Full-automatic ultrasonic detection map intelligent evaluation method and device

Abstract

The invention discloses a full-automatic ultrasonic detection map intelligent evaluation method and device. The method comprises the following steps: acquiring a detection map, and reading information of each channel in the detection map; in each band-shaped graph channel, amplitude determination is carried out on each frame of data of the detection graph, a gate position with a signal exceeding a first preset threshold is determined, a defect signal is determined according to the gate position, and band-shaped graph channel defect information is calculated; in each volume channel, inputting the detection atlas into a first detection model for defect detection, determining a first defect signal in the volume channel and corresponding A scanning waveform data, and calculating volume channel defect information; in the TOFD channel, inputting the detection atlas into a second detection model for detection, determining a second defect signal in the TOFD channel and corresponding A scanning waveform data, and calculating TOFD channel defect information; and according to the defect information of the strip graph channel, the volume channel and the TOFD channel, based on a joint weighting algorithm, obtaining a final defect detection result of the detection graph.

Description

Technical Field

[0001] The invention relates to a fully automatic ultrasonic detection spectrum intelligent evaluation method and device. Background Art

[0002] In recent years, with the rapid development of pipeline engineering construction in my country, there are more and more large-diameter and high-wall thickness oil and gas pipelines, and automatic welding technology has been widely adopted. For the detection of groove unfusion defects that are more likely to occur in automatic welding, AUT technology has become the preferred detection method and has been increasingly recognized and applied in the domestic industry.

[0003] As pipeline construction develops towards digitalization and intelligence, the intelligence of AUT detection technology is particularly important. At present, the AUT detection data of long-distance oil and gas pipelines need to be manually judged, which is time-consuming and labor-intensive. Due to differences in judges, it may lead to missed evaluations. Therefore, it is urgent to develop an intelligent evaluation method for AUT detection maps, which can perform intelligent evaluation on the collected data and then be manually confirmed to improve the efficiency and accuracy of AUT detection map evaluation and realize intelligent evaluation of AUT detection results. At present, no intelligent evaluation method for AUT detection maps has been proposed at home and abroad. Summary of the invention

[0004] In order to better realize the intelligent evaluation of AUT detection images, the embodiment of the present invention provides a method and device for fully automatic intelligent evaluation of ultrasonic detection images.

[0005] In a first aspect, an embodiment of the present invention provides a fully automatic ultrasonic detection pattern intelligent evaluation method, the method comprising:

[0006] Obtaining a detection spectrum, and reading information of each channel in the detection spectrum;

[0007] In each strip chart channel, an amplitude determination is performed on each frame of the detection spectrum, a gate position where a signal exceeds a first preset threshold is determined, and whether the signal exceeding the first preset threshold is a defect signal is determined according to the gate position, and defect information of the strip chart channel is calculated respectively;

[0008] In each volume channel, the detection spectrum is input into the first detection model to perform defect detection, a first defect signal in the volume channel is determined, and the defect information of the volume channel is calculated according to the A-scan waveform data corresponding to the first defect signal;

[0009] In the TOFD channel, the detection spectrum is input into the second detection model to perform defect detection, a second defect signal in the TOFD channel is determined, and the TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal;

[0010] According to the strip chart channel defect information, volume channel defect information and TOFD channel defect information, based on a joint weighted algorithm, the final defect detection result of the detection spectrum is calculated.

[0011] In an optional implementation of the embodiment of the present invention, in each strip chart channel, an amplitude determination is performed on each frame of the detection spectrum, a gate position where a signal exceeds a first preset threshold is determined, and whether a signal exceeding the first preset threshold is a defect signal is determined according to the gate position, and strip chart channel defect information is calculated respectively, including:

[0012] In each strip chart channel, an amplitude determination is performed on each frame of data of the detection spectrum to determine a gate position where the signal exceeds a first preset threshold value;

[0013] Determining whether the gate position is within a preset range;

[0014] If yes, determining that the signal exceeding the first preset threshold is a defect signal;

[0015] According to the defect signal judgment results in each strip chart channel, the strip chart channel defect information is calculated respectively; the strip chart channel defect information includes defect height, depth and length.

