually at a short distance, the targets are obvious and single, the rapid development of current ocean exploration cannot be met, machine learning algorithms show prominent superiority in recent years, more achievements are obtained particularly in the field of target recognition, only the current underwater laser target recognition is limited to a method of target contour extraction-feature extraction-traditional classifier classification of single target recognition, the steps are complicated, and the recognition efficiency and the recognition accuracy are low, so that the invention has the advantages of high accuracy, high recognition speed, high intelligence degree, and urgent and important significance at present and can be used for automatic underwater laser image target recognition.
Disclosure of Invention
The invention aims to provide a system which has high accuracy, high recognition speed and high intelligence degree and can be used for automatically recognizing underwater laser image targets, aims to realize automatic feature extraction and target recognition of the underwater targets based on the underwater laser images, adopts laser images containing one or more underwater targets as input, improves the quality of the underwater laser images through image enhancement, models an underwater laser image target recognition model based on a built underwater laser image database, performs newly acquired underwater laser target recognition by using the built recognition model, and finally outputs a recognition result. The method realizes automatic feature extraction and target identification, can identify a plurality of targets in the same underwater laser image, has the advantages of high accuracy, high speed, strong reliability and the like, and solves the defects that the conventional underwater laser target identification steps are complicated (target segmentation, feature extraction, classifier modeling and the like are required), is only suitable for a single target, and has low identification accuracy, low speed and the like.
The purpose of the invention is realized by the following technical scheme: an automatic target identification system based on an underwater laser image comprises an underwater target laser image acquisition module, an automatic target identification system based on an underwater laser image and a display module which are sequentially connected, wherein the automatic target identification system based on the underwater laser image comprises an underwater laser image database, a preprocessing module, an underwater laser image target modeling module, an underwater laser image target automatic identification module and an underwater target identification output module.
The underwater target laser image acquisition module acquires an underwater laser image containing a target through a laser transmitter and a receiver.
Furthermore, the underwater laser image database is used for storing all the underwater laser target images acquired in history and target position and category information contained in the images, so that a data basis is provided for the underwater laser image target modeling module, meanwhile, the module can update the newly acquired underwater laser images in real time from the laser image acquisition device, the database content is perfected, so that a basis is provided for updating the model, and for the newly acquired underwater laser images, the target positions and the categories in the images need to be artificially marked and then stored in the database.
Furthermore, the preprocessing module is used for preprocessing the underwater laser target image, and due to the absorption and scattering effects of a water body on laser, the underwater laser image has more noise spots, low contrast and fuzzy target and is not beneficial to target identification, wherein the only difference between the preprocessing of data in the underwater laser image database and the preprocessing of the newly acquired underwater target laser image is that the data used by the preprocessing is divided into a training set and a verification set, so that the model effect can be verified in the modeling module and an ideal model can be finally obtained, in addition, the preprocessing module for the underwater laser target image is mainly completed by adopting the following processes:
for the convenience of subsequent operations, image normalization is firstly performed:
let x j For a point in the image, the laser image is processed as follows to obtain a normalized feature
Wherein x min Is x j Minimum value of (1), x max Is x j Maximum value of (d):
the image definition is improved by adopting gray level conversion, the target area is highlighted, and the visual effect is enhanced:
dividing the laser image into L gray levels according to gray values, wherein the number of pixels of the ith gray level is n i Probability of occurrence p i ＝n i N, where n is the number of all pixels, the gray scale conversion of each gray scale is performed according to the following formula
Wherein
Is an accumulated gray value.
Due to the particularity of laser propagation under water, a laser image often contains a large amount of speckle noise, thermal noise and other various noise types, and noise points may be increased after gray level conversion is performed on the laser image, so that image denoising is performed firstly in order to improve image quality and facilitate subsequent identification, and as the related noise types are many and the noise problem is more serious compared with the image under natural light or an air medium, a space domain and transform domain double denoising method is adopted:
in the spatial domain, the following mode is adopted, and the value of the jth point after filtering is as follows:
where l represents each pixel within the filter kernel, ω l The best weight to debug for reality.
