Method for detecting spinning box, detection device, electronic device, storage medium, and program

Abstract

A spinning box detection method, a detection apparatus, an electronic device, a storage medium, and a program are provided. [Solution] The method includes, when it is determined that the spinning box satisfies a preset defect detection condition, acquiring a detection target image of the spinning box, inputting the detection target image into a target detection model, and obtaining a target detection result of the spinning box, wherein the target detection model is used to detect whether or not there is a defect in the spinning box and obtain a target detection result, and the target detection result includes at least one of a total number of defects, defect locations, and defect types.

Description

[Technical field]

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a spinning box detection method, a detection apparatus, an electronic device, a storage medium and a program. [Background technology]

[0002] In the chemical fiber industry, in the process of using a spinning box to produce chemical fiber yarn, since the spinning box operates for a long time, it is easy for abnormalities to occur in the spinning box, such as floating yarn, yarn breakage, wrong yarn, frayed yarn, tilted oil nozzle, tilted yarn guide hook, yarn not entering the yarn guide hook, etc. Therefore, operators usually need to inspect the spinneret, yarn, yarn guide hook, oil nozzle, etc. in the spinning box. However, this inspection method is highly dependent on human experience and has low efficiency, which affects the production quality and production efficiency of yarn. Summary of the Invention [Problem to be solved by the invention]

[0003] The present disclosure provides a spinning box detection method, a detection apparatus, an electronic device, a storage medium, and a program to solve or alleviate one or more technical problems in the prior art. [Means for solving the problem]

[0004] In a first aspect, the present disclosure provides a method for detecting a spinning box, the method comprising: When it is determined that the spinning box satisfies a preset defect detection condition, acquiring a detection target image of the spinning box; inputting the detection target image into a target detection model to obtain a target detection result of the spinning box, the target detection model being used to detect whether there is a defect in the spinning box and obtain a target detection result, the target detection result including at least one of a total defect number, a defect location, and a defect type.

[0005] In a second aspect, the present disclosure provides a detection device for a spinning box, the device comprising: an acquisition unit for acquiring a detection target image of the spinning box when it is determined that the spinning box satisfies a preset defect detection condition; a detection unit for inputting the detection target image into a target detection model to obtain a target detection result of the spinning box, the target detection model being used to detect whether there is a defect in the spinning box and obtain a target detection result, the target detection result including at least one of a total defect number, a defect location, and a defect type.

[0006] In a third aspect, the present disclosure provides an electronic device, the device comprising: At least one processor; a memory in communication with the at least one processor; The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, cause the implementation of any one of the methods in the embodiments of the present disclosure.

[0007] In a fourth aspect, there is provided a non-transitory computer-readable storage medium having stored thereon computer instructions for causing a computer to perform any one of the methods in the embodiments of the present disclosure.

[0008] In a fifth aspect, a program is provided, which, when executed by a processor, implements any one of the methods in the embodiments of the present disclosure. Effect of the Invention

[0009] The beneficial effects of the technical solution provided by the present disclosure at least include: a target detection model can be used to perform defect detection on the spinning box to obtain a target detection result, thus improving the accuracy of defect detection on the spinning box; compared with the traditional manual inspection, the solution of the present disclosure does not need to rely on human experience to perform defect detection on the spinning box, realizing an automated process of defect detection on the spinning box, saving a large amount of labor and time costs, and thereby improving the detection efficiency of the spinning box.

[0010] It should be understood that the contents described herein are not intended to describe key points or key features of the embodiments of the present disclosure, nor are they used to limit the scope of the present disclosure. Other features of the present disclosure are facilitated by the following specification. [Brief description of the drawings]