[0016] In an optional implementation of the embodiment of the present invention, the first detection model includes the preset feature network, the ROI area, the fully connected layer, the classifier and the bounding box regressor; the first detection model is obtained by training in the following manner:

[0017] Inputting the pre-built training sets of each volume channel into the pre-built initial detection model for training, and obtaining defect detection results and the trained initial detection model;

[0018] Comparing the defect detection result with the actual defect result, and updating the trained initial detection model;

[0019] The above training process is repeated until the accuracy of the defect detection result meets the preset conditions, thereby obtaining the first detection model.

[0020] In an optional implementation manner of the embodiment of the present invention, in each volume channel, the detection spectrum is input into a first detection model to perform defect detection, a first defect signal in the volume channel is determined, and the volume channel defect information is calculated according to the A-scan waveform data corresponding to the first defect signal, including:

[0021] In each volume channel, the detection map is input into the first detection model, and an image feature map is generated through the preset feature network;

[0022] Input the image feature map into the ROI region pooling to obtain a feature vector of fixed length;

[0023] Inputting the fixed-length feature vector passing through the fully connected layer into the classifier for defect detection, and inputting the bounding box regressor to correct the target bounding box coordinates, and determining the first defect signal in the volume channel;

[0024] The volume channel defect information is calculated based on the A-scan waveform data corresponding to the first defect signal.

[0025] In an optional implementation of the embodiment of the present invention, after the fixed-length feature vector passing through the fully connected layer is input into the classifier for defect detection and input into the bounding box regressor to correct the target bounding box coordinates, the method further includes:

[0026] Based on the non-maximum suppression algorithm, the target boxes whose intersection-over-union ratio is greater than a second preset threshold in the output result of the first detection model are removed to obtain the defect detection result.

[0027] In an optional implementation of the embodiment of the present invention, the second detection model includes the preset feature network, the ROI area, the fully connected layer, the classifier and the bounding box regressor; the second detection model is obtained by training in the following manner:

[0028] Input the pre-built TOFD channel training set into the pre-built initial detection model for training, and obtain the defect detection results and the trained initial detection model;

[0029] Comparing the defect detection result with the actual defect result, and updating the trained initial detection model;

[0030] The above training process is repeated until the accuracy of the defect detection result meets the preset conditions, thereby obtaining the second detection model.

[0031] In an optional implementation of an embodiment of the present invention, in the TOFD channel, the detection spectrum is input into a second detection model for defect detection, a second defect signal in the TOFD channel is determined, and the TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal, including:

[0032] In the TOFD channel, the detection map is input into the second detection model, and an image feature map is generated through the preset feature network;

[0033] Input the image feature map into the ROI region pooling to obtain a feature vector of fixed length;

[0034] Inputting the fixed-length feature vector passing through the fully connected layer into the classifier for defect detection, and inputting the bounding box regressor to correct the target bounding box coordinates to determine the second defect signal in the TOFD channel;

[0035] The TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal.

[0036] In an optional implementation of the embodiment of the present invention, after the fixed-length feature vector passing through the fully connected layer is input into the classifier for defect detection and input into the bounding box regressor to correct the target bounding box coordinates, the method further includes:

[0037] Based on the non-maximum suppression algorithm, the target boxes whose intersection-over-union ratio is greater than a second preset threshold in the output result of the second detection model are removed to obtain the defect detection result.

[0038] In a second aspect, an embodiment of the present invention provides a fully automatic ultrasonic detection spectrum intelligent evaluation device, the device comprising:

[0039] A reading module, used to obtain a detection spectrum and read information of each channel in the detection spectrum;

[0040] A first determination module is used to perform amplitude determination on each frame of the detection spectrum in each strip chart channel, determine a gate position where a signal exceeds a first preset threshold, determine whether a signal exceeding the first preset threshold is a defect signal according to the gate position, and calculate defect information of the strip chart channel respectively;

[0041] A second determination module is used to input the detection spectrum into the first detection model to perform defect detection in each volume channel, determine the first defect signal in the volume channel, and calculate the volume channel defect information according to the A-scan waveform data corresponding to the first defect signal;

[0042] A third determination module is used to input the detection spectrum into a second detection model to perform defect detection in the TOFD channel, determine a second defect signal in the TOFD channel, and calculate the TOFD channel defect information according to the A-scan waveform data corresponding to the second defect signal;

[0043] The weighted calculation module is used to calculate the final defect detection result of the detection spectrum based on the strip chart channel defect information, volume channel defect information and TOFD channel defect information based on a joint weighted algorithm.