And under a transform domain, performing wavelet packet decomposition on the laser image, filtering the high-frequency part, restoring to obtain a de-noised image with high-frequency noise removed, and manually setting a filtering threshold according to an effect.
If the number of samples in the database is more than 1000, extracting 40% of data from the database as a training set, and taking the rest data as a verification set; for types with a sample number less than 1000, 60% of the data is extracted for training and the remainder is used for validation. The recognition effect of the model can be viewed through the validation set.
Further, the underwater laser image target modeling module comprehensively utilizes the strong feature extraction capability of the neural network and the rapid classification capability of machine learning to establish a high-accuracy underwater laser image target automatic identification model, the model can automatically learn how to extract effective features and identify the effective features based on a training set, and the model is specifically realized as follows:
inputting an underwater laser image into a Resnet18 network to extract a characteristic diagram, obtaining positions of a plurality of targets in the image by adopting an RPN network based on the characteristic diagram to obtain candidate frames, only keeping frames with the coincidence degree of more than 95% with the candidate frames in a label through filtering, namely, frames with the targets existing, and discarding the rest frames without the targets, after obtaining the position frames of the plurality of targets, dividing each target area into 16 × 16, dividing each small area into 4 parts in an average way, obtaining a value of a central position through bilinear interpolation of each part, and taking an average value of the four values as a value of the small area, so that a plurality of target matrixes with the size of 16 × 16 are obtained, and each matrix can be regarded as a sample for identification.
The classification network consists of three layers of input layer, hidden layer and output layer, the number of nodes is P, M and O, for the ith input sample
Its predicted value
Calculated by the following formula
Wherein gamma is o ＝[γ o1 …γ oM ] T O =1, \ 8230where O is the connection weight of the intermediate layer and the output layer, w m ＝[w m1 …w mP ] T M =1, \ 8230, M is the connection weight g of the input layer and the middle layer as a kernel function, and b is a bias term. The value of gamma is obtained by solving the following objective function
Wherein phi = [ gamma ] 1 …γ O ] T Where λ is the regularization parameter, ε is the prediction error, N is the total number of all samples,
the model is optimized by further modifying the parameters by observing the test results of the model in the validation set. Finally, a model C is obtained.
Further, the underwater laser image target automatic identification module is used for acquiring a newly acquired underwater target laser image processed by the preprocessing module
Identifying, namely directly inputting the preprocessed image into an end-to-end target identification model C integrating feature extraction, candidate frame selection and identification for the newly acquired underwater laser image to obtain the positions, sizes and types of all targets contained in the image
Further, the underwater target recognition output module outputs the recognition result, and the recognition result is multi-target recognition, so that the output is not only the type, but also the position, the size and the type of each target. Thereby finally realizing the automatic identification of the underwater laser image target.
Further, the display module outputs and displays the position, size and type of the underwater laser multiple targets obtained by the underwater target identification output module through the display screen.
The technical conception of the invention is as follows: the invention realizes the automatic feature extraction and target recognition of underwater targets based on underwater laser images, the automatic target recognition system based on the underwater laser images adopts the laser images containing one or more underwater targets as input, improves the quality of the underwater laser images through image enhancement, carries out the automatic feature extraction and recognition model modeling of the underwater laser targets based on the built underwater laser image database, carries out the recognition of the newly obtained underwater laser targets by using the built recognition model, and finally outputs the recognition result. The method realizes automatic feature extraction and target identification, can identify a plurality of targets in the same underwater laser image, has the advantages of high accuracy, high speed, strong reliability and the like, and solves the defects that the conventional underwater laser target identification steps are complicated (target segmentation, feature extraction, classifier modeling and the like are required), is only suitable for a single target, and has low identification accuracy, low speed and the like.
The invention has the following beneficial effects: 1. the underwater laser target image is updated through the database, so that real-time updating is realized, and the adaptability to a new target type is improved; 2. by preprocessing an airspace and a transform domain of the underwater laser, the image quality is improved, high-quality target characteristics are provided, and the identification accuracy is improved; 3. the feature extraction and identification are automatically completed by the model; 4. a plurality of targets in one image can be identified simultaneously; 5. the newly acquired underwater laser image can be automatically identified end to end without grading or manual participation, and the identification process is simple and quick;
drawings
FIG. 1 is a hardware connection diagram of an underwater laser image based automatic target identification system;
fig. 2 is a functional block diagram of an underwater laser image-based automatic target identification system.
