CN117611856B - Method for clustering and analyzing echo data of small target of interest in synthetic aperture sonar image - Google Patents
Method for clustering and analyzing echo data of small target of interest in synthetic aperture sonar image Download PDFInfo
- Publication number
- CN117611856B CN117611856B CN202311322405.2A CN202311322405A CN117611856B CN 117611856 B CN117611856 B CN 117611856B CN 202311322405 A CN202311322405 A CN 202311322405A CN 117611856 B CN117611856 B CN 117611856B
- Authority
- CN
- China
- Prior art keywords
- interest
- echo data
- synthetic aperture
- aperture sonar
- cluster analysis
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000007621 cluster analysis Methods 0.000 claims abstract description 67
- 238000004458 analytical method Methods 0.000 claims abstract description 23
- 238000001514 detection method Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 38
- 230000004913 activation Effects 0.000 claims description 24
- 238000012360 testing method Methods 0.000 claims description 9
- 238000004140 cleaning Methods 0.000 claims description 7
- 230000014509 gene expression Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 28
- 230000006870 function Effects 0.000 description 25
- 230000008569 process Effects 0.000 description 8
- 235000004257 Cordia myxa Nutrition 0.000 description 6
- 244000157795 Cordia myxa Species 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 239000011435 rock Substances 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000013211 curve analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000001303 quality assessment method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- 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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/05—Underwater scenes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The invention provides a clustering analysis method for small target echo data of interest of a synthetic aperture sonar image, which comprises the following steps: processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method; correlating the position information of the small object of interest with the echo data to obtain the echo data of the small object of interest; the echo data of the small target of interest is input into a pre-established and trained cluster analysis model, so that real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image is realized; the cluster analysis model employs a modified AutoEncoder network. The method analyzes the underwater small object of interest in the echo domain, improves the interpretability of the underwater small object of interest, and provides an effective solution for the task of the underwater object fine identification based on the SAS image.
Description
Technical Field
The invention relates to the field of underwater acoustic signal processing, in particular to a clustering analysis method for small target echo data of interest of a synthetic aperture sonar image.
Background
The synthetic aperture sonar (SYNTHETIC APERTURE SONAR, SAS) is a high-resolution underwater imaging sonar, and the basic principle is that a virtual large aperture is formed by using the movement of a small aperture matrix, so that the high resolution of the azimuth direction is obtained. Compared with the common side-scan sonar, the SAS has the most remarkable advantages that the azimuth resolution is higher, and the theoretical resolution is irrelevant to the target distance and the adopted sound wave frequency band. The synthetic aperture sonar image target detection task plays an important role in autonomous navigation and search of an underwater unmanned platform. The underwater small targets of the synthetic aperture sonar images have the problems of intra-class difference and inter-class similarity, and under the condition of lacking priori knowledge, the underwater small targets are difficult to confirm only from an image domain. The array element echo data contains feature information rich in targets, and further analysis of small targets of interest is expected to be realized by means of the echo data.
AutoEncoder is a very important unsupervised learning method in deep learning, and can automatically learn from a large amount of unlabeled data to obtain effective features contained in the data. It is essentially an unsupervised form of a deep belief network (Deep belief nets, DBN) consisting of a number of layer-by-layer iterations of the constraint boltzmann machines (RESTRICTED BOLTZMANN MACHINES, RBM). In the preprocessing stage (Per-training), autoEncoder is the same as the training mode of the DBN, and a large amount of unlabeled data is utilized to enable the initial value perception of the model parameters to be in a reasonable range; in the reverse tuning step (Fine-tuning), the DBN model uses the Wake-sleep algorithm to Fine tune the parameters of the model, and AutoEncoder first constructs a symmetrical network for generating the original input data, this process is called expansion (Unrolling), then Fine-tunes the parameters of the model by using the difference between the original data and the generated data, the whole process does not need to use tag data, and after enough iterative operations, the model can accurately reconstruct the original input data. The noise reduction self-encoder proposed by Vincent et al further improves the stability of the encoder network, and in the process of training the noise reduction self-encoder, noise added data is input for training, so that the output of the decoder is the same as the data without noise added, and better noise resistance is obtained to improve the clustering precision. When there are more hidden nodes than input nodes, the self-encoder loses the ability to automatically learn the sample features, and some constraint is required on the hidden nodes. Xu et al propose a sparse self-encoder, which is based on a conventional self-encoder by adding some sparsity constraints. This sparsity is directed to the hidden layer neurons of the self-encoder, which achieve a sparse effect by suppressing most of the output of the hidden layer neurons. If the input signals can be perfectly reconstructed through the sparse expressions of the hidden layer neurons, the sparse expressions are described to contain most of the main characteristics of the input signals and can be regarded as a simple representation of the input data, so that the dimension of the data is greatly reduced on the basis of ensuring the reconstruction accuracy of the model, and the performance of the model is greatly improved.
