CN113030966A - Method for quickly sensing effective target in satellite-borne SAR original echo domain - Google Patents

Method for quickly sensing effective target in satellite-borne SAR original echo domain Download PDF

Info

Publication number
CN113030966A
CN113030966A CN202110182601.9A CN202110182601A CN113030966A CN 113030966 A CN113030966 A CN 113030966A CN 202110182601 A CN202110182601 A CN 202110182601A CN 113030966 A CN113030966 A CN 113030966A
Authority
CN
China
Prior art keywords
target
echo
effective
network
original echo
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
Application number
CN202110182601.9A
Other languages
Chinese (zh)
Inventor
张盼
张颖而
金仲和
皇甫江涛
黄毅
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN202110182601.9A priority Critical patent/CN113030966A/en
Publication of CN113030966A publication Critical patent/CN113030966A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/9004—SAR image acquisition techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
    • G01S13/937—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of marine craft
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/417—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section involving the use of neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Ocean & Marine Engineering (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a method for quickly sensing an effective target in an original echo domain of a satellite-borne SAR. The invention constructs an intelligent target rapid sensing network based on the AlexNet network, acquires high-quality training data to train the network, identifies whether the original echo of the input system contains an effective target, can accurately sense the target under a small calculation amount and ensures the instantaneity. The method can be used for intelligently and efficiently perceiving the effective target from the original electromagnetic echo domain directly, and the satellite-borne SAR receiver can efficiently and quickly screen and preprocess the target of interest in the echo while receiving the echo signal of the observed region, so that the consumption of the computing resource and the energy resource of the subsequent imaging of the system is reduced, the efficient utilization rate of the computing resource and the energy resource of the system is improved, and the intelligent and efficient satellite-borne SAR echo domain target perceiving system capable of perceiving the prejudice is realized.

