CN113807324A - Sonar image recognition method and device, electronic equipment and storage medium - Google Patents
Sonar image recognition method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113807324A CN113807324A CN202111286463.5A CN202111286463A CN113807324A CN 113807324 A CN113807324 A CN 113807324A CN 202111286463 A CN202111286463 A CN 202111286463A CN 113807324 A CN113807324 A CN 113807324A
- Authority
- CN
- China
- Prior art keywords
- sonar image
- data set
- image data
- learning model
- recognition
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Signal Processing (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a sonar image identification method, a sonar image identification device, electronic equipment and a storage medium, wherein the method comprises the steps of respectively preprocessing a forward-looking sonar image data set, a multi-beam depth-sounding sonar image data set and a side-scan sonar image data set; respectively constructing a migration learning model corresponding to the forward-looking sonar image data set, the multi-beam depth sounding sonar image data set and the side-scan sonar image data set, and training to obtain three corresponding groups of classification results; the recognition rate of the three groups of classification results is obtained, the three groups of sonar image recognition results are fused in a D-S evidence decision theory algorithm, and the sonar image recognition results are finally obtained.
Description
Technical Field
The invention relates to the technical field of sonar image identification, in particular to a sonar image identification method and device, electronic equipment and a storage medium.
Background
In recent years, with the rapid development of economic society and the need for national infrastructure, the demand for detection and identification of underwater minerals and undersea structures has been increasing. By utilizing the sonar image recognition technology, large-area scanning recognition can be carried out on underwater targets, and the recognition efficiency is improved. However, due to the characteristics of few underwater target data sets, the identification difficulty is always high. Currently, common sonar technologies generally include a forward-looking sonar, a side-scan sonar, a multi-beam depth-finding sonar, and the like, where the forward-looking sonar is an active sonar that "focuses" a sound wave signal in a predetermined direction, i.e., forms a beam in a desired scanning direction. An acoustic signal emitted by a sonar is transmitted underwater and reflected by obstacles such as reefs, naval vessels, fish schools and the like to generate an echo signal. The sonar receives echo signals and converts the echo signals into electric signals, the receiver converts the received electric signals of each array element into a form suitable for equipment processing, and finally forms a sonar scanning image through a series of data processing. The side scan sonar is mainly used for detecting objects and seabed structures, and two transducers arranged on two sides emit sound wave pulses and digitize the sound wave pulses so as to obtain seabed images. The SSS displays the echo energy intensity along a timeline to obtain a continuous image of the seafloor. The side scan sonar can present the image that highly becomes more meticulous, can not only show the existence of object, can also launch a band through the material classification multi-beam depth sounding sonar of strong reflection and weak reflection show object simultaneously, can measure the depth of water value of a line in the twinkling of an eye, along with the removal of hull, the multi-beam sonar line moves into the face, alright in different within ranges measurement depth of water value, be one of the main technique in the present measurement depth of water field.
Different sonars have different recognition results on different targets, and the selection of proper sonar equipment is also very important. Therefore, it is important how to improve the efficiency of sonar image recognition, how to improve the capability of sonar image recognition in a small data set sample, and how to merge the recognition results of a plurality of sonar detection techniques.
Currently, underwater target identification usually uses an underwater robot to carry a side scan sonar to scan a submarine target, and technicians manually identify the scanned data. The method has relatively low recognition and detection efficiency and is not beneficial to large-area application. The sonar image target detection technology is generally to identify a target by combining the result of classifier identification in a mode of manually designing feature extraction such as image segmentation and chrominance feature extraction. This requires a large amount of training data and is single in character and is not widely used in practice.
Disclosure of Invention
In view of the above, the invention provides a sonar image identification method, which is characterized in that three sonar identification results are identified and fused to an instrument through a machine self-learning mode in a migration learning mode, so that dependence on sonar image quantity is reduced, and a sonar image identification result is improved.
