Radio frequency signal source jam detection method, system, computer equipment and storage medium
The RF signal source of LCD TV is detected through the convolutional neural network model, which solves the problem of insufficient detection efficiency and accuracy of RF signal source lag in the existing technology, realizes real-time and accurate detection of LCD TV radio frequency circuits, and improves RF frequency Reliability of the circuit.
Patent Information
- Application Number
- CN202210864550.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-21
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2042-07-21
AI Technical Summary
The prior art lacks efficiency and accuracy when detecting the stuck phenomenon of RF signal source of LCD TVs, especially the treadmill phenomenon caused by frequency drift under different ambient climate conditions, which is difficult to effectively solve.
Convolutional neural network model combined with RF signal data sets, by obtaining RF signal source images at different sensitivity, establishing data sets and training models, capturing video frame streams in real time for detection, determining whether there is a lag area, and improving the detection of RF circuits accuracy.
Real-time and accurate detection of LCD TV radio frequency circuits is realized, detection efficiency and accuracy are improved, dependence on manual judgments is reduced, and the reliability of RF circuits is improved.
Smart Images

Figure CN115278216B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of radio frequency circuit testing technology, and in particular to a method, system, computer equipment and storage medium for detecting jamming of a radio frequency signal source. Background Art
[0002] The high-frequency tuner (tuner) of LCD TV, also known as high-frequency head, is a key component for TV to receive cable broadcast TV signals. Traditional analog TV tuners complete the process of receiving amplification, gating, frequency conversion, and image and sound demodulation in an analog manner. If there is distortion and distortion in it, the quality of the received image and accompanying sound will deteriorate. Poor-quality high-frequency heads will cause frequency drift under different environmental and climatic conditions, causing the phenomenon of running platform.
[0003] LCD TV RF signal RF performance test usually uses AMM300 RF signal generator, which is a portable all-in-one digital TV signal generator. After setting the frequency information and playing the code stream, the decibel strength of the attenuated generator output signal is tested to test the sensitivity of the RF PCB circuit, such as Figure 3a and 3b As shown, the tester manually determines whether the video bit stream is stuck at different sensitivities to verify whether the sensitivity of the digital cable TV radio frequency (RF) PCB circuit meets the factory standards. Summary of the invention
[0004] Embodiments of the present invention provide a method, system, computer device and storage medium for detecting jamming of a radio frequency signal source, aiming to improve the testing efficiency and accuracy of radio frequency circuits.
[0005] In a first aspect, an embodiment of the present invention provides a method for detecting a jam of a radio frequency signal source, comprising:
[0006] Get the TV size, select the corresponding modulation signal, and set the signal parameters;
[0007] Controlling the RF signal generator to play the code stream according to the signal parameters, and controlling the decibel intensity of the RF signal output by the attenuation generator, so as to obtain RF signal source images under different sensitivities, and using the obtained RF signal source images as sample data to establish a RF signal data set;
[0008] Constructing a convolutional neural network model, and using the radio frequency signal data set to train the convolutional neural network model;
[0009] The video frame stream played by the TV is captured in real time, and the video frame stream is detected through the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and the judgment result is used as the detection result of the RF circuit in the TV.
[0010] In a second aspect, an embodiment of the present invention provides a radio frequency signal source jam detection system, including:
[0011] The signal parameter setting module is used to obtain the TV size, select the corresponding modulation signal, and set the signal parameters;
[0012] A signal source image acquisition module, used to control the RF signal generator to play the code stream according to the signal parameters, and to control the decibel intensity of the RF signal output by the attenuation generator, so as to obtain RF signal source images under different sensitivities, and use the obtained RF signal source images as sample data to establish a RF signal data set;
[0013] A model building and training module, used to build a convolutional neural network model and train the convolutional neural network model using the radio frequency signal data set;
[0014] The detection module is used to capture the video frame stream played by the TV in real time, and detect the video frame stream through the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and use the judgment result as the detection result of the radio frequency circuit in the TV.
[0015] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for detecting a jamming of a radio frequency signal source as described in the first aspect is implemented.
[0016] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for detecting a jamming of a radio frequency signal source as described in the first aspect is implemented.