[0044] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned fully automatic ultrasonic detection image intelligent evaluation method.

[0045] In a fourth 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 fully automatic ultrasonic detection image intelligent evaluation method as described above is implemented.

[0046] The beneficial effects of the above technical solution provided by the embodiment of the present invention include at least:

[0047] The fully automatic ultrasonic detection map intelligent evaluation method provided by the embodiment of the present invention uses the first detection model and the second detection model established based on the deep recognition algorithm to respectively identify defects in the volume channel and TOFD channel of the AUT detection map, and combines the logical judgment method to identify defects in the strip chart channel. Finally, the defect detection results of the TOFD channel, the volume channel and the strip chart channel are comprehensively evaluated to obtain the final evaluation result of the detection map. This method focuses on solving the problems of low efficiency and human misjudgment of the AUT detection map evaluation, reduces missed judgments and time and manpower resource consumption, improves the efficiency and quality of the AUT detection map evaluation at the construction site, and realizes the intelligent evaluation of the AUT detection results; ensures the quality of AUT detection on the long-distance pipeline project site, gives full play to the technical advantages of intelligent detection, and accelerates the intelligent process of AUT detection technology.

[0048] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0049] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0051] Figure 1 A schematic diagram of the steps of the fully automatic ultrasonic detection pattern intelligent evaluation method provided by an embodiment of the present invention;

[0052] Figure 2 A flow chart of a fully automatic ultrasonic testing pattern intelligent evaluation method provided by an embodiment of the present invention;

[0053] Figure 3 A schematic diagram of an intelligent evaluation interface of a detection spectrum provided by an embodiment of the present invention;

[0054] Figure 4A schematic diagram of the structure of a fully automatic ultrasonic detection image intelligent evaluation device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0056] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0057] It should also be understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0058] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]", depending on the context.

[0059] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0060] References to "one embodiment" or "some embodiments" etc. described in the present specification mean that one or more embodiments of the present invention include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0061] It should be understood that the order of execution of the steps in the following embodiments does not imply a precedence of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0062] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0063] The inventors found that in the prior art, the AUT detection data of long-distance oil and gas pipelines must be manually judged, which is time-consuming and laborious. Due to differences in judges, it may lead to missed evaluations. Therefore, it is urgent to develop an intelligent evaluation method for AUT detection maps, which can perform intelligent evaluation on the collected data and then confirm it manually to improve the efficiency and accuracy of AUT detection map evaluation and realize intelligent evaluation of AUT detection results. However, no intelligent evaluation method for AUT detection maps has been proposed at home and abroad. Based on this, the inventors have made the present invention after further research and development, providing a fully automatic ultrasonic detection map intelligent evaluation method and device.

[0064] Embodiment 1

[0065] The embodiment of the present invention provides a fully automatic ultrasonic detection spectrum intelligent evaluation method, referring to Figure 1 As shown, the method includes:

[0066] S101: Acquire a detection spectrum, and read information of each channel in the detection spectrum.

[0067] In an embodiment of the present invention, an AUT test map to be tested is obtained, the number of strip chart channels and volume channels of the test map, and the data information stored in each channel are read, so that the data of each channel is read, and the strip chart channel, volume channel and TOFD channel of the test map can be intelligently judged respectively, and the wall thickness information is read to calculate the depth information of the defect.

[0068] S102: In each strip chart channel, an amplitude determination is performed on each frame of the detection spectrum to determine a gate position where a signal exceeds a first preset threshold value, and based on the gate position, it is determined whether the signal exceeding the first preset threshold value is a defect signal, and the strip chart channel defect information is calculated respectively.

[0069] In the above step S102, in each strip chart channel, the amplitude of each frame of the detection spectrum is determined, the gate position where the signal exceeds the first preset threshold is determined, and whether the signal exceeding the first preset threshold is a defect signal is determined according to the gate position, and the strip chart channel defect information is calculated respectively, which specifically includes:

[0070] In each strip chart channel, an amplitude determination is performed on each frame of data of the detection spectrum to determine a gate position where the signal exceeds a first preset threshold value;

[0071] Determining whether the gate position is within a preset range;

[0072] If yes, determining that the signal exceeding the first preset threshold is a defect signal;

[0073] According to the defect signal judgment results in each strip chart channel, the strip chart channel defect information is calculated respectively; the strip chart channel defect information includes defect height, depth and length.