The technical scheme is as follows: a sonar image identification method comprises the following steps: respectively preprocessing a forward-looking sonar image data set, a multi-beam depth sounding sonar image data set and a side-scan sonar image data set;
respectively constructing a migration learning model corresponding to the forward-looking sonar image data set, the multi-beam depth sounding sonar image data set and the side-scan sonar image data set, and training to obtain three corresponding groups of classification results;
and respectively solving the recognition rates of the three groups of classification results, and putting the classification results into a D-S evidence decision theory algorithm to fuse the recognition results of the three groups of sonar images to obtain sonar image recognition results.
A sonar image recognition device comprises an acquisition module, a data acquisition module and a data acquisition module, wherein the acquisition module is used for acquiring a forward-looking sonar image data set, a multi-beam depth-sounding sonar image data set and a side-scan sonar image data set;
the preprocessing module is used for preprocessing a forward-looking sonar image data set, a multi-beam sounding sonar image data set and a side-scan sonar image data set;
the training module is used for respectively constructing a transfer learning model corresponding to the forward-looking sonar image data set, the multi-beam sounding sonar image data set and the side-scan sonar image data set, training and acquiring three corresponding groups of classification results;
and the fusion module is used for solving the recognition rate of the three groups of classification results and putting the recognition rate into a D-S evidence decision theory algorithm to fuse the recognition results of the three groups of sonar images so as to finally obtain the sonar image recognition result.
An electronic device, comprising: a processor, a memory;
the memory for storing a computer program;
the processor is used for executing the sonar image identification method according to any one of the above by calling the computer program.
A computer-readable storage medium storing a computer program for implementing the sonar image recognition method as described in any one of the above when executed by a processor.
Compared with the prior art, the invention has the advantages that: the data quantity required by using the transfer learning is less, and meanwhile, the data set features of the image recognition have universality, so that a similar deep learning recognition effect can be achieved by using less data; by using the self-adaptive migration model method, the recognition result can be obtained under the condition of original small data volume, and then the recognition effect is improved by improving the model parameters through the data and the labels obtained in the recognition process; by using the information fusion method, a certain type of data source with low recognition rate can be fused under the condition of other data sources with high recognition rate, so that the overall decision judgment result is improved, and the influence of a certain characteristic with low recognition rate on the overall recognition result is reduced; by combining the two methods, the target accuracy is improved in a small data volume sonar image data set. Meanwhile, the method is a recognition method of homologous heterogeneous data, a recognition strategy of comprehensive judgment is performed according to different sonar image characteristics of the same target, a sonar image recognition method of transfer learning and information fusion is integrated, the method is better applied to the field of image sonars, the accuracy of recognition can be effectively improved, the training time is reduced, the stability of a system is improved, and the influence of single sonar image characteristics on the recognition process is avoided.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used for describing the embodiments will be briefly introduced below. It is obvious that the drawings in the following description are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
FIG. 1 is a system flow architecture diagram of the present invention;
FIG. 2 is a flow chart of transfer learning;
FIG. 3 is a feature fusion flow diagram.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The same reference numerals denote the same or similar parts in the drawings, and thus, a repetitive description thereof will be omitted.
Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the disclosure. One skilled in the relevant art will recognize, however, that the subject matter of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, devices, implementations, or operations have not been shown or described in detail to avoid obscuring aspects of the disclosure.
It is to be understood by those skilled in the art that the drawings are merely schematic representations of exemplary embodiments, and that the blocks or processes shown in the drawings are not necessarily required to practice the present disclosure and are, therefore, not intended to limit the scope of the present disclosure.
The following detailed description of embodiments of the invention is provided in conjunction with the accompanying drawings:
embodiments of this patent are described in conjunction with fig. 1-3.
A sonar image identification method comprises the following steps: respectively preprocessing a forward-looking sonar image data set, a multi-beam depth sounding sonar image data set and a side-scan sonar image data set; respectively constructing a migration learning model corresponding to the forward-looking sonar image data set, the multi-beam depth sounding sonar image data set and the side-scan sonar image data set, and training to obtain three corresponding groups of classification results; and solving the recognition rate of the three groups of classification results, and fusing the recognition results of the three groups of sonar images in a D-S (Dempster-Shafer) evidence decision theory algorithm to finally obtain a sonar image recognition result.