[0017] The embodiment of the present invention provides a method, system, computer device and storage medium for detecting jamming of a radio frequency signal source, the method comprising: obtaining the size of a television, selecting a corresponding modulation mode signal, and setting signal parameters; controlling the radio frequency signal generator to play a code stream according to the signal parameters, and controlling the decibel strength of the radio frequency signal output by the attenuation generator, thereby obtaining radio frequency signal source images under different sensitivities, and using the obtained radio frequency signal source images as sample data to establish a radio frequency signal data set; constructing a convolutional neural network model, and using the radio frequency signal data set to train the convolutional neural network model; capturing a video frame stream played by a television in real time, and detecting the video frame stream through the trained convolutional neural network model to determine whether there is a jamming area in the video frame stream, and using the judgment result as the detection result of the radio frequency circuit in the television. The embodiment of the present invention uses the convolutional neural network model to detect jamming of the radio frequency signal source on the video frame stream captured in real time, so as to detect the sensitivity of the radio frequency circuit in the television in real time, thereby improving the reliability of the radio frequency circuit in the television. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.
[0019] Figure 1 Schematic diagram of the process of the detection method of the embodiment of the present invention;
[0020] Figure 2 Another schematic diagram of the detection method according to an embodiment of the present invention;
[0021] Figure 3a This is the signal source screen where the attenuation strength of the RF signal generator is lower than the standard value;
[0022] Figure 3b The attenuation strength of the RF signal generator meets the test standard signal source screen;
[0023] Figure 4 It is a principle diagram of an algorithm of a detection method according to an embodiment of the present invention;
[0024] Figure 5 A convolutional neural network model diagram of the detection method according to an embodiment of the present invention;
[0025] Figure 6a , 6b , 6c, and 6d are schematic diagrams of image processing of a detection method according to an embodiment of the present invention;
[0026] Figure 7A schematic diagram of marking a jamming area in a detection method according to an embodiment of the present invention;
[0027] Figure 8a is an exemplary schematic diagram of a detection method according to an embodiment of the present invention;
[0028] Figure 8b Another exemplary schematic diagram of the detection method according to an embodiment of the present invention;
[0029] Fig. 9 4 is a schematic block diagram of a detection device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0031] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.
[0032] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0033] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0034] See below Figure 1 , Figure 1 A flowchart of a method for detecting jamming of a radio frequency signal source provided by an embodiment of the present invention specifically includes steps S101 to S104.
[0035] S101, obtaining the TV size, selecting the corresponding modulation signal, and setting the signal parameters;
[0036] S102, controlling the RF signal generator to play a code stream according to the signal parameter, and controlling the decibel intensity of the RF signal output by the attenuation generator, so as to obtain RF signal source images under different sensitivities, and using the obtained RF signal source images as sample data to establish a RF signal data set;
[0037] S103, constructing a convolutional neural network model, and using the radio frequency signal data set to train the convolutional neural network model;
[0038] S104, capturing the video frame stream played by the TV in real time, and detecting the video frame stream through the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and using the determination result as the detection result of the radio frequency circuit in the TV.
[0039] In this embodiment, the RF signal generator is used to set the frequency information, play the code stream, and attenuate the decibel strength of the RF signal output by the generator to obtain the RF signal source image at different sensitivities, thereby establishing a RF signal data set including the RF signal source image at different sensitivities. At the same time, a convolutional neural network model is constructed, and the constructed convolutional neural network model is trained and tested using the RF signal data set. The convolutional neural network model is then used to perform RF signal source jam detection on the real-time captured video frame stream, and qualified or unqualified video images are distinguished in real time, thereby realizing the sensitivity detection of the RF circuit (or RF signal source) in the TV, thereby improving the reliability of the RF circuit in the TV.
[0040] The rapid development of artificial intelligence technology has made image recognition a hot field, and computer vision methods have also entered real life from theory. Convolutional neural network (CNN) is one of the representative algorithms of deep learning, which has rich data representation learning capabilities. Unlike traditional algorithms, CNN can learn the most useful features through automatic adjustment of parameters. The trained convolutional neural network model can replace the human eye in judging the jamming of RF signal sources in RF performance testing.
[0041] This embodiment acquires images of RF signal sources at different sensitivities, establishes a data set, and uses a convolutional neural network model to recognize images to detect video images with unknown attenuation decibel intensity. The prediction accuracy is high. Since the convolutional neural network has a certain degree of invariance to geometric transformation, deformation, and illumination, the accuracy of image detection of RF signal sources can be greatly improved. In the detection process, as long as the RF signal source recognition system calls the camera module to capture the video frame stream played by the LCD TV in real time, the system imports each frame of the picture into the trained convolutional neural network model for detection. No manual visual judgment is required, and the system has a high degree of intelligence.
[0042] In one embodiment, the signal parameters include bandwidth, forward error correction coding, number of samples, maximum bit rate, frequency range, and test frequency;
[0043] The radio frequency signal data set is divided into a training set, an auxiliary training set and a test set.