Description

Method for quickly sensing effective target in satellite-borne SAR original echo domain
Technical Field
The invention relates to the technical field of target perception in satellite-borne SAR original radar echo data, in particular to a rapid intelligent perception method of an effective target in a satellite-borne SAR original echo domain based on a lightweight AlexNet network.
Background
With the rapid development of the space technology field, the field of microsatellites is also greatly developed, and the field of space-based remote sensing imaging has great requirements in various aspects such as military and civil use. Compared with the traditional optical imaging, the satellite-borne SAR imaging has the advantages of no limitation of illumination, no influence of weather, all-weather and all-day work and the like, and is continuously applied to more fields. Because the space of the microsatellite platform is limited, the carried load and energy of the microsatellite platform are limited, for the perception of a long-distance target, the energy received by a radar receiver is inversely proportional to the 4 th power of the distance of the microsatellite platform, meanwhile, the traditional SAR target imaging is a system of 'after-knowing' and does not have the capability of intelligent target perception, the target detection is a target identification process based on an image domain, and the process needs to consume a large amount of computing resources and energy. Therefore, if the target can be intelligently and effectively perceived in the original echo domain, the energy consumption and the computing resource consumption of the SAR satellite platform can be reduced while the target is rapidly identified.
The invention mainly carries out high-efficiency intelligent perception on the effective target in the received original electromagnetic echo data in the SAR imaging process, thereby providing more accurate positioning for the subsequent imaging area, reducing unnecessary calculation process and energy consumption and realizing a more efficient and autonomous SAR imaging system. The conventional SAR imaging process is an imaging mechanism based on a fixed mode, does not have target perception capability in the imaging signal processing process, belongs to a system of 'after-know after-sense', and performs target detection in an image domain after SAR imaging is completed, so that the whole imaging process needs to consume a large amount of computing resources and energy resources. The invention provides a novel electromagnetic target perception method for intelligently and efficiently perceiving an effective target from an original electromagnetic echo domain. Since the space of the satellite-borne SAR is limited and the only energy source is solar energy, the computing resources and energy resources on the platform are very important. The method extracts effective target information from the original echo domain, thereby improving the high-efficiency utilization rate of system computing resources and energy resources and realizing an intelligent high-efficiency satellite-borne SAR echo domain target sensing system with a 'foreknowledge and foreknowledge'.
The invention realizes intelligent perception of effective information in the echo signal by using a neural network method, thereby enabling an SAR signal processing platform to screen out accurate information while receiving the echo signal and realizing intelligent extraction of an effective target of an original echo signal in a complex environment. The invention constructs an intelligent perception network system, trains the network by utilizing positive and negative samples, and identifies whether the original echo of the input system contains an effective target. The invention intelligently senses the effective information in the original echo domain, can greatly reduce unnecessary redundant information, and accurately images the specific region, thereby enhancing the intelligent sensing capability of the system and greatly improving the efficiency of the system and the utilization rate of energy resources.
Disclosure of Invention
The invention aims to provide a method for quickly sensing an effective target in an original echo domain of a satellite-borne SAR (synthetic aperture radar), aiming at the defect of insufficient sensing capability of a satellite-borne SAR imaging system.
The invention is realized by the following technical scheme:
a method for quickly sensing an effective target in a satellite-borne SAR original echo domain comprises the following steps: aiming at a sea scene, an original echo received by a radar receiver contains an effective ship target and sea clutter invalid information, and an intelligent target perception network is designed to perceive valuable region blocks in the original echo, and the method specifically comprises the following steps:
step 1, constructing an intelligent target perception network by using an AlexNet network, wherein the intelligent target perception network comprises 5 convolutional layers, 3 full-connection layers and a Softmax layer; effective feature extraction in original echo data is achieved by convolution of the 5 convolution layers of the AlexNet network and input echo signals, convolution kernels in the convolution layers are equivalent to filters, parameters of the convolution kernels are equivalent to corresponding weights, effective separation of amplitude information, target features, target contours and clutter features in the original echo signals is achieved through convolution operation, the 3 full-connection layers are connected with a front feature extraction channel, and a target sensing result is output by a Softmax layer, so that an intelligent target sensing network is constructed;
step 2, acquiring training sample data, dividing the training sample data into a positive sample and a negative sample, and labeling each sample correspondingly;
step 3, inputting the training samples and the corresponding labels into the intelligent target perception network designed in the step 1, and training the intelligent target perception network;
and 4, inputting unknown echo data into the trained network, and judging whether the echo data contains an effective target or not through a Sigmod value.
Further, the step 1 of constructing the intelligent target-aware network specifically includes the following steps:
the target perception network structure constructed by utilizing the AlexNet network is as follows: the method comprises the steps that a first layer of convolution layer is convolution kernel with the size of 11 x 11, a second layer of convolution layer is convolution kernel with the size of 5 x 5, a third layer to a fifth layer of convolution kernel with the size of 3 x 3, the convolution layers are output and then processed through a Maxpool maximum stratification layer, and then are processed through full connection layers with the sizes of 9216, 4096 and 4096 respectively, through SoftMax classification, intelligent perception of effective targets in input data is achieved, original echo data need to be normalized firstly, the signal amplitude of the original echo data is enabled to be 0-1, and then the original echo data are input into the network.