[0011] [Figure 1] 1 is a schematic flow chart of a method for detecting a spinning box according to an embodiment of the present disclosure; [Figure 2(a)] FIG. 2 is a schematic diagram illustrating a configuration of a first network layer according to an embodiment of the present disclosure. [Figure 2(b)] FIG. 2 is a schematic diagram illustrating a configuration of an example of a first network layer according to an embodiment of the present disclosure. [Figure 3(a)] FIG. 2 is a schematic diagram illustrating a configuration of a first sub-network layer according to an embodiment of the present disclosure. [Figure 3(b)] FIG. 2 is a schematic diagram illustrating a configuration of an example of a second network layer according to an embodiment of the present disclosure. [Figure 3(c)] FIG. 13 is a schematic diagram illustrating a configuration of another example of a second sub-network layer according to an embodiment of the present disclosure. [Figure 3(d)] FIG. 2 is a schematic diagram illustrating a configuration of a feature weight module according to an embodiment of the present disclosure. [Figure 3(e)] FIG. 1 is a schematic diagram showing the configuration of a multi-view attention module according to one embodiment of the present disclosure. [Figure 4]2 is a schematic flow chart of a spinning box detection method according to an embodiment of the present disclosure; [Diagram 5] FIG. 2 is a schematic diagram showing the configuration of a detection device for a spinning box according to an embodiment of the present disclosure. [Figure 6] FIG. 1 is a block diagram of an electronic device for implementing a spinning box detection method according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the several accompanying drawings indicate the same or similar components or elements. The accompanying drawings are not necessarily drawn to scale. It should be understood that the drawings depict only some examples provided by the present disclosure and should not be considered as limiting the scope of the present disclosure.

[0013] The present disclosure will now be described in more detail with reference to the accompanying drawings, in which like reference numbers represent the same or similar elements, and in which various aspects of the embodiments are illustrated, and which, unless otherwise noted, are not necessarily drawn to scale.

[0014] Furthermore, in order to better explain the present disclosure, many specific details are described in the following specific examples. Those skilled in the art should understand that the present disclosure can be similarly implemented without some details. In some examples, methods, means, components, circuits, etc. that are well known to those skilled in the art are not described in detail so that the gist of the present disclosure is clear.

[0015] During the spinning process of the spinning box, because the spinning box operates for a long time, it is easy to have abnormalities such as floating yarn, yarn breakage, wrong yarn, loose yarn, tilt of oil nozzle, tilt of yarn guide hook, yarn not entering into yarn guide hook, etc. in the spinning box. Generally, workers need to inspect the spinneret, yarn, yarn guide hook, oil nozzle, etc. in the spinning box. However, this inspection method is highly dependent on human experience, and the defects such as floating yarn and loose yarn in the spinning box are relatively small and difficult to distinguish, so the manual inspection method has problems such as low efficiency and easy to make mistakes, which further affects the production quality and production efficiency of yarn. Therefore, an automated processing process is needed to efficiently detect defects in the spinning box.

[0016] Based on this, the present disclosure provides a spinning box detection method to solve the above problems.

[0017] Specifically, Figure 1 is a schematic flow chart of a method for detecting a spinning box according to an embodiment of the present disclosure, which can be selectively applied to electronic devices such as personal computers, servers, server clusters, and the like.

[0018] In addition, the method includes at least part of the following contents: As shown in FIG.

[0019] In step S101, when it is determined that the spinning box satisfies the preset defect detection condition, a detection target image of the spinning box is obtained.

[0020] Here, in one embodiment, the detection target image of the spinning box is an image collected by an image collecting device for the spinning box. The image collecting device can specifically include a yarn path patrol inspection robot (built-in device having an image collecting function such as a camera), for example, the yarn path patrol inspection robot is used to take a picture of the spinning box to obtain a detection target image, or a video is collected for a preset time length for the spinning box to obtain a plurality of consecutive video frames, and one image is selected and extracted from the consecutive video frames as a detection target image, thus facilitating the subsequent defect detection of the spinning box.

[0021] Further, in one embodiment, the detection target image includes image content of at least a spinneret, a yarn, a yarn guide hook, and an oil nozzle.

[0022] Furthermore, in one specific example, a detection target image can be obtained by the following method, specifically, when it is determined that the spinning box satisfies a preset defect detection condition, obtaining a detection target image of the spinning box (for example, the above step S101) specifically includes the following:

[0023] When it is determined that any one of the following conditions is satisfied, the detection target image of the spinning box is acquired.

[0024] Condition 1: The detection time of the spinning box reaches a preset detection timing.

[0025] Condition 2: It is determined that a winding package to be downgraded is present among the multiple winding packages obtained based on the spinning box.

[0026] Condition 3: It is determined that the number of wound yarn packages to be downgraded among the multiple wound yarn packages obtained based on the spinning box is greater than a first threshold value.

[0027] Here, the first threshold value is an empirical value, and can be set based on actual demand in practical applications, and the present disclosure does not specifically limit the first threshold value.