[0044] In this embodiment, samples are selected according to the different sizes of LCD TVs, and modulation signals are further selected and signal parameters are set. For example, 32-inch, 43-inch and 65-inch LCD TVs are respectively used as samples, and four modulation mode signals, namely QPSK (Quadrature Phase ShiftKeying), 16QAM (Quadrature Amplitude Modulation), QAM64 (Quadrature Amplitude Modulation) and 256QAM (Quadrature Amplitude Modulation), are selected, the bandwidth is set to 8MHz, forward error correction code FEC=1 / 2, sampling number FFT Size=8K, maximum bit rate is 22.860661Mbps, the frequency range is from 474MHz to 858MHz, and the interval is 8MHz as a unit, and the test frequency is set.
[0045] Using the AMM300 RF signal generator, attenuating the decibel intensity of the generator output signal, and obtaining RF signal source images at different sensitivities, the RF signal data set was established. Sensitivity refers to the minimum received signal level that the high-frequency tuner of an LCD TV can receive under certain bit error rate performance conditions. The unit of sensitivity is decibel milliwatt dBm, that is, decibel intensity.
[0046] The receiving sensitivity of the LCD TV high-frequency tuner is usually around -77dBm. The larger the negative value, the lower the signal strength, the higher the sensitivity, and the better the signal that the RF circuit can receive. When the sensitivity is unqualified, the quality of the signal received by the high-frequency tuner will deteriorate, and then each frame of the decoded picture will be abnormal, and the picture will freeze. Figure 3a As shown in Figure 2, when the sensitivity is low, the image will be abnormal, such as Figure 3b The picture shown is the picture when the sensitivity is normal.
[0047] In a specific embodiment, the RF signal data set includes a training set and a test set. The training set includes 4830 images of RF signal sources of three sizes of LCD TVs at different sensitivities; the auxiliary training set includes 3381 images of RF signal sources of three sizes of LCD TVs at different sensitivities; the test set includes 1449 images of RF signal sources of three sizes of LCD TVs at different sensitivities.
[0048] In one embodiment, in combination Figure 4 , the step S103 comprises:
[0049] Constructing a first convolutional layer and a first pooling layer, and using the output of the first convolutional layer as the input of the first pooling layer; wherein the convolution kernel size of the first convolutional layer is 3×3, and the first pooling layer is a maximum pooling layer with a size of 3×3;
[0050] Construct a second convolutional layer and a second pooling layer, and use the output of the first pooling layer as the input of the second convolutional layer, and use the output of the second convolutional layer as the input of the second pooling layer; wherein the convolution kernel size of the second convolutional layer is 3×3, and the second pooling layer is a maximum pooling layer with a size of 3×3;
[0051] Constructing a third convolutional layer and a third pooling layer, and using the output of the second pooling layer as the input of the third convolutional layer, and using the output of the third convolutional layer as the input of the third pooling layer; wherein the convolution kernel size of the third convolutional layer is 3×3, and the third pooling layer is a maximum pooling layer with a size of 3×3;
[0052] A fully connected layer and a Sigmoid classification layer are constructed, and the output of the third pooling layer is used as the input of the fully connected layer, and the output of the fully connected layer is used as the input of the Sigmoid classification layer, so as to construct a convolutional neural network model based on 2D-CNN.
[0053] Further, combined with Figure 5 In one embodiment, the step S103 further includes:
[0054] The sample data in the radio frequency signal data set is adjusted to a size of 100×100×3 and input into the convolutional neural network model based on 2D-CNN;
[0055] The sample data of size 100×100×3 is processed sequentially through the first convolution layer of 16 3×3 convolution kernels, the first ReLU function, and the first maximum pooling layer of 3×3 to obtain the first feature map of size 50×50×16;
[0056] The first feature map of size 50×50×16 is processed by the second convolution layer of 32 3×3 convolution kernels, the second ReLU function, and the first maximum pooling layer of 3×3 in sequence to obtain a second feature map of size 25×25×32;
[0057] The second feature map of size 25×25×32 is processed sequentially through a third convolution layer of 64 3×3 convolution kernels, a third ReLU function, and a third 3×3 maximum pooling layer to obtain a third feature map of size 12×12×64;
[0058] Using a Flatten layer to perform one-dimensional processing on the third feature map to obtain a one-dimensional feature vector;
[0059] Input the feature vector into the fully connected layer, and output corresponding neurons from the first neural network layer and the second neural network layer in the fully connected layer; wherein the first neural network layer includes 136 neurons and the second neural network layer includes 16 neurons;
[0060] The neurons are classified using a Sigmoid classification layer, and the classification result is used as the output result of the convolutional neural network model based on 2D-CNN.