In the field of traditional image recognition, the recognition accuracy of a CNN (Convo l ut i ona l neural network) convolutional neural network reaches the level of artificial recognition, and meanwhile, the derived VGG16 and ResNet models achieve a higher level of processing a large number of image data sets. However, in the field of sonar imaging, since a sonar image is not a natural photographed image but an image formed by an electric signal of a sonar scanning echo, the relevance to an image in reality is not great; the sonar image field is different from the traditional image identification field, the data acquisition is complex, the acquisition cost is high, the data volume is small, and the traditional image identification method is not applicable, so that the sonar image cannot be directly identified by the convolutional neural network. Meanwhile, the feature extraction mode of the sonar image target is different from the data source acquisition mode. Therefore, for the recognition of different underwater targets, even if transfer learning is introduced, ideal recognition capability for a certain sonar image cannot be achieved. In the migration learning step, the initial training set is migrated from other model data, and can be synchronously updated and learned according to the change of test data in the test process, so that the adaptability of the model is improved, and the identification accuracy of a single model is improved. The method combines the transfer learning and the D-S evidence decision theory algorithm, integrates the recognition set of the forward-looking sonar image, the multi-beam depth-sounding sonar image and the side-scan sonar image, and improves the recognition accuracy of the sonar image.
On the basis of the above embodiment, the present invention further provides an implementation manner, that is, the step of constructing the migration learning model includes: selecting a CNN network structure, and training by using an I mageNet data set to obtain a CNN target network; selecting a VGG16 network structure, and training by using an I mageNet data set to obtain a VGG16 target network; selecting an FCN network structure, and training by using an I mageNet data set to obtain an FCN (full convolution neural network) target network; and substituting the forward-looking sonar image features into a CNN network structure to obtain a CNN migration learning model, substituting the side-scan sonar image features into a VGG16 network structure to obtain a VGG16 migration learning model, and substituting the multi-beam sounding sonar image features into an FCN network structure to obtain an FCN migration learning model.
In the present embodiment, since image scanning characteristics of a forward-looking sonar image, a multi-beam depth-sounding sonar image, and a side-scan sonar image are greatly different from each other, only one kind of migration learning model is used to have a very low accuracy in sonar image recognition. The sonar image fusion method has the advantages that the selection of the feature recognition model and the selection of the decision strategy are difficult, the feature recognition model needs to meet the feature requirements of different sonar images, and meanwhile, the selection of network parameters and the data needs need to be adapted. According to the method, an evidence decision theory algorithm is selected from decision algorithms, decision analysis is carried out on the judgment result of the homologous heterogeneous data, the overall judgment effect is improved from a decision layer, and the prediction accuracy of the recognition result is improved. By combining the two methods, the influence of low recognition rate of a single sonar image in a special environment is avoided, and finally the target accuracy is improved in a small data volume data set and multi-angle sonar image data.
On the basis of the above embodiment, the method further comprises the step of training by using a transfer learning model, wherein the step of training by using the transfer learning model comprises the following steps: adjusting a corresponding CNN migration learning model, a VGG16 migration learning model and an FCN migration learning model according to the image quantity of the forward-looking sonar image data set, the multi-beam depth-sounding sonar image data set and the side-scan sonar image data set; correspondingly substituting the forward-looking sonar image data set, the multi-beam sounding sonar image data set and the side-scan sonar image data set into a CNN migration learning model, a VGG16 migration learning model and an FCN migration learning model to obtain three corresponding groups of classification results. In this embodiment, according to the respective characteristics of different sonar images, the different number of each sonar image set, and the identification requirements, it is a migration learning mode with relatively optimal efficiency and accuracy to perform migration training on a forward-looking sonar image with a CNN network structure, to perform training on a side-scan sonar image with a VGG16 target network, and to perform training on a multi-beam depth-finding sonar image with an FCN network structure. The multi-beam sounding sonar image data set is relatively less in acquisition and is suitable for a VGG16 transfer learning model; the FCN migration learning model can classify sonar images at a pixel level and is suitable for the recognition process of side-scan sonar images. Different data sets of the method are suitable for different network training models, the obtained recognition accuracy is the optimal solution under respective networks, and then the optimal solution is put into a fusion module, so that the recognition accuracy of the overall judgment of the target is improved. The method further selects an evidence decision theory algorithm from the decision algorithms on the basis of the three models, and carries out decision analysis on the data judgment results of the common source heterogeneity, so that the overall judgment effect is improved, and the prediction accuracy of the identification result is improved.