Further, the method for acquiring training sample data comprises the following steps: using the real seaClutter data and a theoretical model are respectively formed into two batches of sample data of a positive sample and a negative sample, and a sample database is established; the positive sample generation method is as follows: 1-5 effective sparse targets are randomly distributed in the field of view of the satellite-borne SAR radar, original echo signals are generated through simulation by using coordinates of the targets, the size of an original echo matrix is 800 × 500 pixel points, the signal-to-interference ratio SJR is 0dB, the noise distribution is 0 as the mean value, and the variance is sigma2Generating 3000 samples by the Gaussian white noise, and labeling each sample correspondingly; the negative sample generation method is as follows: no effective target exists in the field range of the space-borne SAR radar, only sea clutter is randomly distributed, a sea clutter model obeys Rayleigh distribution, an original echo signal is generated by simulating a randomly generated field, the size of an original echo matrix is 800 × 500 pixel points, the signal-to-interference ratio SJR is 0dB, the noise distribution is 0 as a mean value, and the variance is sigma2The sea clutter is a Rayleigh distribution function with a Rayleigh distribution parameter sigma of 10, 3000 generated samples are provided, and corresponding labels are made for the samples.
Furthermore, before the intelligent target perception network is trained, the training samples need to be normalized, so that the signal amplitude is between 0 and 1.
Further, the perception of an effective target in the original echo is judged according to an output threshold value of the Softmax layer, if the Sigmod value is larger than 0.6, the echo data contains the interested target, and otherwise, the echo data does not contain the effective target.
Aiming at a sea scene, an original echo received by a radar receiver contains not only an effective ship target but also ineffective information such as sea clutter and the like, and the traditional target detection is required to be carried out in an image domain after SAR imaging is finished, so that a large amount of calculation and resource utilization are required, and the real-time property of the target detection is difficult to ensure; the invention realizes intelligent sensing of effective information in echo signals by using a neural network method, designs an intelligent sensing network, and intelligently senses valuable region blocks in original echoes, so that an SAR signal processing platform can screen out accurate information while receiving the echo signals, and intelligent and efficient extraction of effective targets of the original echo signals in complex environments is realized. According to the invention, an original echo intelligent target perception network system is constructed through the AlexNet which is a lightweight network, high-quality positive and negative samples are obtained by using sea clutter data and a theoretical model design to train the network, whether the original echo of an input system contains an effective target or not is identified, the target can be accurately perceived under a small calculation amount, and the instantaneity is ensured. The invention intelligently senses the effective information in the original echo domain, can greatly reduce unnecessary redundant information, and accurately images the specific region, thereby enhancing the intelligent sensing capability of the system and greatly improving the efficiency of the system and the utilization rate of energy resources.
Drawings
Fig. 1 is a schematic time domain diagram of an LFM echo signal.
Fig. 2 is a schematic diagram of the frequency domain of the LFM echo signal.
Fig. 3 is a schematic diagram of a sea clutter histogram distribution.
Fig. 4 is a schematic diagram of a positive sample/targeted training sample in a sparse target field.
Fig. 5 is a schematic diagram of negative samples/no-target training samples in a sparse target field.
Fig. 6 is a schematic diagram of a smart sensor network structure.
FIG. 7 is a graph illustrating the recognition rate of the training set and the verification set according to the change of the epoch times.
FIG. 8 is a graph showing the loss function of the training set and the validation set according to the change of the epoch times.
Fig. 9 is a schematic diagram of an intelligent perception framework for effective information in a raw echo.
Detailed Description
The technical scheme of the invention is further explained in detail in the following by combining the attached drawings.
In the embodiment, a computer simulation method is mainly adopted for verification, and all steps and conclusions are verified to be correct on MATLAB-R2019 a; the specific implementation steps are as follows:
step 1: and constructing an imaging model of the satellite-borne SAR system.
The model parameters of the spaceborne SAR system are shown in table 1.
TABLE 1 spaceborne SAR System parameters
Parameter(s) Value of
Height of track 850km
Velocity in azimuth direction 7.1km/s
Pulse repetition period 1700Hz
Center frequency of LFM signal 9.8GHz
Bandwidth of LFM signal 150MHz
Distance resolution 1m
According to the parameters of the satellite-borne SAR system in the table 1, the transmitter transmits pulse modulation signals with the bandwidth of 150MHz to an observation area in the ocean, and the expression is shown as follows.
xtr(t)＝A*exp(j2πf0t+jπμt2)
Wherein A represents the amplitude of the signal, f0Representing the carrier center frequency and mu the signal modulation slope.
The time domain diagram of the transmitted signal is shown in fig. 1 and the frequency domain diagram is shown in fig. 2.
For an offshore target scene, the existence of sea clutter is inevitable. To make the system more realistic, a sea clutter model is introduced into the electromagnetic echo to identify valid information in the original electromagnetic echo. The sea clutter histogram obeys rayleigh distribution, the amplitude spectrum histogram is shown in fig. 3, and the probability density function expression can be written as:
Figure BDA0002941838940000051
wherein z is the sea clutter variance of the ruiy distribution, and t is a time sequence; fig. 3 is a graph of a histogram distribution and an ideal rayleigh distribution function of actual IPIX sea clutter radar data. As can be seen from fig. 3, the actual sea clutter model substantially follows the rayleigh function distribution.
The radar receiver receives an echo signal after target reflection, and the expression of the echo signal is as follows:
xrec(t)＝x(t)+n(t)+fcluster(t)
wherein n (t) is 0 as a mean and σ as a variance2White Gaussian noise of (f)cluster(t) is the sea clutter echo signal, and x (t) is the effective information.
Step 2: processing a training sample:
in order to enable the system to sense whether the input echo contains a valid target or not more quickly, a sample training set needs to be constructed. A high quality training data set is crucial to the outcome of the network training. The sample training set comprises a positive sample and a negative sample, wherein the positive sample contains effective targets, and the negative sample only contains ineffective target information such as sea clutter and the like.
The specific sample generation method is as follows:
and (3) positive sample generation: 1-5 effective sparse targets are randomly distributed in the field of view of the satellite-borne SAR radar, original echo signals are generated in a simulation mode by using the coordinates of the targets, the size of an original echo matrix is 800 × 500 pixel points, the signal-to-interference ratio SJR is 0dB, and noise is dividedCloth is mean 0 and variance is σ2The number of generated samples is 3000, and a corresponding label is made for each sample, and a schematic diagram of a positive sample training sample is shown in fig. 4.