[0061] In this embodiment, the first convolution layer uses 16 3×3 convolution kernels to deconvolve the 100×100×3 input image (i.e., the sample data), and after the first ReLU function processing, passes through a 3×3 first maximum pooling layer to obtain a 50×50×16 feature map. The second convolution layer uses 32 3×3 convolution kernels to deconvolve the 50×50×16 first feature map output by the first maximum pooling layer, and after the second ReLU function processing, passes through a 3×3 second maximum pooling layer to obtain a 25×25×32 second feature map. The third convolution layer uses 64 3×3 convolution kernels to deconvolve the 25×25×32 feature map output by the second maximum pooling layer, and after the third ReLU function processing, passes through a 3×3 third maximum pooling layer to obtain a 12×12×64 third feature map. The three-dimensional third feature map is input to the Flatten layer (for flattening the array), and is processed into one dimension to transition to the fully connected layer. The fully connected layer includes two layers of neural networks. Based on the golden model, the first layer of neural network (i.e. Figure 5 The first hidden layer in the neural network contains 136 neurons, and the second layer of the neural network (i.e. Figure 5 The second hidden layer in the image processing method contains 16 neurons. The one-dimensional feature vector output of the Flatten layer is input into each neuron of the first hidden layer to obtain the output of 136 neurons. The output is input into each neuron of the second hidden layer to obtain the output of 16 neurons. Finally, the output of 16 neurons is input into the Sigmoid classification layer to obtain the classification result of the corresponding RF signal source jamming.
[0062] In one embodiment, the step S103 further includes:
[0063] Using the training set in the radio frequency signal data set to train the 2D-CNN based convolutional neural network model until the loss function of the 2D-CNN based convolutional neural network model converges;
[0064] Select sample data with detection errors during the training process, and continue to train the 2D-CNN based convolutional neural network model using the auxiliary training set until the detection results of the 2D-CNN based convolutional neural network model for the sample data in the test set are stable.
[0065] In this embodiment, when the loss function of the convolutional neural network model based on 2D-CNN converges, the trained convolutional neural network model based on 2D-CNN is tested on the validation set. For images with incorrect classification detection of RF signal source jamming, some images of the same type as the images with incorrect detection are found in the auxiliary training set and added to the training set. The CNN convolutional neural network is trained continuously until the detection results on the test set are stable. The training is stopped to obtain a trained 2D-CNN convolutional neural network model. The stable detection results on the test set refer to that when the 2D-CNN convolutional neural network model is tested on the test set, the classification detection accuracy of the RF signal source jamming corresponding to each frame image in each video frame stream is 100%, and the model accurately grasps the classification detection results of each frame image at more than 98%, which basically meets the requirements of production testing.
[0066] In one embodiment, the step S104 includes:
[0067] Call the camera module to capture the video frame stream played on the TV in real time;
[0068] Detecting a current frame in the video frame stream by using a trained 2D-CNN-based convolutional neural network model;
[0069] If no jamming is detected in the current frame, continue to detect the next frame;
[0070] If a jam is detected in the current frame, the jam area of the current frame is marked and displayed in real time.
[0071] In this embodiment, when using a convolutional neural network model based on 2D-CNN to detect the jamming of the RF signal source, for any frame in the video frame stream, if no jamming is detected, the next frame will be continued for detection. If a jamming is detected, the corresponding jamming area is marked on the frame for easy backtracking. In a specific application scenario, a camera or other device can be used to capture the video frame stream in real time. For example, a camera module is used to capture the image of the AMM300 RF signal generator attenuating an unknown intensity signal source in real time. The video frame stream input to the convolutional neural network model based on 2D-CNN is 30FPS. Therefore, when using the trained convolutional neural network model based on 2D-CNN to detect the jamming image of the unknown RF signal source, 30 unknown images are input per second and the size is 100×100×3.
[0072] In one embodiment, if a jam is detected in the current frame, the jam area of the current frame is marked and displayed in real time, including:
[0073] Performing image processing on the current frame with stuck image; wherein the image processing includes: grayscale processing, binarization processing, expansion processing, corrosion processing and / or closing processing;
[0074] The Canny operator is used to perform edge detection on the current frame to obtain a basic outline of the current frame, and a freeze area of the current frame is marked according to the basic outline.