In recent years, various excellent performances obtained by convolutional neural networks have directly driven the generation of convolutional self-encoders. Strictly speaking, a convolutional self-encoder belongs to a special case of a traditional self-encoder, and the convolutional layer and the pooling layer are used for replacing the original full-connection layer. The conventional self-encoder generally uses a full-connection layer, which has no influence on a one-dimensional signal, but loses spatial information on a two-dimensional image or video signal, and the convolution self-encoder can well reserve the spatial information of the two-dimensional signal by adopting convolution operation. The convolutional self-encoder is very similar to the conventional self-encoder, and the main difference is that the convolutional self-encoder performs linear transformation on an input signal in a convolutional manner, and weights thereof are shared, as in the convolutional neural network.
In view of the fact that the type of the underwater small object of interest is largely unknown, the distribution characteristic of the small object of interest can be better reflected based on the unsupervised cluster analysis model AutoEncoder. However, the full-connection-based AutoEncoder model describes echo data as a whole, lacks the capability of expressing local characteristics of echoes, and has the influence on the cluster analysis effect. In view of the foregoing, a method suitable for cluster analysis of underwater interesting small target echo data is urgently needed at present to improve accuracy and efficiency of target refined identification.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a clustering analysis method for echo data of a small object of interest of a synthetic aperture sonar image, which realizes further confirmation of the small object of interest under water based on an improved AutoEncoder network; the method is a data-driven cluster analysis method, and improves the adaptability of the network to underwater small objects of interest by improving the network structure and the activation function.
In order to achieve the above object, the present invention provides a method for cluster analysis of echo data of a small object of interest in a synthetic aperture sonar image, the method comprising:
processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method;
correlating the position information of the small object of interest with the echo data to obtain the echo data of the small object of interest;
the echo data of the small target of interest is input into a pre-established and trained cluster analysis model, so that real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image is realized;
the cluster analysis model employs a modified AutoEncoder network.
Preferably, the improved AutoEncoder network includes an encoder and a decoder, wherein,
The encoder is used for converting input data into hidden layer expression through a nonlinear mapping;
The decoder is used for remapping the hidden layer expression into input data;
The encoder and decoder each comprise 3 convolutional layers and 3 fully-concatenated layers, the 3 convolutional layers of the encoder and decoder being fully-symmetrical, the 3 fully-concatenated layers of the encoder and decoder being fully-symmetrical.
Preferably, the encoder output h 1 satisfies the following equation:
h1＝σ(W1x+b1)
Where W 1 and b 1 are the weights and offsets of the encoder, respectively, x is the echo data of the small object of interest, σ is the activation function:
Wherein the bias α is a learnable parameter;
the decoder output y satisfies the following equation:
y＝σ(W2h1+b2)
Where W 2 and b 2 are the weight and bias of the decoder, respectively.
Preferably, the method further comprises a training step of a cluster analysis model, specifically comprising:
Building a training set;
and sequentially inputting training set data into the improved AutoEncoder network to perform model training, and obtaining a trained cluster analysis model when training requirements are met.
Preferably, the training set establishment includes:
Collecting original interesting small target echo data from a real underwater environment by using a synthetic aperture sonar;
cleaning and screening the collected echo data;
And randomly dividing the cleaned and screened echo data into a training set and a testing set according to a standard data set format.