And (3) negative sample generation: effective targets do not exist in the field range of the satellite-borne SAR radar, only sea clutter distributed randomly exists, and a sea clutter model obeys Rayleigh distribution. The randomly generated field of view simulation generates original echo signals, the size of an original echo matrix is 800 × 500 pixel points, the signal-to-interference ratio SJR is 0dB, the noise distribution is an average value of 0, and the variance is sigma2The sea clutter is a rayleigh distribution function with a rayleigh distribution parameter sigma of 10. The number of generated samples is 3000, and corresponding labels are made for each sample, and a schematic diagram of a negative sample training sample is shown in fig. 5.
To this end, the generation of the training set of positive and negative samples of the original echo is completed.
And step 3: the intelligent target perception network design specifically comprises the following steps:
the first 5 convolutional layers of the AlexNet network realize effective feature extraction in original echo data by convolution with an input echo signal, convolution kernels in the convolutional layers are equivalent to filters, parameters of the convolution kernels are equivalent to corresponding weights, and effective separation of features such as amplitude information, target features, target contours and clutter in the original echo signal is realized through convolution operation. The three full-connection layers are connected with the front feature extraction channel, and the Softmax layer outputs a target sensing result to construct a target sensing network. The structure diagram of the intelligent sensing network is shown in fig. 6, wherein the first layer of convolution layer is a convolution kernel with a size of 11 × 11, the second layer of convolution layer is a convolution kernel with a size of 5 × 5, the third layer to the fifth layer of convolution layer are convolution kernels with a size of 3 × 3, the convolution layers are output and then processed by a MaxPool maximum pooling layer, and then are processed by full connection layers with sizes of 9216, 4096 and 4096 respectively, and intelligent sensing of effective targets in input data is realized by SoftMax classification.
Step 4, intelligent target perception network training, which specifically comprises the following steps:
and 4.1, classifying the Positive and Negative sample Data sets to be trained in the step 2 according to the labels, respectively placing the Positive and Negative sample Data sets into two folders of Positive _ Data and Negative _ Data, and making corresponding labels for each sample.
And 4.2, carrying out normalization processing on the sample to be trained to enable the signal amplitude to be between 0 and 1, avoiding the overfitting phenomenon in the training process, wherein the data normalization expression is shown as the following formula.
Figure BDA0002941838940000061
X is the original value in the data, Xmax、XminThe data are respectively the maximum value and the minimum value in the data, and X' is the value after the data are normalized.
And 4.3, inputting the sample to be trained with the label into the intelligent sensing network system designed in the step 3, and training a feasible target intelligent sensing network. In the training process, 50 samples are divided into a group, and batch training is performed, so that the convergence speed of the training samples is increased. The learning rate in the training process is set to be 0.001, the cross entropy is used as a loss function, 2400 samples in the positive and negative samples are used as training samples, 600 samples are used as verification samples, and meanwhile, the generalization capability of the system is improved through a cross verification method.
The graph of the effective recognition rate and the loss function of the trained system along with the change of the epoch times is shown in fig. 7 and 8. As can be seen from fig. 7 and fig. 8, after a very short 10 epoch training, the effective recognition rate of the system can be close to 99%, the sunshi function of the system is also in a convergence state, and at the same time, the effective recognition rate of the verification set is also close to 99%, and the system also rapidly reaches the convergence state.
And 5, inputting unknown radar echo data into the trained intelligent perception system after normalization processing, and judging whether the echo data contains an effective target or not through a Sigmod value. The Sigmod value represents the confidence of whether the target exists or not in the output, and generally, the Sigmod value is greater than 0.6, which indicates that the probability that the original echo contains the effective target is relatively high. Therefore, if the Sigmod value is greater than 0.6, it is determined that the echo data contains the target of interest, otherwise, the echo data does not contain a valid target, and the decision expression is shown as the following formula.
Figure BDA0002941838940000071
Fig. 9 is a schematic diagram of an intelligent perception framework for effective information in a raw echo.
By testing the system, the system can sense whether the target echo raw data contains a valid target within 8.626us after an unidentified echo raw data is input.
Meanwhile, in order to verify the high efficiency of the novel method provided by the invention, the time consumption of the method is compared and verified with that of the traditional imaging method. The traditional method does not have the capability of intelligently sensing the target, and the effective target in the image can be identified only after all imaging processing is finished. The time consumption of the intelligent target perception method was compared to the conventional imaging method by setting the ratio of the effective target ROI in the target field, and the results are shown in table 2.
TABLE 2 comparison of imaging times for two methods
Figure BDA0002941838940000072
As can be seen from table 2, compared with the conventional imaging method, the method can more efficiently and intelligently sense the effective target in the original echo, thereby increasing the imaging efficiency of the system, and having the intelligent sensing effect of "first-known first-sense".
The invention provides a novel electromagnetic target perception method for intelligently and efficiently perceiving an effective target from an original electromagnetic echo domain, which can realize that a satellite-borne SAR receiver can efficiently and quickly screen and preprocess an interested target in an echo while receiving an echo signal of an observed region, thereby reducing the consumption of computing resources and energy resources of subsequent imaging of a system. Since the space of the satellite-borne SAR is limited and the only energy source is solar energy, the computing resources and energy resources on the platform are very important. The method extracts effective target information from the original echo domain, thereby improving the high-efficiency utilization rate of system computing resources and energy resources and realizing an intelligent high-efficiency satellite-borne SAR echo domain target sensing system with a 'foreknowledge and foreknowledge'. The previous description of the disclosed examples is provided to enable any person skilled in the art to make or use the present invention. The invention has not been described in detail in part of the common general knowledge of those skilled in the art.