[0075] In this embodiment, when marking the stuck area, firstly, the current frame with the stuck area is processed, for example, Figure 6a -d, grayscale the current frame, convert it into a grayscale image, and then perform binarization ( Figure 6a ), weighted calculation is performed on each pixel in the image according to the distance from the center point according to the Gaussian function, and the Gaussian weighted value of the local neighborhood block is calculated and the image is binarized to remove the noise interference from the outside. In morphology, a 3×3 convolution kernel is used, and the current frame is expanded with an iteration number of 2 ( Figure 6b ), and use a 4×4 convolution kernel to perform an erosion operation on the current layer with an iteration number of 2 ( Figure 6c ). Then close the current frame. After completing the image processing, use the Canny operator to perform edge detection on the current frame to obtain the corresponding basic contour, such as Figure 6d , thereby marking the stuck area, and also being able to obtain the decibel image classification result of the RF signal source to determine whether the decibel intensity of the output signal of the attenuation signal generator received by the RF circuit meets the factory specifications.
[0076] In actual use, the camera module can be used to obtain the RF signal source video frame stream played by the LCD TV in real time, so as to detect the RF signal source in real time for a 29FPS, 21S video frame stream. During the detection, each frame of the RF signal source image is subjected to image processing such as binarization, expansion, and corrosion, and the Canny operator is used for edge detection. By performing real-time dynamic recognition on each frame of the image, marking the recognition result and judging the accuracy probability of the result, when the RF signal source image is unqualified, the unqualified area, that is, the jamming area, is marked. Figure 7 shown.
[0077] In a specific application scenario, when detecting a video frame stream of 29FPS and lasting 21S, the video freezes at 3-4S, 10-11S, and 19-20S when judged by the human eye. By statistically analyzing the data detected by the detection method provided by the embodiment of the present invention, the probability graph of PASS and the probability graph of NG of the signal source detection are drawn in combination with the recognition result of each frame, as shown in FIG. Figure 8a and 8b It can be seen that when the detection method provided by the embodiment of the present invention identifies each frame of image, the 104-114 frames and the 116-127 frames (3-4S), the 290-303 frames (10-11S), and the 562-578 frames (19-20S) are all identified as unqualified images. The judgment result is consistent with the human eye judgment of the tester, and the probability of determining the unqualified image for each frame is above 98%.
[0078] In a specific embodiment, the CPU processing efficiency can reach up to 30FPS. After reading and processing the image, it can judge the accuracy on the processing screen of each frame rate and mark the stuck area. Theoretically, the frame rate of the video frame stream output after the test is completed should be 30FPS, but in fact, the system needs a certain amount of time to call the model and program for processing, about ten milliseconds, that is to say, the actual frame rate of the processed video frame stream will decrease to a certain extent, so 29FPS is the actual frame rate of the output video after the system judges the processing.
[0079] In one embodiment, if Figure 2 As shown, first, the AMM300 RF signal generator plays the code stream, and at the same time, by attenuating the decibel strength of the generator output signal, the LCD TV receives the RF signal and outputs an unknown image, and then captures the frame stream in real time through the camera module, and then uses the convolutional neural network model based on 2D-CNN to detect the real-time captured frame stream to determine whether it is qualified. For qualified images, the screen outputs PASS in real time and continues to detect the next frame. For unqualified images, image processing is performed on them, such as binarization, corrosion, expansion and closure processing, and the basic outline is obtained through edge detection, so as to mark the jammed area in real time and display NG on the screen.
[0080] Fig. 9 A schematic block diagram of a radio frequency signal source jam detection system 900 provided in an embodiment of the present invention, the system 900 includes:
[0081] The signal parameter setting module 901 is used to obtain the TV size, select the corresponding modulation signal, and set the signal parameters;
[0082] The signal source image acquisition module 902 is used to control the RF signal generator to play the code stream according to the signal parameters, and control the decibel intensity of the RF signal output by the attenuation generator, so as to obtain the RF signal source image under different sensitivities, and use the obtained RF signal source image as sample data to establish a RF signal data set;
[0083] A model building and training module 903, used to build a convolutional neural network model and train the convolutional neural network model using the radio frequency signal data set;
[0084] The detection module 904 is used to capture the video frame stream played by the TV in real time, and detect the video frame stream through the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and use the judgment result as the detection result of the RF circuit in the TV.
[0085] In one embodiment, the signal parameters include bandwidth, forward error correction coding, number of samples, maximum bit rate, frequency range, and test frequency;
[0086] The radio frequency signal data set is divided into a training set, an auxiliary training set and a test set.