[0028] In this way, the solution of the present disclosure can trigger defect detection for the spinning box when at least one of the above-mentioned conditions is met (for example, trigger the image collection device to collect images for the spinning box, and further obtain the detection target image of the spinning box) to obtain the detection result of defect detection, thus quickly knowing the current defect situation of the spinning box, and subsequently laying the foundation for improving the spinning quality in the spinning box.

[0029] In step S102, the detection target image is input into a target detection model to obtain a target detection result of the spinning box.

[0030] Wherein, the target detection model is used for detecting whether there is a defect in the spinning box, and obtaining a target detection result, the target detection result including at least one of a total defect number, a defect position, and a defect type, so that the defect data in the spinning box can be obtained more accurately than the conventional manual detection, and the efficiency of defect detection can be further improved.

[0031] Further, in one embodiment, detecting defects in the spinning box may specifically refer to detecting defects such as floating yarns, yarn breakage, wrong yarns, inclination of the yarn guide hook, inclination of the oil nozzle, and foreign objects in the yarn path.

[0032] Further, in one embodiment, the main types of yarns according to the solution of the present disclosure may include one or more of partially oriented yarns (POY), fully drawn yarns (FDY), etc. For example, the types of wound yarns may specifically include polyester partially oriented yarns, polyester fully drawn yarns, polyester drawn yarns, polyester textured yarns, etc.

[0033] In this way, the present disclosure can use a target detection model to perform defect detection on the spinning box and obtain a target detection result, thus improving the accuracy of defect detection on the spinning box; and compared with the traditional manual detection, the present disclosure can perform defect detection on the spinning box without relying on human experience, realizing an automated process of defect detection on the spinning box, saving a large amount of labor costs and time costs, and thereby improving the detection efficiency of the spinning box.

[0034] Furthermore, since automated and accurate defect detection can be performed on the spinning box, measures can then be taken in a timely manner to generate presentation information, warning information, etc. to prompt operators or robots to repair the spinning box based on the obtained detection result (e.g., target detection result), thus providing support for further improving the spinning production quality and production efficiency.

[0035] In one example, the target detection model may specifically be a Dynamic Weights-Based Wavelet Attention Neural Network (DWWA-Net), or may be other models with other target detection capabilities, and the present disclosure is not limited thereto.

[0036] Furthermore, in one example, the target detection model may be a dynamic weight-based wavelet attention neural network, or other models obtained by improving based on Dynamic Wavelet Convolution Networks (DWCNet), and the present disclosure is not limited thereto.

[0037] Moreover, in one embodiment, the target detection model includes at least a first network layer, a second network layer, and a third network layer.

[0038] Specifically, the first network layer is used to filter out background noise from the input target image to obtain a low-level feature map, for example, the first network layer is a dynamic wavelet convolution network.Furthermore, as shown in FIG. 2(a), the first network layer includes at least an initial convolution module, a weight distribution module, a wavelet convolution module and a feature fusion module, the initial convolution module is used to at least perform wavelet transform on the target image to obtain a first low-frequency feature map, the weight distribution module is used to obtain a target weight coefficient of the feature map of the input image, the wavelet convolution module is used to perform wavelet transform on the target image to obtain a global feature map representing low-frequency components and high-frequency components, and the feature fusion module is used to perform fusion processing of the first low-frequency feature map and the global feature map according to the target weight coefficient to filter out background noise and obtain a low-level feature map.

[0039] Furthermore, the second network layer is used to perform feature enhancement processing on the low-level feature map to obtain a high-level feature map. For example, in one example, the second network layer can be specifically used to obtain a high-level feature map by enhancing key feature information in the low-level feature map and suppressing irrelevant feature information in the low-level feature map, such as suppressing feature information irrelevant to the yarn package.

[0040] Furthermore, the third network layer is used to perform defect recognition based on the high-level feature map and obtain target detection results.

[0041] In this way, the solution disclosed herein provides a concrete example of a model for rapid defect detection in the spinning box, thus realizing efficient detection of defects in the spinning box, and further improving the accuracy of spinning box defect detection, especially for weak defects (e.g., floating yarns, loose yarns, etc.) that are difficult to detect in the spinning box, and laying the foundation for subsequently improving the spinning production quality and production efficiency.