[0074] In an embodiment of the present invention, when performing intelligent evaluation on the strip chart channel, firstly, the amplitude of each frame of data in each strip chart channel is determined in turn, that is, the A-scan amplitude is calculated and compared with the first preset threshold, and the signal exceeding the first preset threshold is recorded, wherein the first preset threshold can be set according to the fully automatic ultrasonic detection evaluation standard, for example, the strip chart channel evaluation gate threshold is set to 40%, that is, the first preset threshold can be set to 40%.

[0075] Then, the gate position when the signal exceeds the first preset threshold is automatically calculated. The gate position can be calculated by calculating the gate length and center position when the signal exceeds the first preset threshold, and then calculating the position where the highest wave deviates from the center. According to the calculation result of the gate position, the accuracy weight of the defect signal is given, that is, the preset range of the given gate position is determined to determine whether the signal exceeding the first preset threshold is a defect signal. If the gate position is within the preset range, the signal exceeding the first preset threshold is considered to be a defect signal; if the gate position is not within the preset range, the signal exceeding the first preset threshold is considered not to be a defect signal.

[0076] Finally, according to the different defect signal judgment results of each strip chart channel, the relevant information such as the circumferential position, amplitude, circumferential length, partition and defect type of the defect signal are output respectively, and the defect height, depth and length information are calculated.

[0077] S103: In each volume channel, the detection spectrum is input into the first detection model to perform defect detection, a first defect signal in the volume channel is determined, and the volume channel defect information is calculated according to the A-scan waveform data corresponding to the first defect signal.

[0078] In an embodiment of the present invention, the first detection model includes the preset feature network, the ROI area, the fully connected layer, the classifier and the bounding box regressor; the first detection model is obtained by training in the following manner:

[0079] Inputting the pre-built training sets of each volume channel into the pre-built initial detection model for training, and obtaining defect detection results and the trained initial detection model;

[0080] Comparing the defect detection result with the actual defect result, and updating the trained initial detection model;

[0081] The above training process is repeated until the accuracy of the defect detection result meets the preset conditions, thereby obtaining the first detection model.

[0082] In an embodiment of the present invention, a first detection model based on a deep learning recognition algorithm is established to intelligently judge the volume channel of the detection map. In the process of training the first detection model, a deep learning training data set is first established. Since an AUT detection map contains multiple volume channels, each volume channel training set is established by channel. In order to increase the data in the volume channel training set, the inventor developed a volume channel data annotation software to annotate the defect data in the AUT detection map and increase the data capacity of the volume channel training set. Among them, the training data in the volume channel training set is a typical defect image, which is a part of the image displayed by the AUT detection data. The initial detection model is trained with the typical defect image.

[0083] In an embodiment of the present invention, the pre-constructed initial detection model may be a Faster RCNN model, which includes a preset feature network, an ROI region, a fully connected layer, a classifier, and a bounding box regressor. The volume channel training set is input into the initial detection model for training, the defect detection result is output, and the trained initial detection model is obtained. The defect detection result is compared with the corresponding real defect result, the parameters of the initial detection model are optimized, and the trained initial detection model is updated. The above training process is repeated to continuously update the trained initial detection model until the accuracy of the defect detection result meets the preset conditions, and the first detection model is obtained.

[0084] In the embodiment of the present invention, in each volume channel, the detection spectrum is input into the first detection model to perform defect detection, a first defect signal in the volume channel is determined, and the volume channel defect information is calculated according to the A-scan waveform data corresponding to the first defect signal, which specifically includes:

[0085] In each volume channel, the detection map is input into the first detection model, and an image feature map is generated through the preset feature network;

[0086] Input the image feature map into the ROI region pooling to obtain a feature vector of fixed length;

[0087] Inputting the fixed-length feature vector passing through the fully connected layer into the classifier for defect detection, and inputting the bounding box regressor to correct the target bounding box coordinates, and determining the first defect signal in the volume channel;

[0088] The volume channel defect information is calculated based on the A-scan waveform data corresponding to the first defect signal.