On the other hand, the invention provides a synthetic aperture sonar image interesting small target echo data cluster analysis system, which comprises:
The target detection module is used for processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method;
The echo correlation module is used for correlating the position information of the small object of interest with echo data to obtain the echo data of the small object of interest; and
The analysis output module is used for inputting echo data of the small target of interest into a pre-established and trained cluster analysis model to realize real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image;
the cluster analysis model employs a modified AutoEncoder network.
Preferably, the cluster analysis model is deployed on an edge computing platform.
Compared with the prior art, the invention has the advantages that:
1. The invention combines the detection result of the synthetic aperture sonar image target with the clustering analysis result of echo data, and provides an intelligent analysis model of the underwater small object of interest, which solves the problem of low accuracy of identifying the underwater small object of interest in the existing method in a data driving mode;
2. The improved AutoEncoder network structure designed by the invention is a data-driven unsupervised neural network, good characteristic representation is obtained from original data by minimizing data reconstruction errors, samples are divided into a plurality of disjoint clusters, so that the similarity of the samples in the same cluster is higher, the similarity of the samples among different clusters is lower, and further confirmation of underwater small objects of interest is realized.
Drawings
FIG. 1 is a frame of a method and system for cluster analysis of small objects of interest under water by using a synthetic aperture sonar;
FIG. 2 is a representation of a small object of interest cluster analysis method under water based on the improvement AutoEncoder used in the present invention;
FIG. 3 is a modified AutoEncoder network architecture for use with the present invention;
FIG. 4 is a graph comparing training loss functions of the improved AutoEncoder network and AutoEncoder provided by the present invention;
FIG. 5 is a graph comparing the effects of the cluster analysis of the improved AutoEncoder network and AutoEncoder provided by the present invention, wherein FIG. 5 (a) is a graph of the effects of the cluster analysis of the improved AutoEncoder network, and FIG. 5 (b) is a graph of the effects of the cluster analysis of the AutoEncoder network;
FIG. 6 is a graph of the effect of the cluster analysis for ReLU as the network activation function for improvement AutoEncoder provided by the present invention, wherein FIG. 6 (a) is a graph of the ReLU activation function and FIG. 6 (b) is a graph of the effect of the cluster analysis;
FIG. 7 is a graph of the effect of the cluster analysis with the improved AutoEncoder network activation function CeLU, wherein FIG. 7 (a) is a graph of CeLU activation function and FIG. 7 (b) is a graph of the effect of the cluster analysis;
FIG. 8 is a graph of the effect of the cluster analysis with the improved AutoEncoder network activation function SeLU provided by the present invention, wherein FIG. 8 (a) is a graph of SeLU activation function and FIG. 8 (b) is a graph of the effect of the cluster analysis;
FIG. 9 is a graph showing the effect of the cluster analysis with PreLU as the activation function of the improved AutoEncoder-Mixed network, wherein FIG. 9 (a) is a PreLU activation function curve and FIG. 9 (b) is a graph showing the effect of the cluster analysis.
Detailed Description
The invention provides a clustering analysis method for small target echo data of interest of a synthetic aperture sonar image, which comprises the following steps:
processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method;
correlating the position information of the small object of interest with the echo data to obtain the echo data of the small object of interest;
the echo data of the small target of interest is input into a pre-established and trained cluster analysis model, so that real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image is realized;
the cluster analysis model employs a modified AutoEncoder network.
The system of the invention comprises: the system comprises an echo extraction module, a data set making module, a model training module and a platform deployment module;
the echo extraction module is used for capturing underwater small-object echo data of interest;
The data set making module is used for collecting underwater scene data, cleaning echo data of a small object of interest underwater and making an echo data set;
the model training module is used for initializing, training and testing parameters of the cluster analysis model;
the platform deployment module is used for deploying the trained cluster analysis model to the embedded platform and is used for real-time online small-object-of-interest cluster analysis tasks.
The synthetic aperture sonar image preprocessing module further comprises: the system comprises a synthetic aperture sonar submodule, a target detection submodule and an echo data association submodule;
The synthetic aperture sonar submodule is used for processing the received array element data to obtain a real-time synthetic aperture sonar image;
The target detection submodule is used for automatically detecting a small target of interest in the synthetic aperture sonar image to obtain the position information of the small target of interest;
the echo data association sub-module is used for associating the position information of the small object of interest in the SAS image with echo data to obtain accurate underwater small object echo data of interest;
Optionally, the data set making module submodule further includes: the device comprises a data acquisition sub-module, a data cleaning sub-module and an echo data set manufacturing sub-module;
The data acquisition submodule acquires the interested small target echo data of the synthetic aperture sonar image from the real environment;
the data cleaning submodule confirms and screens the effectiveness of the echo data of the small target of interest;
The echo data set making submodule randomly divides data into a training set and a testing set according to a standard data set format.