Claims (5)

1. A method for quickly sensing an effective target in a satellite-borne SAR original echo domain is characterized by comprising the following steps: aiming at a sea scene, an original echo received by a radar receiver contains an effective ship target and sea clutter invalid information, and an intelligent target perception network is designed to perceive valuable region blocks in the original echo, and the method specifically comprises the following steps:
step 1, constructing an intelligent target perception network by using an AlexNet network, wherein the intelligent target perception network comprises 5 convolutional layers, 3 full-connection layers and a Softmax layer; effective feature extraction in original echo data is achieved by convolution of the 5 convolution layers of the AlexNet network and input echo signals, convolution kernels in the convolution layers are equivalent to filters, parameters of the convolution kernels are equivalent to corresponding weights, effective separation of amplitude information, target features, target contours and clutter features in the original echo signals is achieved through convolution operation, the 3 full-connection layers are connected with a front feature extraction channel, and a target sensing result is output by a Softmax layer, so that an intelligent target sensing network is constructed;
step 2, acquiring training sample data, dividing the training sample data into a positive sample and a negative sample, and labeling each sample correspondingly;
step 3, inputting the training samples and the corresponding labels into the intelligent target perception network designed in the step 1, and training the intelligent target perception network;
and 4, inputting unknown echo data into the trained network, and judging whether the echo data contains an effective target or not through a Sigmod value.
2. The method for quickly sensing the effective target in the original echo domain of the spaceborne SAR according to claim 1, wherein the step 1 of constructing the intelligent target sensing network specifically comprises the following steps:
the target perception network structure constructed by utilizing the AlexNet network is as follows: the method comprises the steps that a first layer of convolution layer is convolution kernel with the size of 11 x 11, a second layer of convolution layer is convolution kernel with the size of 5 x 5, a third layer to a fifth layer of convolution kernel with the size of 3 x 3, the convolution layers are output and then processed through a Maxpool maximum stratification layer, and then are processed through full connection layers with the sizes of 9216, 4096 and 4096 respectively, through SoftMax classification, intelligent perception of effective targets in input data is achieved, original echo data need to be normalized firstly, the signal amplitude of the original echo data is enabled to be 0-1, and then the original echo data are input into the network.
3. The method for quickly sensing the effective target in the original echo domain of the spaceborne SAR according to claim 1, wherein the method for acquiring the training sample data is as follows: respectively forming two batches of sample data of a positive sample and a negative sample by using real sea clutter data and a theoretical model, and establishing a sample database; the positive sample generation method is as follows: 1-5 effective sparse targets are randomly distributed in the field of view of the satellite-borne SAR radar, original echo signals are generated through simulation by using the coordinates of the targets, the size of an original echo matrix is 800 × 500 pixel points, the signal-to-interference ratio SJR is 0dB, the noise distribution is 0 as the mean value, and the variance is
Figure RE-DEST_PATH_IMAGE002
Generating 3000 samples by the Gaussian white noise, and labeling each sample correspondingly; the negative sample generation method is as follows: no effective target exists in the field range of the space-borne SAR radar, only sea clutter is randomly distributed, a sea clutter model obeys Rayleigh distribution, an original echo signal is generated by simulating a randomly generated field, the size of an original echo matrix is 800 × 500 pixel points, the signal-to-interference ratio SJR is 0dB, the mean value of noise distribution is 0, and the variance is
Figure RE-DEST_PATH_IMAGE002A
White Gaussian noise, sea noiseThe wave is a Rayleigh distribution function with a Rayleigh distribution parameter sigma of 10, 3000 samples are generated, and corresponding labels are made for the samples.
4. The method for rapidly sensing the effective target in the original echo domain of the spaceborne SAR according to claim 1 is characterized in that a training sample needs to be normalized before an intelligent target sensing network is trained, and the signal amplitude of the training sample is 0-1.
5. The method for rapidly sensing the effective target in the original echo domain of the spaceborne SAR as claimed in claim 1, wherein the sensing of the effective target in the original echo is judged according to an output threshold value of a Softmax layer, if a Sigmod value is larger than 0.6, the echo data contains the target of interest, otherwise, the echo data does not contain the effective target.
CN202110182601.9A 2021-02-08 2021-02-08 Method for quickly sensing effective target in satellite-borne SAR original echo domain Pending CN113030966A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110182601.9A CN113030966A (en) 2021-02-08 2021-02-08 Method for quickly sensing effective target in satellite-borne SAR original echo domain