Further, in the implementation process of the invention, the preprocessing process of the forward-looking sonar image data set, the multi-beam depth-sounding sonar image data set and the side-scan sonar image data set in the migration learning process further comprises the steps of respectively selecting 80% of each group of image data sets as a training set and 20% of each group of image data sets as a testing set.
Correspondingly substituting the forward-looking sonar image data set, the multi-beam sounding sonar image data set and the side-scan sonar image data set into the CNN migration learning model, the VGG16 migration learning model and the FCN migration learning model to obtain three corresponding groups of classification results, and evaluating the classification performance of different networks by using a verification set, and returning to the step 7 for retraining if the over-fitting or recognition effect is not obvious. Until the classification results of the three groups of models meet the requirements.
The method adopts a D-S evidence decision theory algorithm to fuse three groups of sonar image recognition results. D-S (Dempster-Shafer) evidence decision theory based on uncertain reasoning is commonly used in the fields of target tracking, pattern recognition, medical images and the like. In DS evidence theory, a complete set of basic propositions (hypotheses) that are mutually incompatible is called a recognition framework, representing all possible answers to a question, but only one of which is correct. The following examples illustrate
Assuming that four types of targets need to be identified, the identification framework of the abstract system is as follows: and according to the prior probability, all the combination conditions of the theta are not required to be completely considered, and only the probability model classification conditions of four different types of targets are considered.
The basic probability distribution function is: m { A }, m { B }, m { C }, m { D }. According to the D-S theory, the mass function under the current recognition framework can be expressed as follows:
According to the above formula, to find the synthesis rule mABCDFirst, the normalization coefficient 1-K is calculated. There is only one target tentatively identified, so A, B, C, D intersect each other empty:
first, calculating a conflict factor K:
and secondly, calculating BPA combined with the recognition results of different types of image sonars according to evidence rules:
the mass function value of the recognition result combination of the A-type target is as follows:
In the invention, the DS theory is used for fusing three images, the recognition rate is obtained according to the three groups of classification results, and the recognition results of the three groups of sonar images are fused in a D-S evidence decision theory algorithm to finally obtain the sonar image recognition result.
The invention further provides a sonar image recognition device, which comprises an acquisition module, a data acquisition module and a data acquisition module, wherein the acquisition module is used for acquiring a forward-looking sonar image data set, a multi-beam depth-sounding sonar image data set and a side-scan sonar image data set; the preprocessing module is used for preprocessing a forward-looking sonar image data set, a multi-beam sounding sonar image data set and a side-scan sonar image data set in a segmentation, denoising or other modes; the training module is used for respectively constructing corresponding migration learning models corresponding to the forward-looking sonar image data set, the multi-beam sounding sonar image data set and the side-scan sonar image data set for training, then inputting an initial recognition result obtained by the migration learning models into the models again as a training set for training, adjusting network parameters, increasing the training data volume, improving the self-adaptive capacity of the migration learning models and acquiring three corresponding groups of classification results; and the fusion module is used for solving the recognition rate of the three groups of classification results and putting the recognition rate into a D-S evidence decision theory algorithm to fuse the recognition results of the three groups of sonar images so as to obtain the sonar image recognition result.
On the basis of the device, the device further comprises a pre-learning module, a pre-learning module and a pre-learning module, wherein the pre-learning module is used for selecting a CNN network structure, training by using an ImageNet data set and constructing a CNN target network; selecting a VGG16 network structure, training by using an ImageNet data set, and constructing a VGG16 target network; and selecting an FCN network structure, and training by using the I mageNet data set to construct an FCN target network.
On the basis of the above embodiment, the present invention further includes a verification module for evaluating classification performance of different networks by a verification set.
Further, the present invention provides an electronic device, including: a processor, a memory;
the memory for storing a computer program;
the processor is configured to execute the sonar image recognition method according to any one of the above embodiments by calling the computer program.