[0089] In an embodiment of the present invention, after obtaining the trained first detection model, the first detection model can be used to perform intelligent recognition of detection maps in different volume channels. When performing defect target detection, the first detection model mainly includes four basic steps: feature extraction, candidate region generation (Region Proposal Network, RPN), region of interest pooling (Region of interest pooling, ROI pooling) and target classification and bounding box coordinate correction. First, the features of the input image are extracted through a preset feature network to generate an image feature map. Then, the feature map is output to the RPN to generate a candidate region (i.e., the ROI region). The candidate region features are unified into a feature vector of a fixed length through the ROI region pooling operation. Finally, the feature vectors after the fully connected layer are respectively input into the Softmax classifier to predict the target category and the bounding box regressor to correct the target bounding box coordinates. Among them, the preset feature network can adopt the ResNet50 feature network.

[0090] In the embodiment of the present invention, after the fixed-length feature vector passing through the fully connected layer is input into the classifier for defect detection and input into the bounding box regressor to correct the target bounding box coordinates, the method further includes:

[0091] Based on the non-maximum suppression algorithm, the target boxes whose intersection-over-union ratio is greater than a second preset threshold in the output result of the first detection model are removed to obtain the defect detection result.

[0092] In the embodiment of the present invention, in order to obtain more accurate detection results, a non-maximum suppression algorithm (NMS) can also be used to remove the target box whose intersection over union (IOU) is greater than a second preset threshold in the output result of the first detection model to obtain a defect detection result. The second preset threshold is set according to actual conditions, for example, it can be set to 0.5.

[0093] In an embodiment of the present invention, defect detection is required to be performed on the detection spectrum through the first detection model in different volume channels. After the defect detection result is obtained to identify the defect signal, the A-scan waveform data corresponding to the defect signal is called to judge the circumferential position, amplitude information and depth information of the defect, and the circumferential position, amplitude, partition and type of the circumferential position of the circumferential weld of the defect signal of the current channel are returned. Among them, the partition information can calculate the defect depth, and the circumferential position of the circumferential weld can correspond to the defect length.

[0094] S104: In the TOFD channel, the detection spectrum is input into a second detection model to perform defect detection, a second defect signal in the TOFD channel is determined, and TOFD channel defect information is calculated according to A-scan waveform data corresponding to the second defect signal.

[0095] In an embodiment of the present invention, the second detection model includes the preset feature network, the ROI area, the fully connected layer, the classifier and the bounding box regressor; the second detection model is obtained by training in the following manner:

[0096] Input the pre-built TOFD channel training set into the pre-built initial detection model for training, and obtain the defect detection results and the trained initial detection model;

[0097] Comparing the defect detection result with the actual defect result, and updating the trained initial detection model;

[0098] The above training process is repeated until the accuracy of the defect detection result meets the preset conditions, thereby obtaining the second detection model.

[0099] In an embodiment of the present invention, a second detection model based on a deep learning recognition algorithm is established to perform intelligent judgment on the TOFD channel of the detection map. In the process of training the second detection model, a deep learning training data set is first established. Since an AUT detection map contains a TOFD channel, a TOFD channel training set is established. In order to increase the data in the TOFD channel training set, the inventor developed a TOFD channel data annotation software to annotate the defect data in the AUT detection map and increase the data capacity of the TOFD channel training set. Among them, the training data in the TOFD channel training set is a typical defect image, which is a part of the AUT detection data display image, and the initial detection model is trained with the typical defect image.

[0100] In an embodiment of the present invention, the pre-constructed initial detection model may be a Faster RCNN model, which includes a preset feature network, an ROI region, a fully connected layer, a classifier, and a bounding box regressor. The TOFD channel training set is input into the initial detection model for training, the defect detection result is output, and the trained initial detection model is obtained. The defect detection result is compared with the corresponding real defect result, the parameters of the initial detection model are optimized, and the trained initial detection model is updated. The above training process is repeated to continuously update the trained initial detection model until the accuracy of the defect detection result meets the preset conditions, and a second detection model is obtained.

[0101] In an embodiment of the present invention, in the TOFD channel, the detection spectrum is input into the second detection model to perform defect detection, a second defect signal in the TOFD channel is determined, and the TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal, which specifically includes:

[0102] In the TOFD channel, the detection map is input into the second detection model, and an image feature map is generated through the preset feature network;

[0103] Input the image feature map into the ROI region pooling to obtain a feature vector of fixed length;

[0104] Inputting the fixed-length feature vector passing through the fully connected layer into the classifier for defect detection, and inputting the bounding box regressor to correct the target bounding box coordinates to determine the second defect signal in the TOFD channel;

[0105] The TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal.