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110182601.9A CN113030966A (en) 2021-02-08 2021-02-08 Method for quickly sensing effective target in satellite-borne SAR original echo domain

Publications (1)

Publication Number Publication Date
CN113030966A true CN113030966A (en) 2021-06-25

Family

ID=76461184

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110182601.9A Pending CN113030966A (en) 2021-02-08 2021-02-08 Method for quickly sensing effective target in satellite-borne SAR original echo domain

Country Status (1)

Country Link
CN (1) CN113030966A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114049551A (en) * 2021-10-22 2022-02-15 南京航空航天大学 ResNet 18-based SAR raw data target identification method

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108229404A (en) * 2018-01-09 2018-06-29 东南大学 A kind of radar echo signal target identification method based on deep learning
CN109766811A (en) * 2018-12-31 2019-05-17 复旦大学 The end-to-end detection and recognition methods of sea ship in a kind of satellite-borne SAR image
CN110414414A (en) * 2019-07-25 2019-11-05 合肥工业大学 SAR image Ship Target discrimination method based on the fusion of multi-layer depths of features
CN111126335A (en) * 2019-12-31 2020-05-08 珠海大横琴科技发展有限公司 SAR ship identification method and system combining significance and neural network
CN111458711A (en) * 2020-04-24 2020-07-28 北京卫星信息工程研究所 Satellite-borne dual-band SAR system and detection method of ship target
US20210003697A1 (en) * 2019-07-02 2021-01-07 Wuyi University Method and apparatus for end-to-end sar image recognition, and storage medium