[0087] In one embodiment, the model building and training module 903 includes:
[0088] A first construction unit is used to construct a first convolutional layer and a first pooling layer, and use the output of the first convolutional layer as the input of the first pooling layer; wherein the convolution kernel size of the first convolutional layer is 3×3, and the first pooling layer is a maximum pooling layer with a size of 3×3;
[0089] A second construction unit is used to construct a second convolutional layer and a second pooling layer, and use the output of the first pooling layer as the input of the second convolutional layer, and use the output of the second convolutional layer as the input of the second pooling layer; wherein the convolution kernel size of the second convolutional layer is 3×3, and the second pooling layer is a maximum pooling layer with a size of 3×3;
[0090] A third construction unit is used to construct a third convolutional layer and a third pooling layer, and use the output of the second pooling layer as the input of the third convolutional layer, and use the output of the third convolutional layer as the input of the third pooling layer; wherein the convolution kernel size of the third convolutional layer is 3×3, and the third pooling layer is a maximum pooling layer with a size of 3×3;
[0091] The fourth construction unit is used to construct a fully connected layer and a Sigmoid classification layer, and use the output of the third pooling layer as the input of the fully connected layer, and use the output of the fully connected layer as the input of the Sigmoid classification layer, so as to construct a convolutional neural network model based on 2D-CNN.
[0092] In one embodiment, the model building and training module 903 further includes:
[0093] An input adjustment unit, used for adjusting the sample data in the radio frequency signal data set to a size of 100×100×3 and inputting the sample data into the convolutional neural network model based on 2D-CNN;
[0094] A first feature map acquisition unit is used to process the sample data of a size of 100×100×3 through a first convolution layer of 16 3×3 convolution kernels, a first ReLU function, and a first pooling layer of 3×3 in sequence to obtain a first feature map of a size of 50×50×16;
[0095] A second feature map acquisition unit is used to process the first feature map of size 50×50×16 through a second convolution layer of 32 3×3 convolution kernels, a second ReLU function, and a 3×3 second pooling layer in sequence to obtain a second feature map of size 25×25×32;
[0096] A third feature map acquisition unit is used to process the second feature map of size 25×25×32 through a third convolution layer of 64 3×3 convolution kernels, a third ReLU function, and a third pooling layer of 3×3 in sequence to obtain a third feature map of size 12×12×64;
[0097] A one-dimensional processing unit, used for performing one-dimensional processing on the third feature map by using a Flatten layer to obtain a one-dimensional feature vector;
[0098] A vector input unit, used to input the feature vector into the fully connected layer, and the first layer neural network and the second layer neural network in the fully connected layer output corresponding neurons; wherein the first layer neural network includes 136 neurons, and the second layer neural network includes 16 neurons;
[0099] The neuron classification unit is used to classify the neurons using a Sigmoid classification layer, and use the classification result as the output result of the convolutional neural network model based on 2D-CNN.
[0100] In one embodiment, the model building and training module 903 further includes:
[0101] A model training unit, used to train the 2D-CNN based convolutional neural network model using a training set in the radio frequency signal data set until a loss function of the 2D-CNN based convolutional neural network model converges;
[0102] An auxiliary training unit is used to select sample data with detection errors during the training process, and continue to train the 2D-CNN based convolutional neural network model using the auxiliary training set until the detection results of the 2D-CNN based convolutional neural network model for the sample data in the test set are stable.
[0103] In one embodiment, the detection module 904 includes:
[0104] The frame stream capture unit is used to call the camera module to capture the video frame stream played by the TV in real time;
[0105] A current frame detection unit, used to detect the current frame in the video frame stream by using a trained 2D-CNN based convolutional neural network model;
[0106] A next frame detection unit, used to continue detecting the next frame if no jamming is detected in the current frame;
[0107] The jam marking unit is used to mark and display the jam area of the current frame in real time if jam is detected in the current frame.
[0108] In one embodiment, the jam marking unit includes:
[0109] A grayscale processing unit, used to perform grayscale processing on the current frame with stuck state to obtain a corresponding grayscale image;
[0110] A binarization processing unit is used to perform weighted calculation on the distance between each pixel point and the center point in the grayscale image according to a Gaussian function, calculate the Gaussian weighted value of the local neighborhood block and binarize the image to remove noise interference;
[0111] A dilation operation unit is used to perform a dilation operation on the binarized image using a 3×3 convolution kernel and multiple iterations;
[0112] An erosion processing unit, used to perform erosion processing on the image after the dilation operation by using a 3×3 convolution kernel and multiple iterations;
[0113] A closing processing unit, used for performing closing processing on the image after the corrosion processing to obtain a closed image corresponding to the current frame;
[0114] The edge detection unit is used to perform edge detection on the closed image of the current frame using the Canny operator to obtain the basic outline of the current frame, and mark the jamming area of the current frame according to the basic outline to obtain the decibel image classification result of the RF circuit, so as to determine whether the decibel intensity of the output signal of the attenuation signal generator received by the RF circuit meets the factory specifications.