A computer-readable storage medium storing a computer program for implementing a sonar image recognition method according to any one of the above embodiments when executed by a processor.
Compared with the sonar image identification effect under the traditional single characteristic, the method for identifying the sonar image by combining the multiple characteristics can effectively improve the identification accuracy, meanwhile, the use of data volume can be reduced by using transfer learning, the accuracy of the multiple sonar image characteristics can be fused by using a D-S evidence decision theory algorithm, and the identification accuracy is improved by using a decision algorithm. By combining the two methods, the influence of low recognition rate of a single sonar image in a special environment is avoided, and finally the target accuracy is improved in a small amount of sonar image data set.
The invention relates to a method for identifying homologous and heterogeneous data, which is an identification strategy for comprehensively judging different sonar image characteristics of the same target. By combining transfer learning and the traditional deep learning method, the multi-source sonar image data model is extracted, the method is better applied to the field of image sonar, the overall recognition accuracy of the system is improved, and negative influences on the recognition process caused by low recognition accuracy of a single sonar image feature in a specific environment are avoided
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiment or embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims (10)
1. A sonar image recognition method is characterized by comprising the following steps:
respectively preprocessing a forward-looking sonar image data set, a multi-beam depth sounding sonar image data set and a side-scan sonar image data set;
respectively constructing a migration learning model corresponding to the forward-looking sonar image data set, the multi-beam depth sounding sonar image data set and the side-scan sonar image data set, and training to obtain three corresponding groups of classification results;
and respectively solving the recognition rates of the three groups of classification results, and putting the classification results into a D-S evidence decision theory algorithm to fuse the recognition results of the three groups of sonar images to obtain sonar image recognition results.
2. The sonar image recognition method according to claim 1, wherein the migration learning model constructing step includes:
selecting a CNN network structure, and training by using an I mageNet data set to obtain a CNN target network; selecting a VGG16 network structure, and training by using an I mageNet data set to obtain a VGG16 target network; selecting an FCN network structure, and training by using an I mageNet data set to obtain an FCN target network;
and substituting the forward-looking sonar image features into a CNN network structure to obtain a CNN migration learning model, substituting the side-scan sonar image features into a VGG16 network structure to obtain a VGG16 migration learning model, and substituting the multi-beam sounding sonar image features into an FCN network structure to obtain an FCN migration learning model.
3. The sonar image-recognition method according to claim 2, wherein the training with the migration learning model includes:
adjusting a corresponding CNN migration learning model, a VGG16 migration learning model and an FCN migration learning model according to the image quantity of the forward-looking sonar image data set, the multi-beam depth-sounding sonar image data set and the side-scan sonar image data set;
correspondingly substituting the forward-looking sonar image data set, the multi-beam sounding sonar image data set and the side-scan sonar image data set into a CNN migration learning model, a VGG16 migration learning model and an FCN migration learning model to obtain three corresponding groups of classification results.
4. The sonar image recognition method according to claim 1, wherein preprocessing a forward-looking sonar image dataset, a multi-beam depth-finding sonar image dataset, and a side-scan sonar image dataset further comprises selecting 80% of each set of image datasets as a training set and 20% of each set of image datasets as a testing set, respectively.
5. The sonar image identification method according to claim 3, wherein a forward looking sonar image data set, a multi-beam depth finding sonar image data set, and a side scan sonar image data set are correspondingly substituted into a CNN migration learning model, a VGG16 migration learning model, and an FCN migration learning model, and the process of obtaining three corresponding groups of classification results further comprises the steps of evaluating classification performance of different networks by using a verification set, and returning to step 7 for retraining if over-fitting or identification effect is not obvious. Until the classification results of the three groups of models meet the requirements.