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108229404A (en) * 2018-01-09 2018-06-29 东南大学 A kind of radar echo signal target identification method based on deep learning
CN109766811A (en) * 2018-12-31 2019-05-17 复旦大学 The end-to-end detection and recognition methods of sea ship in a kind of satellite-borne SAR image
US20210003697A1 (en) * 2019-07-02 2021-01-07 Wuyi University Method and apparatus for end-to-end sar image recognition, and storage medium
CN110414414A (en) * 2019-07-25 2019-11-05 合肥工业大学 SAR image Ship Target discrimination method based on the fusion of multi-layer depths of features
CN111126335A (en) * 2019-12-31 2020-05-08 珠海大横琴科技发展有限公司 SAR ship identification method and system combining significance and neural network
CN111458711A (en) * 2020-04-24 2020-07-28 北京卫星信息工程研究所 Satellite-borne dual-band SAR system and detection method of ship target

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
SIZHE CHEN等: "Target Classification Using the Deep Convolutional Networks for SAR Images", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 *
任硕良: "基于卷积神经网络和迁移学习的SAR目标识别", 《中国优秀博硕士学位论文全文数据库(硕士)·信息科技辑》 *
李洁: "基于CFAR与深度学习的SAR图像舰船目标检测研究", 《中国优秀博硕士学位论文全文数据库(硕士)·信息科技辑》 *
石洪基 等: "基于改进区域卷积神经网络的SAR图像船只检测方法", 《中国海洋大学学报》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114049551A (en) * 2021-10-22 2022-02-15 南京航空航天大学 ResNet 18-based SAR raw data target identification method
CN114049551B (en) * 2021-10-22 2022-08-05 南京航空航天大学 ResNet 18-based SAR raw data target identification method

Similar Documents

Publication Publication Date Title
CN114202696A (en) SAR target detection method and device based on context vision and storage medium
CN110348288A (en) A kind of gesture identification method based on 77GHz MMW RADAR SIGNAL USING
CN107527023B (en) Polarized SAR image classification method based on superpixels and topic models
CN109241982A (en) Object detection method based on depth layer convolutional neural networks
CN113240047B (en) SAR target recognition method based on component analysis multi-scale convolutional neural network
CN103065136A (en) Method for recognizing collaborative target in SAR (Synthetic Aperture Radar) image based on visual attention mechanism
Qu et al. Human activity recognition based on WRGAN-GP-synthesized micro-doppler spectrograms
CN111008979A (en) Robust night image semantic segmentation method
Wang et al. IDUDL: Incremental double unsupervised deep learning model for marine aquaculture SAR images segmentation
CN107392140A (en) Polarization SAR terrain classification method with incremental learning is decomposed based on low-rank sparse
CN113030966A (en) Method for quickly sensing effective target in satellite-borne SAR original echo domain
Guo et al. Recognition rate versus substitution rate curve: An objective utility assessment criterion of simulated training data
CN112801928B (en) Attention mechanism-based millimeter wave radar and visual sensor fusion method
Humayun et al. YOLO-OSD: Optimized ship detection and localization in multiresolution SAR satellite images using a hybrid data-model centric approach
Meihong Vehicle detection method of automatic driving based on deep learning
CN117216517A (en) Radiation source individual identification method based on contrast generation sample enhancement
Tang et al. SAR deception jamming target recognition based on the shadow feature
CN109031298B (en) Closed-loop feedback judgment method for self-adaptive imaging of cognitive radar resources
CN115856811A (en) Micro Doppler feature target classification method based on deep learning
CN110991354A (en) ISAR image identification method based on inverse space transformation network
CN113627480B (en) Polarization SAR image classification method based on reinforcement learning
CN114049551B (en) ResNet 18-based SAR raw data target identification method
Yu et al. Convolutional Neural Networks’ Applications in Automatic Target Recognition for Synthetic Aperture Images
Guo et al. Deep Model Based Road User Classification Using mm-Wave Radar
Liao et al. LPI radar waveform recognition based on semi-supervised model all mean teacher

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210625