[0042] Furthermore, in one embodiment, the initial convolution module is specifically used to convolute the detection target image to obtain a first convolution feature map, and then wavelet transform the obtained first convolution feature map to obtain a first low-frequency feature map, for example, in one example, a convolution process is performed on the detection target image to obtain a feature map after the convolution process (for example, a first convolution feature map), then wavelet decomposition is performed on the feature map after the convolution process to obtain a plurality of feature maps, and finally, the obtained plurality of feature maps are input to a low-pass filter to perform a convolution process of the low-pass filter to obtain a first low-frequency feature map. In this way, a feature map for obtaining low noise can be effectively extracted, which lays the foundation for improving the accuracy of defect detection thereafter.

[0043] Furthermore, in one embodiment, the weight distribution module is used to convolute the detection target image to obtain a second convolution feature map, wavelet transform the obtained second convolution feature map to obtain a second low-frequency feature map, and nonlinearly transform the obtained second low-frequency feature map to obtain a target weight coefficient. In this way, the solution of the present disclosure can fully utilize the importance of the low-noise low-frequency feature map in the original image (e.g., the detection target image) to adjust the background noise in the detection target image, and can easily identify effective feature information, especially for minute defects (e.g., floating threads, loose threads), which lays a foundation for improving the accuracy of defect detection.

[0044] Further, in one embodiment, the feature fusion module filters out background noise, element-wise processes (e.g., element-wise multiplies) the first low-frequency feature map and the target weighting factor to obtain a low-level feature map, element-wise processes the global feature map and the target weighting factor, and element-wise adds the results obtained from the element-wise processing, thereby extracting useful feature information and improving the detection accuracy of yarn defects in the spinning box.

[0045] In this way, the feature map fusion module can effectively combine the extracted local feature information (e.g., the first low-frequency feature map) and the global feature information (e.g., the global feature map) to obtain a high-quality feature map, and thus, it is easier to better collect effective feature information of defects, further improve the accuracy of defect detection, and subsequently lay the foundation for improving the production quality and production efficiency of spinning.

[0046] For example, as shown in FIG. 2(b), the first network layer is a DWCNet, which includes three branches, in which an image to be detected is input to the convolution layer via branch 1 (i.e., the circled number 1 in the figure) for convolution processing, the result of which is input to the wavelet neural network to obtain a first low-frequency feature map, an image to be detected is input to the convolution layer via branch 2 (i.e., the circled number 2 in the figure) for convolution processing, the result of which is input to the wavelet neural network to obtain a second low-frequency feature map, and the second low-frequency feature map is input to the convolution layer via branch 3 (i.e., the circled number 4 in the figure) for convolution processing, the result of which is input to the wavelet neural network to obtain a second low-frequency feature map. The local feature information (e.g., the first low-frequency feature map) is multiplied by the first low-frequency feature map, and the global feature map is multiplied by the first low-frequency feature map. The results obtained after elemental processing are added element by element to filter out background noise and obtain a low-level feature map. In this way, by combining the extracted local feature information (e.g., the first low-frequency feature map) and global feature information (e.g., the global feature map), a high-quality feature map can be obtained, which further facilitates the collection of image feature information, further improves the accuracy of defect inspection, and lays the foundation for subsequently improving the production quality and production efficiency of spinning.

[0047] In another example of the present disclosure, the second network layer includes at least a first sub-network layer, a second sub-network layer, and a third sub-network layer.

[0048] Here, the first sub-network layer is used for performing feature extraction on the low-level feature map, and fusing the extracted feature maps to obtain M initial fusion feature maps, where M is an integer equal to or greater than 2; the second sub-network layer is used for performing key feature extraction on each initial fusion feature map in the M initial fusion feature maps, and performing feature enhancement on the key feature information extracted by each initial fusion feature map to obtain M target enhanced feature maps; and the third sub-network layer is used for fusing the obtained M target enhanced feature maps to obtain a high-level feature map. In this way, clear and abundant feature information (such as a high-level feature map) can be obtained, and the accuracy of defect detection can be further improved, especially for weak defects (such as floating yarns, loose yarns, etc.) that are difficult to detect in the spinning box, which significantly improves the accuracy of weak defect detection, and lays the foundation for improving the production quality and production efficiency of spinning.

[0049] Furthermore, in one example, as shown in FIG. 3(a), the j-th initial fused feature map of the M initial fused feature maps is obtained based on the following method.