Detailed Description
The invention is further illustrated below with reference to the figures and examples:
referring to fig. 1 and 2, an underwater target laser image acquisition 1, an underwater laser image-based target automatic identification system 2 and a display module 3 are sequentially connected, wherein the underwater laser image-based target automatic identification system 2 comprises an underwater laser image database 4, a preprocessing module 5, an underwater laser image target modeling module 6, an underwater laser image target automatic identification module 7 and an underwater target identification output module 8.
The underwater target laser image acquisition module 1 acquires an underwater laser image containing a target through a laser transmitter and a receiver.
The underwater laser image database 4 is used for storing all the underwater laser target images acquired in history and target position and category information contained in the images, so that a data basis is provided for the underwater laser image target modeling module 6, meanwhile, the module can update the newly acquired underwater laser images in real time from the laser image acquisition device, the database content is perfected, so that a basis is provided for updating the model, and for the newly acquired underwater laser images, the target positions and the categories in the images need to be marked artificially and then stored in the database.
The preprocessing module 5 is used for preprocessing an underwater laser target image, and the underwater laser image has many noise spots, low contrast, fuzzy target and is not beneficial to target identification due to the absorption and scattering effect of a water body on laser, wherein the only difference between the preprocessing of data in the underwater laser image database 4 and the preprocessing of a newly acquired underwater target laser image is that the data used by the former are divided into a training set and a verification set so as to verify the model effect in the modeling module and finally obtain an ideal model, in addition, the preprocessing module 5 for the underwater laser target image is mainly completed by adopting the following processes:
for the convenience of subsequent operations, image normalization is firstly performed:
let x j For a point in the image, the laser image is processed as follows to obtain a normalized feature
Wherein x min Is x j Minimum value of (1), x max Is x j Maximum value of (d):
the image definition is improved by adopting gray level conversion, the target area is highlighted, and the visual effect is enhanced:
dividing the laser image into L gray levels according to gray values, wherein the number of pixels of the ith gray level is n i Probability of occurrence p i ＝n i N, where n is the number of all pixels, the gray scale conversion of each gray scale is performed according to the following formula
Wherein
Is an accumulated gray value.
Due to the particularity of laser propagation under water, a laser image often contains a large amount of speckle noise, thermal noise and other various noise types, and noise points may be increased after gray level conversion is performed on the laser image, so that image denoising is performed firstly in order to improve image quality and facilitate subsequent identification, and as the related noise types are many and the noise problem is more serious compared with the image under natural light or an air medium, a space domain and transform domain double denoising method is adopted:
in the spatial domain, the following mode is adopted, and the value of the jth point after filtering is as follows:
where l represents each pixel within the filter kernel, ω l The best weight to debug for reality.
And under a transform domain, performing wavelet packet decomposition on the laser image, filtering the high-frequency part, restoring to obtain a de-noised image with high-frequency noise removed, and manually setting a filtering threshold according to an effect.
If the number of samples in the database is more than 1000, extracting 40% of data from the database as a training set, and taking the rest data as a verification set; for types with a sample size less than 1000, 60% of the data is extracted for training and the remainder is used for validation. The recognition effect of the model can be viewed through the validation set.
The underwater laser image target modeling module 6 comprehensively utilizes the strong feature extraction capability of a neural network and the rapid classification capability of machine learning to establish a high-accuracy automatic underwater laser image target identification model, the model can automatically learn how to extract effective features and identify the effective features based on a training set, and the model is specifically realized as follows:
inputting an underwater laser image into a Resnet18 network to extract a characteristic diagram, obtaining positions of a plurality of targets in the image by adopting an RPN network based on the characteristic diagram to obtain candidate frames, only keeping frames with the coincidence degree of more than 95% with the candidate frames in a label through filtering, namely, frames with the targets existing, and abandoning the rest frames without the targets, after obtaining the position frames of the plurality of targets, dividing each target area into 16 × 16, equally dividing each small area into 4 parts, obtaining a value of a central position through bilinear interpolation of each part, and taking an average value of the four values as a value of the small area, so that a plurality of target matrixes with the size of 16 × 16 are obtained, and each matrix can be regarded as a sample to be identified.