Optionally, the model training submodule further includes: the system comprises a parameter setting module, a modified AutoEncoder network module and a model test module.
The parameter setting sub-module is used for completing the parameter initialization work required by model training;
The improved AutoEncoder network sub-module is used for realizing cluster analysis of the echo data of the small target of interest;
The model test module sub-module is used for monitoring the model training state in real time.
Optionally, the platform deployment submodule further includes: a model deployment sub-module and a result output sub-module;
The model deployment sub-module is used for transplanting the trained model to an edge computing platform;
and the result output sub-module is used for displaying and outputting the small target clustering effect of interest.
The technical scheme of the invention is described in detail below with reference to the accompanying drawings and examples.
Example 1
The embodiment 1 of the invention provides a clustering analysis method for small target echo data of interest of a synthetic aperture sonar image. Including echo extraction, dataset production, model training, and platform deployment.
Collecting submarine small-object-of-interest data by using a synthetic aperture sonar, cleaning the data and generating a target attribute analysis data set; secondly, initializing training parameters, training an improved AutoEncoder model, and carrying out quality assessment on a training result; thirdly, deploying the trained target detection model and the cluster analysis model to an edge computing platform, and building a small target cluster analysis method of interest, so as to realize online detection and cluster analysis of the small target of interest under water. The general flow chart is shown in fig. 1, and the specific steps are as follows:
Step 1, intelligent analysis data set production of underwater small object of interest
Step 1-1, acquiring original interesting small target echo data from a real underwater environment by utilizing a synthetic aperture sonar acquisition sub-module;
step 1-2, cleaning and screening echo data by using a playback program;
step 1-3, a training sample set and a test sample set are established.
Step 2, model training
Step 2-1, setting up environments required by a training platform on a deep learning server, including open source software Anaconda, pytorch, torchvision and the like, and setting model training initialization parameters, including batchsize, epoch, verification_ epochs and the like;
Step 2-2, build up of an improved AutoEncoder network, shown in FIG. 2, consisting of encoder and decoder, denoted AutoEncoder-Mixed. The encoder can be seen as converting the input data into the hidden layer representation by a non-linear mapping, while the decoder is able to remap the hidden layer representation as much as possible into the input data. AutoEncoder-Mixed encoder and decoder respectively contain 3 convolution layer modules and 3 full-connection layers, and decoder and encoder's convolution layer and full-connection layer are symmetrical completely, have guaranteed that the input output dimension is the same.
AutoEncoder-Mixed encoding and decoding processes can be described as:
the coding process comprises the following steps: h 1＝σ(W1x+b1) (1)
The decoding process comprises the following steps: y=σ (W 2h1+b2) (2)
Wherein W 1、b1 is the encoder weight and offset, wherein W 2、b2 is the decoding weight and offset, σ is the activation function, and the calculation formula is:
Where the bias α is a learnable parameter.
AutoEncoder-Mixed loss function is to minimize the error of the output data y and the input data x:
And 2-3, monitoring the training process of the improved AutoEncoder-Mixed network in real time and testing the result, and stopping training when the evaluation index meets the requirement.
Step 3, platform deployment
And 3-1, building a semantic model operation environment on the edge computing platform.
And 3-2, constructing an underwater small object cluster analysis method, wherein the flow is shown in figure 3. Firstly, obtaining the position information of a small object of interest of a high-resolution large-size sonar image by using an object detection method; then, associating the image coordinate information of the small object of interest with the array element domain data, and extracting the corresponding array element domain data; and finally, utilizing an optimal AutoEncoder-Mixed encoder to predict and infer the small object of interest of the sonar image, and realizing attribute analysis of the small object of interest of the sonar image.
And 4, displaying the analysis result of the small target of interest.