[0115] Since the embodiments of the system part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the system part, which will not be repeated here.
[0116] The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed, the steps provided in the above embodiment can be implemented. The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.
[0117] The embodiment of the present invention also provides a computer device, which may include a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps provided in the above embodiment may be implemented. Of course, the computer device may also include various network interfaces, power supplies and other components.
[0118] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of this application.
[0119] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.
Claims
1. A method for detecting jamming of a radio frequency signal source, It is characterized in that include: Get the TV size, select the corresponding modulation signal, and set the signal parameters; According to the signal parameters, the RF signal generator is controlled to play the code stream, and the decibel strength of the RF signal output by the attenuation generator is controlled to obtain RF signal source images under different sensitivities, and the obtained RF signal source images are used as sample data to establish a RF signal data set; Use the RF signal generator to set the frequency information and play the code stream, and attenuate the decibel intensity of the RF signal output by the generator to obtain the RF signal source image under different sensitivities and establish the RF signal data set; Constructing a convolutional neural network model, and using the radio frequency signal data set to train the convolutional neural network model; Capturing the video frame stream played by the TV in real time, and detecting the video frame stream through the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and using the determination result as the detection result of the radio frequency circuit in the TV; The step of constructing a convolutional neural network model and training the convolutional neural network model using the radio frequency signal data set includes: Constructing a first convolutional layer and a first pooling layer, and using the output of the first convolutional layer as the input of the first pooling layer; wherein the convolution kernel size of the first convolutional layer is 3×3, and the first pooling layer is a maximum pooling layer with a size of 3×3; Construct a second convolutional layer and a second pooling layer, and use the output of the first pooling layer as the input of the second convolutional layer, and use the output of the second convolutional layer as the input of the second pooling layer; wherein the convolution kernel size of the second convolutional layer is 3×3, and the second pooling layer is a maximum pooling layer with a size of 3×3; Constructing a third convolutional layer and a third pooling layer, and using the output of the second pooling layer as the input of the third convolutional layer, and using the output of the third convolutional layer as the input of the third pooling layer; wherein the convolution kernel size of the third convolutional layer is 3×3, and the third pooling layer is a maximum pooling layer with a size of 3×3; A fully connected layer and a Sigmoid classification layer are constructed, and the output of the third pooling layer is used as the input of the fully connected layer, and the output of the fully connected layer is used as the input of the Sigmoid classification layer, so as to construct a convolutional neural network model based on 2D-CNN.
2. The method for detecting jamming of a radio frequency signal source according to claim 1, It is characterized in that The signal parameters include bandwidth, forward error correction coding, number of samples, maximum bit rate, frequency range and test frequency; The radio frequency signal data set is divided into a training set, an auxiliary training set and a test set.
3. The method for detecting jamming of a radio frequency signal source according to claim 2, It is characterized in that The step of constructing a convolutional neural network model and training the convolutional neural network model using the radio frequency signal data set further includes: The sample data in the radio frequency signal data set is adjusted to a size of 100×100×3 and input into the convolutional neural network model based on 2D-CNN; The sample data of size 100×100×3 is processed sequentially through the first convolution layer of 16 3×3 convolution kernels, the first ReLU function, and the first pooling layer of 3×3 to obtain the first feature map of size 50×50×16; The first feature map of size 50×50×16 is processed by the second convolution layer of 32 3×3 convolution kernels, the second ReLU function, and the second pooling layer of 3×3 in sequence to obtain a second feature map of size 25×25×32; The second feature map of size 25×25×32 is processed by the third convolution layer of 64 3×3 convolution kernels, the third ReLU function, and the third pooling layer of 3×3 in sequence to obtain a third feature map of size 12×12×64; Using a Flatten layer to perform one-dimensional processing on the third feature map to obtain a one-dimensional feature vector; Input the feature vector into the fully connected layer, and output corresponding neurons from the first neural network layer and the second neural network layer in the fully connected layer; wherein the first neural network layer includes 136 neurons and the second neural network layer includes 16 neurons; The neurons are classified using a Sigmoid classification layer, and the classification result is used as the output result of the convolutional neural network model based on 2D-CNN.