6. A sonar image recognition apparatus is characterized in that,
the system comprises an acquisition module, a data acquisition module and a data acquisition module, wherein the acquisition module is used for acquiring a forward-looking sonar image data set, a multi-beam depth sounding sonar image data set and a side-scan sonar image data set;
the preprocessing module is used for preprocessing a forward-looking sonar image data set, a multi-beam sounding sonar image data set and a side-scan sonar image data set;
the training module is used for respectively constructing a transfer learning model corresponding to the forward-looking sonar image data set, the multi-beam sounding sonar image data set and the side-scan sonar image data set, training and acquiring three corresponding groups of classification results;
and the fusion module is used for solving the recognition rate of the three groups of classification results and putting the recognition rate into a D-S evidence decision theory algorithm to fuse the recognition results of the three groups of sonar images so as to obtain the sonar image recognition result.
7. The apparatus of claim 6, further comprising a pre-learning module to:
selecting a CNN network structure, and training by using an I mageNet data set to construct a CNN target network;
selecting a VGG16 network structure, training by using an I mageNet data set, and constructing a VGG16 target network;
and selecting an FCN network structure, and training by using the I mageNet data set to construct an FCN target network.
8. The apparatus of claim 6, further comprising a verification module configured to verify that the set evaluates classification performance of different networks.
9. An electronic device, comprising: a processor, a memory;
the memory for storing a computer program;
the processor is used for executing the sonar image identification method according to any one of claims 1 to 5 by calling the computer program.
10. A computer-readable storage medium, characterized in that a computer program is stored, which when executed by a processor, implements the sonar image recognition method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111286463.5A CN113807324A (en) | 2021-11-02 | 2021-11-02 | Sonar image recognition method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111286463.5A CN113807324A (en) | 2021-11-02 | 2021-11-02 | Sonar image recognition method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113807324A true CN113807324A (en) | 2021-12-17 |
Family
ID=78938103
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111286463.5A Pending CN113807324A (en) | 2021-11-02 | 2021-11-02 | Sonar image recognition method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113807324A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494603A (en) * | 2022-02-11 | 2022-05-13 | 哈尔滨工程大学 | Simulation sonar image data generation method based on Unity3D |
| CN116188964A (en) * | 2023-01-09 | 2023-05-30 | 中国海洋大学 | Method for carrying out real-time identification on submarine landslide by utilizing multi-beam image |
| CN117437409A (en) * | 2023-12-14 | 2024-01-23 | 华中师范大学 | Deep learning target automatic identification method and system based on multi-view sound image |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446902A (en) * | 2018-09-22 | 2019-03-08 | 天津大学 | A kind of marine environment based on unmanned platform and the comprehensive cognitive method of target |
| CN111723679A (en) * | 2020-05-27 | 2020-09-29 | 上海五零盛同信息科技有限公司 | Face and voiceprint authentication system and method based on deep migration learning |
| CN112163599A (en) * | 2020-09-08 | 2021-01-01 | 北京工商大学 | Image classification method based on multi-scale and multi-level fusion |
| AU2020103613A4 (en) * | 2020-11-23 | 2021-02-04 | Agricultural Information and Rural Economic Research Institute of Sichuan Academy of Agricultural Sciences | Cnn and transfer learning based disease intelligent identification method and system |
| CN112508901A (en) * | 2020-12-01 | 2021-03-16 | 广州大学 | Underwater structure disease identification method, system and device and storage medium |
| CN112613504A (en) * | 2020-12-17 | 2021-04-06 | 上海大学 | Sonar underwater target detection method |
| CN113052215A (en) * | 2021-03-15 | 2021-06-29 | 浙江大学 | Sonar image automatic target identification method based on neural network visualization |
| CN113052012A (en) * | 2021-03-08 | 2021-06-29 | 广东技术师范大学 | Eye disease image identification method and system based on improved D-S evidence |
-
2021
- 2021-11-02 CN CN202111286463.5A patent/CN113807324A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446902A (en) * | 2018-09-22 | 2019-03-08 | 天津大学 | A kind of marine environment based on unmanned platform and the comprehensive cognitive method of target |