Simulation experiment:
The technical effects of the invention are further described below in conjunction with simulation experiments:
In order to verify the effectiveness of the improvement AutoEncoder-Mixd and the influence of related parameters on the performance, experiment 1 is designed, and the AutoEncoder model is taken as a reference to compare and analyze the performance differences of different clustering algorithms. Experiment 2 was designed and the effect of different activation functions on AutoEncoder-Mixed performance was compared using AutoEncoder-Mixed as a study object. The experimental dataset was Gorman and Sejnowski original sonar dataset, comprising sonar reflectance data for two seafloor targets, one of which was a metal cylinder (METAL CYLINDER) and the other was a similarly shaped rock (rocks). The two targets are placed on the sandy seabed surface, then sonar chirp signals are projected from different angles, and finally sonar reflection signals are acquired. The composition of the G & S dataset is shown in table 1.
Table 1g & s dataset
Experiment 1 compares and analyzes AutoEncoder-Mixed performance difference with the classical cluster analysis method AutoEncoder, wherein the model iteration number is 500, the batch size is 64, the learning rate is 0.0005, MSE of two model training processes is recorded respectively, and the G & S data set is tested and verified by using a cluster analysis model trained 500 times. As can be seen from FIG. 4, MSE curves AutoEncoder-Mixed and AutoEncoder-Mixed all have a decreasing trend with increasing iteration times, the MSE curves converge rapidly for the first 200 epochs and for the last 300 epochs, and the MSE curves gradually stabilize. Under the same number of iterations, autoEncoder-Mixed MSE is lower than AutoEncoder MSE, especially 300 before the model iterates. Fig. 5 is a graph comparing the effects of the cluster analysis of the improved AutoEncoder network and AutoEncoder network provided by the present invention, wherein fig. 5 (a) is a graph of the effects of the cluster analysis of the improved AutoEncoder network, and fig. 5 (b) is a graph of the effects of the cluster analysis of the AutoEncoder network. As can be seen from fig. 5, both categories AutoEncoder do not achieve aggregation and are positionally scattered; both categories of object in AutoEncoder-Mixed achieve better aggregation and there are relatively clear cluster boundaries, where '0' represents rocks and '1' represents mines. Considering MSE curve and cluster analysis effect, autoEncoder-Mixed is more suitable for the small object cluster analysis task of sonar image interest than AutoEncoder.
Experiment 2 is aimed at AutoEncoder-Mixed cluster analysis model, reLU, seLU, ceLU and PReLU activation functions are selected to train and test the model, the iteration number of the model is 500, the batch size is 64, and the learning rate is equal to 0.0005. FIG. 6 is a graph of the effect of the cluster analysis for ReLU as the network activation function for improvement AutoEncoder provided by the present invention, wherein FIG. 6 (a) is a graph of the ReLU activation function and FIG. 6 (b) is a graph of the effect of the cluster analysis; FIG. 7 is a graph of the effect of the cluster analysis with the improved AutoEncoder network activation function CeLU, wherein FIG. 7 (a) is a graph of CeLU activation function and FIG. 7 (b) is a graph of the effect of the cluster analysis; FIG. 8 is a graph of the effect of the cluster analysis with the improved AutoEncoder network activation function SeLU provided by the present invention, wherein FIG. 8 (a) is a graph of SeLU activation function and FIG. 8 (b) is a graph of the effect of the cluster analysis; FIG. 9 is a graph showing the effect of the cluster analysis with PreLU as the activation function of the improved AutoEncoder-Mixed network, wherein FIG. 9 (a) is a PreLU activation function curve and FIG. 9 (b) is a graph showing the effect of the cluster analysis. It can be found from fig. 6, fig. 7, fig. 8 and fig. 9 that, under the same model iteration number, the activation functions PReLU have better clustering performance than the three activation functions ReLU, seLU, ceLU, mainly because the value of the part smaller than 0 of PReLU is obtained through training, and the method can better adapt to the requirement of a task.
AutoEncoder-Mixed is a data-driven unsupervised neural network, and by minimizing data reconstruction errors, good characteristic representations are obtained from original data, samples are divided into a plurality of disjoint clusters, and finally the similarity of the samples in the same cluster is higher, and the similarity of the samples among different clusters is lower.