[0050] A convolution process is performed on the j-th layer feature map, and feature extraction is performed to obtain a j+1-th layer feature map, where when j is 1, the 1st layer feature map is obtained by performing a convolution process on the low-level feature map, thereby obtaining M layer-level feature maps; If the value of j is M, the feature map of the Mth layer is directly set as the first initial fusion feature map. When the value of j is an integer between 1 and M-1, the jth layer feature map and the j-1th initial fused feature map are fused to obtain the jth initial fused feature map.

[0051] For example, the first sub-network layer is a feature pyramid network (FPN) layer, the second sub-network layer is a feature enhancement network layer, and the third sub-network layer is a linear addition (Add) processing layer. As shown in FIG. 3(b), first, the low-level feature map is input to the FPN layer to obtain M initial fusion feature maps. Then, each of the obtained initial fusion feature maps is input to the feature enhancement network layer, for example, the feature enhancement network layer performs key feature extraction for each initial fusion feature map, and performs feature enhancement processing for the extracted feature map to obtain M target enhancement feature maps, and finally, the obtained M target enhancement feature maps are input to the linear addition processing layer to obtain a high-level feature map. In this way, it is helpful to improve the accuracy of defect and weak defect detection, and lays the foundation for improving the production quality and production efficiency of spinning.

[0052] Here, in one example, the i-th target enhanced feature map among the M target enhanced feature maps is obtained based on the following method.

[0053] A convolution process is performed on the i-th initial fusion feature map to obtain an i-th weighting factor based on the result of the convolution process, a fusion process is performed on the i+1-th initial fusion feature map and the i-th initial fusion feature map based on the i-th weighting factor to obtain an i-th target fusion feature map, key feature extraction is performed on the obtained i-th target fusion feature map, feature enhancement process is performed on each of the extracted feature maps to obtain multiple i-th initial enhanced feature maps, and the multiple i-th initial enhanced feature maps are fused to obtain the i-th target enhanced feature map, where i is an integer between 1 and M-1.

[0054] Furthermore, when i is M, the M+1th initial fusion feature map may be a preset value, in which case the Mth target fusion feature map is obtained, and then the Mth target enhanced feature map is obtained. Alternatively, to obtain the Mth target enhanced feature map, the following may be done:

[0055] Key feature extraction is performed on the Mth initial fusion feature map, and an enhancement process is performed on each of the extracted feature maps to obtain a plurality of Mth initial enhanced feature maps, and a fusion process is performed on the plurality of Mth initial enhanced feature maps to obtain an Mth target enhanced feature map.

[0056] For example, still take the second sub-network layer as an example of a feature enhancement network layer. In this case, the feature enhancement network layer can further specifically include a feature weight module and a multi-view attention module. As shown in FIG. 3(c), the i-th initial fusion feature map and the i+1-th initial fusion feature map among the M initial fusion feature maps are input into the feature weight module to obtain the i-th target fusion feature map, and the obtained i-th target fusion feature map is input into the multi-view attention module to perform key feature extraction and feature enhancement processing to obtain the i-th target enhanced feature map; and for the M-th initial fusion feature map, the M-th initial fusion feature map is directly used as the M-th target fusion feature map, and the obtained M-th target fusion feature map is input into the multi-view attention module to perform key feature extraction and feature enhancement processing to obtain the M-th target enhanced feature map.

[0057] Furthermore, as shown in Fig. 3(d), the feature weight module first performs upsampling processing on the (i+1)th initial fusion feature map to obtain the processed (i+1)th initial fusion feature map, performs convolution processing on the i-th initial fusion feature map, and performs S-shaped bending function (Sigmod) processing on the result obtained after processing to obtain the i-th weight coefficient (e.g., λ in the figure). i , or 1-λ i ), and then the i-th weight coefficient, e.g., 1-λ i and perform element-by-element multiplication on the i-th initial fused feature map, λ iThen, perform element-by-element multiplication on the i+1th initial fusion feature map processed by the above, and perform element-by-element addition on the obtained result to obtain the ith target fusion feature map. Here, if i is M, the Mth initial fusion feature map does not need to be input to the feature weight module, and can be directly used as the Mth target fusion feature map.