The classification network consists of three layers of input layer, hidden layer and output layer, the number of nodes is P, M and O, for the ith input sample
Its predicted value
Calculated by the following formula
Wherein gamma is o ＝[γ o1 …γ oM ] T O =1, \ 8230where O is the connection weight of the intermediate layer and the output layer, w m ＝[w m1 …w mP ] T M =1, \ 8230, M is the connection weight g of the input layer and the middle layer as a kernel function, and b is a bias term. The value of gamma is obtained by solving the following objective function
Wherein phi = [ gamma ] 1 …γ O ] T Where λ is the regularization parameter, ε is the prediction error, N is the total number of all samples,
the model is optimized by further modifying the parameters by observing the test results of the model in the validation set. Finally obtaining a model C.
The underwater laser image target automatic identification module 7 is used for acquiring a newly acquired underwater target laser image processed by the preprocessing module 5
Identification is carried out, and for the newly acquired underwater laser image, the preprocessed underwater laser image isDirectly inputting the image into an end-to-end target recognition model C integrating feature extraction, candidate frame selection and recognition to obtain the positions, sizes and types of all targets contained in the image
The underwater target identification output module 8 outputs the result of identification, and the result is not only the type but also the position, size and type of each target due to multi-target identification. Thus, automatic underwater laser image target identification is finally realized.
The display module 3 outputs and displays the position, size and type of the underwater laser multiple targets obtained by the underwater target identification output module 8 through a display screen.
The above-described embodiments are intended to illustrate rather than to limit the invention, and any modifications and variations of the present invention are within the spirit of the invention and the scope of the appended claims.
Claims (1)
1. An automatic target identification system based on underwater laser images is characterized in that: the underwater target automatic identification system based on the underwater laser image comprises an underwater laser image database, a preprocessing module, an underwater laser image target modeling module, an underwater laser image target automatic identification module and an underwater target identification output module which are connected in sequence;
the underwater laser image database stores all the historically acquired underwater laser target images and target position and category information contained in the images, and meanwhile, the module can update the newly acquired underwater laser images in real time from the laser image acquisition device, and for the newly acquired underwater laser images, the target positions and the categories in the images need to be artificially marked and then stored in the database;
the preprocessing module is used for preprocessing the underwater laser target image and mainly comprises the following steps:
firstly, normalizing the image and setting x j For a point in the image, the laser image is processed as follows to obtain a normalized feature
Wherein x is min Is x j Minimum value of (1), x max Is x j Maximum value of (d):
improving the image definition by adopting gray level conversion:
dividing the laser image into L gray levels according to gray values, wherein the number of pixels of the ith gray level is n i Probability of occurrence p i ＝n i N, where n is the number of all pixels, the gray scale conversion of each gray scale is performed according to the following formula
Wherein
To accumulate the gray values, c min Is the minimum value of the gray value, h i Is the converted gray level;
carrying out laser image denoising by adopting a space domain and transform domain dual denoising method:
the following mode is adopted in a spatial domain, and the value of the jth point after filtering is as follows:
where l represents each pixel within the filter kernel, ω l The optimal weight for actual debugging;
in a transform domain, performing wavelet packet decomposition on a laser image, filtering a high-frequency part, and restoring to obtain a de-noised image without high-frequency noise, wherein a filtering threshold value is artificially set according to an effect;
if the number of samples in the database is more than 1000, extracting 40% of data from the database as a training set, and taking the rest data as a verification set; for types with sample numbers less than 1000, 60% of the data is extracted for training, and the remainder is used for validation; checking the recognition effect of the model through the verification set;