[0106] In an embodiment of the present invention, after obtaining the trained second detection model, the second detection model can be used to perform intelligent recognition of the detection map in the TOFD channel. When performing defect target detection, the second detection model mainly includes 4 basic steps: feature extraction, candidate region generation (Region Proposal Network, RPN), region of interest pooling (Region of interest pooling, ROI pooling) and target classification and bounding box coordinate correction. First, the features of the input image are extracted through a preset feature network to generate an image feature map, and then the feature map is output to the RPN to generate a candidate region (i.e., ROI region). The candidate region features are unified into a feature vector of fixed length through the ROI region pooling operation. Finally, the feature vectors after the fully connected layer are respectively input into the Softmax classifier to predict the target category and the bounding box regressor to correct the target bounding box coordinates. Among them, the preset feature network can adopt the ResNet50 feature network.

[0107] In the embodiment of the present invention, after the fixed-length feature vector passing through the fully connected layer is input into the classifier for defect detection and input into the bounding box regressor to correct the target bounding box coordinates, the method further includes:

[0108] Based on the non-maximum suppression algorithm, the target boxes whose intersection-over-union ratio is greater than a second preset threshold in the output result of the second detection model are removed to obtain the defect detection result.

[0109] In the embodiment of the present invention, in order to obtain more accurate detection results, a non-maximum suppression algorithm (NMS) can also be used to remove the target box whose intersection over union (IOU) is greater than a second preset threshold in the output result of the second detection model to obtain a defect detection result. The second preset threshold is set according to actual conditions, for example, it can be set to 0.5.

[0110] In an embodiment of the present invention, defect detection is performed on the detection spectrum through the second detection model in the TOFD channel. After the defect detection result is obtained and the defect signal is identified, the corresponding A-scan data is called to judge the defect position, and the circumferential position of the ring weld, defect depth, defect type and other information of the defect signal of the current channel are returned to calculate the defect height, depth and length information.

[0111] S105: According to the strip chart channel defect information, the volume channel defect information and the TOFD channel defect information, based on a joint weighted algorithm, a final defect detection result of the detection spectrum is calculated.

[0112] In an embodiment of the present invention, after obtaining the circumferential position, signal amplitude, defect depth, defect type and other information returned by defect detection based on the strip chart channel, volume channel and TOFD channel, multi-channel joint defect recognition is performed. Specifically, the defect amplitude, defect depth and type and other information at the same circumferential position are counted, and the defect accuracy weight is calculated using a joint weighted algorithm according to the defect depth, amplitude information and defect type of different channels, so as to determine the final defect detection result of the detection spectrum. Specifically, the strip chart channel can be used as a reference to first determine the depth, height and circumferential position of the defect, and at the same time determine whether the volume channel and TOFD channel have defects at the same position. If the depth and circumferential position can correspond, then the probability of the display being a defect is very high; if only two channels have defects, the probability of defects is reduced; if only one channel has defects, the probability of defects is the lowest, and finally the final defect detection result of the detection spectrum is calculated.

[0113] The fully automatic ultrasonic detection atlas intelligent evaluation method provided by the present invention can also perform batch intelligent evaluation on existing AUT detection atlases in folders and provide batch atlas evaluation results.

[0114] In the embodiment of the present invention, Figure 2 As shown in the figure, it is a flow chart of the fully automatic ultrasonic detection atlas intelligent evaluation method. The process in the figure first reads the AUT detection data, that is, reads the detection atlas, and reads the strip chart channel, volume channel and pipe wall thickness and other information in the detection atlas, and then, respectively, performs defect identification in the strip chart channel, TOFD channel and volume channel. In the strip chart channel, the A-scan data amplitude and gate position of the detection spectrum are calculated, and the defect recognition is performed according to the calculation results of the A-scan data amplitude and gate position, and the defect signal is determined, thereby calculating the defect depth, height and length information; in the TOFD channel, a suitable deep learning model and feature extraction model are selected to determine the initial detection model, and a TOFD channel training set is constructed to train the initial detection model. After iterative training, a trained second detection model that meets the conditions is obtained. The second detection model is used to perform defect recognition in the TOFD channel and calculate the defect height, depth and length information; in each volume channel, a suitable deep learning model and feature extraction model are selected to determine the initial detection model, and a volume channel training set is constructed to train the initial detection model. After iterative training, a trained first detection model that meets the conditions is obtained. The first detection model is used to perform defect recognition in the volume channel and calculate the defect depth and length information. Finally, after obtaining the defect information of the strip chart channel, TOFD channel and volume channel, a joint weighted algorithm is used to calculate the defect accuracy weight to give the final defect recognition result, that is, the final defect detection result of the detection spectrum.