4. The method for detecting jamming of a radio frequency signal source according to claim 2, It is characterized in that The step of constructing a convolutional neural network model and training the convolutional neural network model using the radio frequency signal data set further includes: Using the training set in the radio frequency signal data set to train the 2D-CNN based convolutional neural network model until the loss function of the 2D-CNN based convolutional neural network model converges; Select sample data with detection errors during the training process, and continue to train the 2D-CNN based convolutional neural network model using the auxiliary training set until the detection results of the 2D-CNN based convolutional neural network model for the sample data in the test set are stable.
5. The method for detecting jamming of a radio frequency signal source according to claim 1, It is characterized in that The real-time capture of the video frame stream played by the TV and the detection of the video frame stream by the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and the determination result is used as the detection result of the radio frequency circuit in the TV, including: Call the camera module to capture the video frame stream played on the TV in real time; Detecting a current frame in the video frame stream by using a trained 2D-CNN-based convolutional neural network model; If no jamming is detected in the current frame, continue to detect the next frame; If a jam is detected in the current frame, the jam area of the current frame is marked and displayed in real time.
6. The method for detecting jamming of a radio frequency signal source according to claim 5, It is characterized in that If a jam is detected in the current frame, the jam area of the current frame is marked and displayed in real time, including: Perform grayscale processing on the current frame with stuck state to obtain a corresponding grayscale image; Performing weighted calculation on the distance between each pixel point and the center point in the grayscale image according to the Gaussian function, calculating the Gaussian weighted value of the local neighborhood block and binarizing the image to remove noise interference; The binary image is expanded using a 3×3 convolution kernel and multiple iterations; The image after the dilation operation is corroded by using a 3×3 convolution kernel and multiple iterations; Perform closing processing on the image after corrosion processing to obtain a closed image corresponding to the current frame; The Canny operator is used to perform edge detection on the closed image of the current frame to obtain the basic outline of the current frame, and the jamming area of the current frame is marked according to the basic outline to obtain the decibel image classification result of the RF circuit, so as to determine whether the decibel intensity of the output signal of the attenuation signal generator received by the RF circuit meets the factory specifications.
7. A radio frequency signal source jam detection system, It is characterized in that include: The signal parameter setting module is used to obtain the TV size, select the corresponding modulation signal, and set the signal parameters; A signal source image acquisition module is used to control the RF signal generator to play the code stream according to the signal parameters, and to control the decibel intensity of the RF signal output by the attenuation generator, so as to obtain RF signal source images under different sensitivities, and use the obtained RF signal source images as sample data to establish a RF signal data set; A model building and training module, used to build a convolutional neural network model and train the convolutional neural network model using the radio frequency signal data set; A detection module is used to capture the video frame stream played by the TV in real time, and detect the video frame stream through the trained convolutional neural network model to determine whether there is a stuck area in the video frame stream, and use the determination result as the detection result of the radio frequency circuit in the TV; The model building and training module includes: A first construction unit is used to construct a first convolutional layer and a first pooling layer, and use the output of the first convolutional layer as the input of the first pooling layer; wherein the convolution kernel size of the first convolutional layer is 3×3, and the first pooling layer is a maximum pooling layer with a size of 3×3; A second construction unit is used to construct a second convolutional layer and a second pooling layer, and use the output of the first pooling layer as the input of the second convolutional layer, and use the output of the second convolutional layer as the input of the second pooling layer; wherein the convolution kernel size of the second convolutional layer is 3×3, and the second pooling layer is a maximum pooling layer with a size of 3×3; A third construction unit is used to construct a third convolutional layer and a third pooling layer, and use the output of the second pooling layer as the input of the third convolutional layer, and use the output of the third convolutional layer as the input of the third pooling layer; wherein the convolution kernel size of the third convolutional layer is 3×3, and the third pooling layer is a maximum pooling layer with a size of 3×3; The fourth construction unit is used to construct a fully connected layer and a Sigmoid classification layer, and use the output of the third pooling layer as the input of the fully connected layer, and use the output of the fully connected layer as the input of the Sigmoid classification layer, so as to construct a convolutional neural network model based on 2D-CNN.
8. A computer device, It is characterized in that It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for detecting jamming of a radio frequency signal source as described in any one of claims 1 to 6 is implemented.
9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for detecting jamming of a radio frequency signal source according to any one of claims 1 to 6 is implemented.