the underwater laser image target modeling module establishes a high-accuracy underwater laser image target automatic identification model, and automatically learns how to extract effective features and identify the effective features based on a training set, and the model is specifically realized as follows:
inputting an underwater laser image into a Resnet18 network to extract a characteristic diagram, acquiring positions of a plurality of targets in the image by adopting an RPN network based on the characteristic diagram to obtain candidate frames, only keeping frames with the coincidence degree of more than 95% with the candidate frames in a label through filtering, namely, considering that the target exists, and considering that the rest frames do not have the target and discard the targets, after obtaining the position frames of the plurality of targets, dividing each target area into 16 × 16, averagely dividing each small area into 4 parts, obtaining a value of a central position of each part through bilinear interpolation, and taking an average value of the four values as a value of the small area, thereby obtaining a plurality of target matrixes with the size of 16 × 16, and regarding each matrix as a sample for identification;
the classification network consists of three layers of input layer, hidden layer and output layer, the number of nodes is P, M and O, for the ith input sample
Its predicted value
Calculated by the following formula
Whereinγ o ＝[γ o1 …γ oM ] T O =1, \ 8230where O is the connection weight of the intermediate layer and the output layer, w m ＝[w m1 …w mP ] T M =1, \ 8230, M is the connection weight of the input layer and the middle layer, g is the kernel function, b is the bias term; the value of gamma is obtained by solving the following objective function
Wherein phi = [ gamma ] 1 …γ O ] T Where λ is the regularization parameter, ε is the prediction error, N is the total number of all samples,
further modifying the parameters by observing the test results of the model in the validation set, thereby optimizing the model; finally obtaining a model C;
the underwater laser image target automatic identification module is used for acquiring a newly acquired underwater target laser image processed by the preprocessing module
Identifying, and directly inputting the preprocessed image into the model C for the newly acquired underwater laser image to obtain the positions, sizes and types of all targets contained in the image
The underwater target identification output module outputs the result obtained by identification, and because the result is multi-target identification, the position, the size and the type of each target are output instead of the type; thus, automatic underwater laser image target identification is finally realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110025071.7A CN112926383B (en) | 2021-01-08 | 2021-01-08 | Automatic target identification system based on underwater laser image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110025071.7A CN112926383B (en) | 2021-01-08 | 2021-01-08 | Automatic target identification system based on underwater laser image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112926383A CN112926383A (en) | 2021-06-08 |
| CN112926383B true CN112926383B (en) | 2023-03-03 |
Family
ID=76163672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110025071.7A Active CN112926383B (en) | 2021-01-08 | 2021-01-08 | Automatic target identification system based on underwater laser image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112926383B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2154842C1 (en) * | 1999-10-14 | 2000-08-20 | Государственное предприятие "Всероссийский научно-исследовательский институт физико-технических и радиотехнических измерений" | Technique for detection and identification of underwater target |
| CN108564109A (en) * | 2018-03-21 | 2018-09-21 | 天津大学 | A kind of Remote Sensing Target detection method based on deep learning |
| CN108596156A (en) * | 2018-05-14 | 2018-09-28 | 浙江大学 | A kind of intelligence SAR radar airbound target identifying systems |
| CN110493509A (en) * | 2019-06-26 | 2019-11-22 | 盐城华昱光电技术有限公司 | A kind of pattern information processing system and processing method of mode of laser group |
| CN110503112A (en) * | 2019-08-27 | 2019-11-26 | 电子科技大学 | A kind of small target deteection of Enhanced feature study and recognition methods |
| EP3709216A1 (en) * | 2018-02-09 | 2020-09-16 | Bayerische Motoren Werke Aktiengesellschaft | Methods and apparatuses for object detection in a scene represented by depth data of a range detection sensor and image data of a camera |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104967832B (en) * | 2015-07-25 | 2016-09-14 | 江苏江北船业有限公司 | A kind of area detecting method that launches a ship based on image procossing |