[0115] In a specific embodiment, Figure 3 As shown in the figure, it is an intelligent evaluation interface of an AUT detection map. After intelligent evaluation of each channel of the detection map, the strip chart channel, volume channel and TOFD channel with defects are automatically marked. The small square in the figure is the marked defect display. The position and depth information of these defects are used to determine whether they are the same defect, and finally the intelligent identification of defects in the AUT detection map is realized.

[0116] Embodiment 2

[0117] Based on the same inventive concept, the embodiment of the present invention also provides a fully automatic ultrasonic detection spectrum intelligent evaluation device, referring to Figure 4 As shown, the device comprises:

[0118] The reading module 101 is used to obtain the detection spectrum and read the information of each channel in the detection spectrum;

[0119] A first determination module 102 is used to perform amplitude determination on each frame of the detection spectrum in each strip chart channel, determine the gate position where the signal exceeds the first preset threshold, and determine whether the signal exceeding the first preset threshold is a defect signal according to the gate position, and calculate the defect information of the strip chart channel respectively;

[0120] The second determination module 103 is used to input the detection spectrum into the first detection model to perform defect detection in each volume channel, determine the first defect signal in the volume channel, and calculate the volume channel defect information according to the A-scan waveform data corresponding to the first defect signal;

[0121] The third determination module 104 is used to input the detection spectrum into the second detection model to perform defect detection in the TOFD channel, determine the second defect signal in the TOFD channel, and calculate the TOFD channel defect information according to the A-scan waveform data corresponding to the second defect signal;

[0122] The weighted calculation module 105 is used to calculate the final defect detection result of the detection map based on the strip chart channel defect information, the volume channel defect information and the TOFD channel defect information based on a joint weighted algorithm.

[0123] Embodiment 3

[0124] Based on the same inventive concept, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the fully automatic ultrasonic detection spectrum intelligent evaluation method described in the above embodiment 1 is implemented.

[0125] Embodiment 4

[0126] Based on the same inventive concept, an embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the fully automatic ultrasonic detection image intelligent evaluation method described in the above-mentioned embodiment 1 is implemented.

[0127] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0128] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0129] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0130] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0131] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A fully automatic ultrasonic detection pattern intelligent evaluation method, characterized in that: include: Obtaining a detection spectrum, and reading information of each channel in the detection spectrum; In each strip chart channel, an amplitude determination is performed on each frame of the detection spectrum, a gate position where a signal exceeds a first preset threshold is determined, and whether the signal exceeding the first preset threshold is a defect signal is determined according to the gate position, and defect information of the strip chart channel is calculated respectively; In each volume channel, the detection spectrum is input into the first detection model to perform defect detection, a first defect signal in the volume channel is determined, and the defect information of the volume channel is calculated according to the A-scan waveform data corresponding to the first defect signal; In the TOFD channel, the detection spectrum is input into the second detection model to perform defect detection, a second defect signal in the TOFD channel is determined, and the TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal; According to the strip chart channel defect information, volume channel defect information and TOFD channel defect information, based on a joint weighted algorithm, the final defect detection result of the detection spectrum is calculated.

2. The method according to claim 1, characterized in that In each strip chart channel, the amplitude of each frame of the detection spectrum is determined, the gate position where the signal exceeds the first preset threshold is determined, and whether the signal exceeding the first preset threshold is a defect signal is determined according to the gate position, and the defect information of the strip chart channel is calculated respectively, including: In each strip chart channel, an amplitude determination is performed on each frame of data of the detection spectrum to determine a gate position where the signal exceeds a first preset threshold value; Determining whether the gate position is within a preset range; If yes, determining that the signal exceeding the first preset threshold is a defect signal; According to the defect signal judgment results in each strip chart channel, the strip chart channel defect information is calculated respectively; the strip chart channel defect information includes defect height, depth and length.