Example 2
The embodiment 2 of the invention provides a small target echo data cluster analysis system of interest of a synthetic aperture sonar image, which is realized based on the method of the embodiment 1, and comprises the following steps:
The target detection module is used for processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method;
The echo correlation module is used for correlating the position information of the small object of interest with echo data to obtain the echo data of the small object of interest;
The analysis output module is used for inputting echo data of the small target of interest into a pre-established and trained cluster analysis model to realize real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image;
the cluster analysis model employs a modified AutoEncoder network.
Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention and are not limiting. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the present invention, which is intended to be covered by the appended claims.
Claims (5)
1. A synthetic aperture sonar image small target echo data cluster analysis method, the method comprising:
processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method;
correlating the position information of the small object of interest with the echo data to obtain the echo data of the small object of interest;
the echo data of the small target of interest is input into a pre-established and trained cluster analysis model, so that real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image is realized;
The cluster analysis model adopts an improved AutoEncoder network;
The improved AutoEncoder network includes an encoder and a decoder, wherein,
The encoder is used for converting input data into hidden layer expression through a nonlinear mapping;
The decoder is used for remapping the hidden layer expression into input data;
The encoder and the decoder each comprise 3 convolution layers and 3 full connection layers, the 3 convolution layers of the encoder and the decoder are completely symmetrical, and the 3 full connection layers of the encoder and the decoder are completely symmetrical;
The encoder output h 1 satisfies the following equation:
h1＝σ(W1x+b1)
Where W 1 and b 1 are the weights and offsets of the encoder, respectively, x is the echo data of the small object of interest, σ is the activation function:
Wherein the bias α is a learnable parameter;
the decoder output y satisfies the following equation:
y＝σ(W2h1+b2)
Where W 2 and b 2 are the weight and bias of the decoder, respectively.
2. The method for clustering analysis of small target echo data of interest of a synthetic aperture sonar image according to claim 1, wherein the method further comprises a training step of a cluster analysis model, and specifically comprises:
Building a training set;
and sequentially inputting training set data into the improved AutoEncoder network to perform model training, and obtaining a trained cluster analysis model when training requirements are met.
3. The method for clustering small target echo data of interest in a synthetic aperture sonar image according to claim 2, wherein the establishing of the training set comprises:
Collecting original interesting small target echo data from a real underwater environment by using a synthetic aperture sonar;
cleaning and screening the collected echo data;
And randomly dividing the cleaned and screened echo data into a training set and a testing set according to a standard data set format.
4. A system based on the synthetic aperture sonar image small target echo data cluster analysis method of claim 1, characterized in that the system comprises:
The target detection module is used for processing the received echo data of the synthetic aperture sonar to obtain a real-time synthetic aperture sonar image, and obtaining the position information of the small target of interest by adopting a target detection method;
The echo correlation module is used for correlating the position information of the small object of interest with echo data to obtain the echo data of the small object of interest; and
The analysis output module is used for inputting echo data of the small target of interest into a pre-established and trained cluster analysis model to realize real-time online attribute analysis of the small target of interest of the synthetic aperture sonar image;
the cluster analysis model employs a modified AutoEncoder network.