| CN109543585A (en) * | 2018-11-16 | 2019-03-29 | 西北工业大学 | Underwater optics object detection and recognition method based on convolutional neural networks |
| CN111626993A (en) * | 2020-05-07 | 2020-09-04 | 武汉科技大学 | Image automatic detection counting method and system based on embedded FEFnet network |
-
2021
- 2021-01-08 CN CN202110025071.7A patent/CN112926383B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2154842C1 (en) * | 1999-10-14 | 2000-08-20 | Государственное предприятие "Всероссийский научно-исследовательский институт физико-технических и радиотехнических измерений" | Technique for detection and identification of underwater target |
| EP3709216A1 (en) * | 2018-02-09 | 2020-09-16 | Bayerische Motoren Werke Aktiengesellschaft | Methods and apparatuses for object detection in a scene represented by depth data of a range detection sensor and image data of a camera |
| CN108564109A (en) * | 2018-03-21 | 2018-09-21 | 天津大学 | A kind of Remote Sensing Target detection method based on deep learning |
| CN108596156A (en) * | 2018-05-14 | 2018-09-28 | 浙江大学 | A kind of intelligence SAR radar airbound target identifying systems |
| CN110493509A (en) * | 2019-06-26 | 2019-11-22 | 盐城华昱光电技术有限公司 | A kind of pattern information processing system and processing method of mode of laser group |
| CN110503112A (en) * | 2019-08-27 | 2019-11-26 | 电子科技大学 | A kind of small target deteection of Enhanced feature study and recognition methods |
Non-Patent Citations (3)
| Title |
|---|
| Effective Crack Damage Detection Using Multilayer Sparse Feature Representation and Incremental Extreme Learning Machine;Baoxian Wang 等;《Applied Sciences》;20190212;第1-21页 * |
| Squeeze excitation densely connected residual convolutional networks for specific emitter identification based on measured signals;Zining Wan 等;《Measurement Science and Technology》;20201127;第1-11页 * |
| 深度残差网络的无人机多目标识别;翟进有 等;《图学学报》;20190228;第40卷(第1期);第158-164页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112926383A (en) | 2021-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110245608B (en) | Underwater target identification method based on half tensor product neural network | |
| CN110472627B (en) | End-to-end SAR image recognition method, device and storage medium | |
| CN109871902B (en) | SAR small sample identification method based on super-resolution countermeasure generation cascade network | |
| CN110084234B (en) | Sonar image target identification method based on example segmentation | |
| CN113469177A (en) | Drainage pipeline defect detection method and system based on deep learning | |
| CN111368633A (en) | AUV-based side-scan sonar image identification method | |
| CN108389188A (en) | A kind of sparse EO-1 hyperion Anomaly target detection method | |
| CN114266892A (en) | Pavement disease identification method and system for multi-source data deep learning | |
| CN112949380B (en) | Intelligent underwater target identification system based on laser radar point cloud data | |
| CN113312993B (en) | Remote sensing data land cover classification method based on PSPNet | |
| CN111563408B (en) | High-resolution image landslide automatic detection method with multi-level perception characteristics and progressive self-learning | |
| CN114266977A (en) | Multi-AUV underwater target identification method based on super-resolution selectable network | |
| CN111783616B (en) | Nondestructive testing method based on data-driven self-learning | |
| Long et al. | Underwater forward-looking sonar images target detection via speckle reduction and scene prior | |
| CN112613504A (en) | Sonar underwater target detection method | |
| CN116468995A (en) | Sonar image classification method combining SLIC super-pixel and graph annotation meaning network | |
| CN115114949A (en) | Intelligent ship target identification method and system based on underwater acoustic signals | |
| Shah et al. | An enhanced YOLOv5 model for fish species recognition from underwater environments | |
| CN112926383B (en) | Automatic target identification system based on underwater laser image | |
| CN116243289A (en) | Unmanned ship underwater target intelligent identification method based on imaging sonar | |
| CN112906458B (en) | Group intelligent optimized underwater laser multi-target end-to-end automatic identification system | |
| CN115631211A (en) | Hyperspectral image small target detection method based on unsupervised segmentation | |
| CN115187855A (en) | Seabed substrate sonar image classification method | |
| CN115223044A (en) | End-to-end three-dimensional ground penetrating radar target identification method and system based on deep learning | |
| CN113554671A (en) | Method and device for converting SAR image into visible light image based on contour enhancement |
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 |