3. The method according to claim 1, characterized in that The first detection model includes the preset feature network, the ROI area, the fully connected layer, the classifier and the bounding box regressor; the first detection model is obtained by training in the following manner: Inputting the pre-built training sets of each volume channel into the pre-built initial detection model for training, and obtaining defect detection results and the trained initial detection model; Comparing the defect detection result with the actual defect result, and updating the trained initial detection model; The above training process is repeated until the accuracy of the defect detection result meets the preset conditions, thereby obtaining the first detection model.

4. The method according to claim 3, characterized in that In each volume channel, the detection spectrum is input into the first detection model to perform defect detection, a first defect signal in the volume channel is determined, and the volume channel defect information is calculated according to the A-scan waveform data corresponding to the first defect signal, including: In each volume channel, the detection map is input into the first detection model, and an image feature map is generated through the preset feature network; Input the image feature map into the ROI region pooling to obtain a feature vector of fixed length; Inputting the fixed-length feature vector passing through the fully connected layer into the classifier for defect detection, and inputting the bounding box regressor to correct the target bounding box coordinates, and determining the first defect signal in the volume channel; The volume channel defect information is calculated based on the A-scan waveform data corresponding to the first defect signal.

5. The method according to claim 4, characterized in that After the fixed-length feature vector passing through the fully connected layer is input into the classifier for defect detection and input into the bounding box regressor to correct the target bounding box coordinates, the method further includes: Based on the non-maximum suppression algorithm, the target boxes whose intersection-over-union ratio is greater than a second preset threshold in the output result of the first detection model are removed to obtain the defect detection result.

6. The method according to claim 1, characterized in that The second detection model includes the preset feature network, the ROI area, the fully connected layer, the classifier and the bounding box regressor; the second detection model is obtained by training in the following manner: Input the pre-built TOFD channel training set into the pre-built initial detection model for training, and obtain the defect detection results and the trained initial detection model; Comparing the defect detection result with the actual defect result, and updating the trained initial detection model; The above training process is repeated until the accuracy of the defect detection result meets the preset conditions, thereby obtaining the second detection model.

7. The method according to claim 6, characterized in that In the TOFD channel, the detection spectrum is input into the second detection model to perform defect detection, a second defect signal in the TOFD channel is determined, and the TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal, including: In the TOFD channel, the detection map is input into the second detection model, and an image feature map is generated through the preset feature network; Input the image feature map into the ROI region pooling to obtain a feature vector of fixed length; Inputting the fixed-length feature vector passing through the fully connected layer into the classifier for defect detection, and inputting the bounding box regressor to correct the target bounding box coordinates to determine the second defect signal in the TOFD channel; The TOFD channel defect information is calculated according to the A-scan waveform data corresponding to the second defect signal.

8. The method according to claim 7, characterized in that After the fixed-length feature vector passing through the fully connected layer is input into the classifier for defect detection and input into the bounding box regressor to correct the target bounding box coordinates, the method further includes: Based on the non-maximum suppression algorithm, the target boxes whose intersection-over-union ratio is greater than a second preset threshold in the output result of the second detection model are removed to obtain the defect detection result.

9. A fully automatic ultrasonic detection pattern intelligent evaluation device, characterized in that: include: A reading module, used to obtain a detection spectrum and read information of each channel in the detection spectrum; A first determination module is used to perform amplitude determination on each frame of the detection spectrum in each strip chart channel, determine a gate position where a signal exceeds a first preset threshold, determine whether a signal exceeding the first preset threshold is a defect signal according to the gate position, and calculate defect information of the strip chart channel respectively; A second determination module is used to input the detection spectrum into the first detection model to perform defect detection in each volume channel, determine the first defect signal in the volume channel, and calculate the volume channel defect information according to the A-scan waveform data corresponding to the first defect signal; A third determination module is used to input the detection spectrum into a second detection model to perform defect detection in the TOFD channel, determine a second defect signal in the TOFD channel, and calculate the TOFD channel defect information according to the A-scan waveform data corresponding to the second defect signal; The weighted calculation module is used to calculate the final defect detection result of the detection spectrum based on the strip chart channel defect information, volume channel defect information and TOFD channel defect information based on a joint weighted algorithm.

10. A computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium. When the instructions are executed on a terminal, the terminal executes the fully automatic ultrasonic detection pattern intelligent evaluation method according to any one of claims 1 to 8.

11. A computer device, characterized in that: It includes a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the fully automatic ultrasonic detection atlas intelligent evaluation method as described in any one of claims 1 to 8.

Images