| CN111723679A (en) * | 2020-05-27 | 2020-09-29 | 上海五零盛同信息科技有限公司 | Face and voiceprint authentication system and method based on deep migration learning |
| CN112163599A (en) * | 2020-09-08 | 2021-01-01 | 北京工商大学 | Image classification method based on multi-scale and multi-level fusion |
| AU2020103613A4 (en) * | 2020-11-23 | 2021-02-04 | Agricultural Information and Rural Economic Research Institute of Sichuan Academy of Agricultural Sciences | Cnn and transfer learning based disease intelligent identification method and system |
| CN112508901A (en) * | 2020-12-01 | 2021-03-16 | 广州大学 | Underwater structure disease identification method, system and device and storage medium |
| CN112613504A (en) * | 2020-12-17 | 2021-04-06 | 上海大学 | Sonar underwater target detection method |
| CN113052012A (en) * | 2021-03-08 | 2021-06-29 | 广东技术师范大学 | Eye disease image identification method and system based on improved D-S evidence |
| CN113052215A (en) * | 2021-03-15 | 2021-06-29 | 浙江大学 | Sonar image automatic target identification method based on neural network visualization |
Non-Patent Citations (1)
| Title |
|---|
| 朱兆彤 等: "一种利用迁移学习训练卷积神经网络的声呐图像识别方法", 《水下无人系统学报》, vol. 28, no. 1, pages 90 - 95 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494603A (en) * | 2022-02-11 | 2022-05-13 | 哈尔滨工程大学 | Simulation sonar image data generation method based on Unity3D |
| CN116188964A (en) * | 2023-01-09 | 2023-05-30 | 中国海洋大学 | Method for carrying out real-time identification on submarine landslide by utilizing multi-beam image |
| CN116188964B (en) * | 2023-01-09 | 2024-01-26 | 中国海洋大学 | Method for carrying out real-time identification on submarine landslide by utilizing multi-beam image |
| CN117437409A (en) * | 2023-12-14 | 2024-01-23 | 华中师范大学 | Deep learning target automatic identification method and system based on multi-view sound image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110472627B (en) | End-to-end SAR image recognition method, device and storage medium | |
| CN113807324A (en) | Sonar image recognition method and device, electronic equipment and storage medium | |
| CN109932708B (en) | Method for classifying targets on water surface and underwater based on interference fringes and deep learning | |
| CN111368633A (en) | AUV-based side-scan sonar image identification method | |
| CN113052200B (en) | Sonar image target detection method based on yolov3 network | |
| Zhao et al. | An automatic ship detection method adapting to different satellites SAR images with feature alignment and compensation loss | |
| CN110991516A (en) | Side-scan sonar image target classification method based on style migration | |
| CN116310837B (en) | SAR ship target rotation detection method and system | |
| CN116468995A (en) | Sonar image classification method combining SLIC super-pixel and graph annotation meaning network | |
| CN115187666A (en) | Deep learning and image processing combined side-scan sonar seabed elevation detection method | |
| CN112613504A (en) | Sonar underwater target detection method | |
| CN117635904A (en) | Dynamic self-adaptive feature-aware credible low-speed unmanned aerial vehicle detection method | |
| CN117456346A (en) | Underwater synthetic aperture sonar image target detection method and system | |
| CN115810144A (en) | Underwater suspended sonar target identification method based on area pre-detection | |
| CN113554123B (en) | Automatic sand production ship identification method based on acousto-optic linkage | |
| CN115937520A (en) | Point cloud moving target segmentation method based on semantic information guidance | |
| CN114283082A (en) | Infrared small target detection method based on attention mechanism | |
| Vohra et al. | Detecting underwater discrete scatterers in echograms with deep learning-based semantic segmentation | |
| Oliveira et al. | Probabilistic positioning of a mooring cable in sonar images for in-situ calibration of marine sensors | |
| Muduli et al. | A Review On Recent Advancements In Signal Processing and Sensing Technologies for AUVs | |
| CN114373118B (en) | Underwater target detection method based on improved YOLOV4 | |
| CN117173549B (en) | Multi-scale target detection method and system for synthetic aperture sonar image under complex scene | |
| Xie et al. | Target detection for forward looking sonar image based on deep learning | |
| US20240185423A1 (en) | System and method of feature detection for airborne bathymetry | |
| Ding et al. | FA-YOLO: An Improved YOLOv5 Sonar Image Target Detection Model Based on Focal Attention |
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 |