5. The synthetic aperture sonar image small target echo data cluster analysis system of claim 4, wherein the cluster analysis model is deployed on an edge computing platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311322405.2A CN117611856B (en) | 2023-10-12 | 2023-10-12 | Method for clustering and analyzing echo data of small target of interest in synthetic aperture sonar image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311322405.2A CN117611856B (en) | 2023-10-12 | 2023-10-12 | Method for clustering and analyzing echo data of small target of interest in synthetic aperture sonar image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117611856A CN117611856A (en) | 2024-02-27 |
| CN117611856B true CN117611856B (en) | 2024-07-09 |
Family
ID=89952233
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311322405.2A Active CN117611856B (en) | 2023-10-12 | 2023-10-12 | Method for clustering and analyzing echo data of small target of interest in synthetic aperture sonar image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117611856B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937085A (en) * | 2010-09-01 | 2011-01-05 | 华南理工大学 | Super-resolution image reconstruction-based synthetic aperture sonar imaging method |
| CN112114300A (en) * | 2020-09-14 | 2020-12-22 | 哈尔滨工程大学 | Underwater weak target detection method based on image sparse representation |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472627B (en) * | 2019-07-02 | 2022-11-08 | 五邑大学 | End-to-end SAR image recognition method, device and storage medium |
| US20220075061A1 (en) * | 2020-09-09 | 2022-03-10 | SparkCognition, Inc. | Machine-learning augmented synthetic aperture sonar |
| CN112395987B (en) * | 2020-11-18 | 2023-07-28 | 西安电子科技大学 | SAR image target detection method based on unsupervised domain adaptive CNN |
| CN114819061A (en) * | 2022-04-02 | 2022-07-29 | 南京航空航天大学 | Sparse SAR target classification method and device based on transfer learning |
| CN115661569A (en) * | 2022-10-17 | 2023-01-31 | 北京航空航天大学 | High-precision fine-grained SAR target detection method |
| CN115995040A (en) * | 2022-10-31 | 2023-04-21 | 西北工业大学 | SAR image small sample target recognition method based on multi-scale network |
-
2023
- 2023-10-12 CN CN202311322405.2A patent/CN117611856B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937085A (en) * | 2010-09-01 | 2011-01-05 | 华南理工大学 | Super-resolution image reconstruction-based synthetic aperture sonar imaging method |
| CN112114300A (en) * | 2020-09-14 | 2020-12-22 | 哈尔滨工程大学 | Underwater weak target detection method based on image sparse representation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117611856A (en) | 2024-02-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Neupane et al. | A review on deep learning-based approaches for automatic sonar target recognition | |
| CN110472627B (en) | End-to-end SAR image recognition method, device and storage medium | |
| CN106355151A (en) | Recognition method, based on deep belief network, of three-dimensional SAR images | |
| Domingos et al. | A survey of underwater acoustic data classification methods using deep learning for shoreline surveillance | |
| CN100370486C (en) | Typhoon center positioning method based on embedded type concealed Markov model and cross entropy | |
| CN101286229B (en) | Sonar image self-adapting division method based on stratified MRF | |
| Liao et al. | ARRU phase picker: Attention recurrent‐residual U‐Net for picking seismic P‐and S‐phase arrivals | |
| CN112526451A (en) | Compressed beam forming and system based on microphone array imaging | |
| Jaramillo et al. | AUV-based bed roughness mapping over a tropical reef | |
| Mohebbi-Kalkhoran et al. | Classifying humpback whale calls to song and non-song vocalizations using bag of words descriptor on acoustic data | |
| Ashraf et al. | Underwater ambient-noise removing GAN based on magnitude and phase spectra | |
| Ji et al. | Full-waveform classification and segmentation-based signal detection of single-wavelength bathymetric LiDAR | |
| CN106067165A (en) | High spectrum image denoising method based on clustering sparse random field | |
| CN117611856B (en) | Method for clustering and analyzing echo data of small target of interest in synthetic aperture sonar image | |
| Roberts et al. | Multiview, broadband acoustic classification of marine fish: a machine learning framework and comparative analysis | |
| Feng et al. | Masking Hierarchical Tokens for Underwater Acoustic Target Recognition With Self-Supervised Learning | |
| Mancusi et al. | Fish sounds: towards the evaluation of marine acoustic biodiversity through data-driven audio source separation | |
| CN117437530A (en) | Synthetic aperture sonar interest small target twin matching identification method and system | |
| CN117451055A (en) | Underwater sensor positioning method and system based on basis tracking noise reduction | |
| CN114764880B (en) | Multi-component GAN reconstructed remote sensing image scene classification method | |
| CN117173550A (en) | Method and system for detecting underwater small target of synthetic aperture sonar image | |
| Oliveira et al. | Probabilistic positioning of a mooring cable in sonar images for in-situ calibration of marine sensors | |
| CN108460773A (en) | A kind of sonar image dividing method based on biased field level set | |
| Wang et al. | Multi-Level Feature Representation Framework With Adaptive Margin Loss for Few-Shot Sonar Images Classification of AUVs. | |
| CN117173549B (en) | Multi-scale target detection method and system for synthetic aperture sonar image under complex scene |
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 |