[0058] Furthermore, as shown in FIG. 3(e), the multi-view attention module first performs a deformable convolution process of the first preset convolution kernel (e.g., 1×1) on the i-th target fusion feature map to extract key features, to obtain a key feature feature map 1; then performs a deformable convolution process of the second preset convolution kernel (e.g., 3×3) on the i-th target fusion feature map to obtain a key feature feature map 2; and performs a deformable convolution process of the third preset convolution kernel (e.g., 5×5) on the i-th target fusion feature map to obtain a key feature feature map 3; then inputs the obtained feature map 1 into the attention submodule A to perform feature enhancement process on the feature map 1 to obtain an i-th initial enhanced feature map 1; similarly, inputs the feature map 2 into the attention submodule B to obtain an i-th initial enhanced feature map 2; inputs the feature map 3 into the attention submodule C to obtain an i-th initial enhanced feature map 3; and finally performs a fusion process (e.g., element addition) of the above three i-th initial enhanced feature maps to obtain an i-th target enhanced feature map.

[0059] In addition to what is shown in FIG. 3(e), a neural network can be added, for example a wavelet neural network, and the wavelet neural network can be used to directly perform feature extraction on the i-th target fusion feature map, to obtain the i-th supplementary feature map, and then the fusion process can be performed together with the above three i-th initial enhanced feature maps.

[0060] In addition, the structure of the attention submodule (e.g., attention submodule A, or attention submodule B, or attention submodule C) included in the multi-view attention module may be a Squeeze and Excitation (SE) network, or other network structures obtained based on an SE network, and the present disclosure is not limited thereto.

[0061] In this way, the present disclosure can make full use of the second sub-network layer to extract key features in the low-level feature map, and then perform feature enhancement processing on the extracted results, thereby directing the network's attention to the potential target (i.e., the defect in the spinning box) and helping to better recognize the defect in the subsequent process, especially for the yarn and background plate in the spinning box whose colors are difficult to distinguish, the key features can be obtained more comprehensively, and the defect detection accuracy can be further improved, laying the foundation for improving the spinning production quality and production efficiency in the future.

[0062] In a further example of the present disclosure, the third network layer includes a fourth sub-network layer and a fifth sub-network layer.

[0063] Here, the fourth sub-network layer is used for selecting candidate regions of false defects in the image based on the high-level feature map, and the fifth sub-network layer is used for recognizing whether there is a defect in the candidate region based on the high-level feature map and the candidate region selected by the frame. In this way, the defects that may exist in the winding package can be efficiently recognized, which then lays a foundation for improving the production quality and production efficiency of the spinning.

[0064] 4 is a schematic flow chart of a method for detecting a spinning box according to an embodiment of the present disclosure. This method can be selectively applied to electronic devices such as personal computers, servers, server clusters, etc. The relevant content of the method shown in FIG. 1 above can also be applied to this embodiment, and the relevant content will not be repeated in this embodiment.

[0065] In addition, the method includes at least part of the following contents: As shown in FIG.

[0066] In step S401, when it is determined that the spinning box satisfies the preset defect detection condition, a detection target image of the spinning box is obtained.

[0067] In step S402, the detection target image is input into a target detection model to obtain a target detection result of the spinning box.

[0068] Here, the target detection model is used to detect whether there is a defect in the spinning box and obtain a target detection result, the target detection result including at least one of a total defect number, a defect location, and a defect type.

[0069] In step S403, judge whether the target detection result of the spinning box meets the preset defect requirement, if yes, proceed to step S405, if not, proceed to step S404.

[0070] In step S404, if it is determined that the target detection result of the spinning box does not satisfy the preset defect requirement, anomaly display information corresponding to the target detection result is generated.

[0071] Here, the abnormality indication information is used to indicate that the spinning box has a defect.

[0072] Furthermore, in one specific example, abnormality display information for prompting an operator or a robot to perform repair processing of the spinning box can be generated at the following timing, thereby further ensuring the production quality and production efficiency of spinning. Specifically, when it is determined that the target detection result of the spinning box does not satisfy the preset defect requirement, generating abnormality display information corresponding to the target detection result (for example, the above step S404) can be specifically: When it is determined that at least one of the following conditions is satisfied, anomaly presentation information corresponding to the target detection result is generated.

[0073] The total number of defects included in the target detection result is greater than a second threshold value; The defect location included in the target detection result is within a preset range of defect locations (e.g., the total defect size is within the preset range of defect locations and exceeds a preset area); and The defect types included in the target detection results are within a preset range of defect types (for example, if the target defect types include defect types included in all of the first detection results, at least one defect type is within the preset range of defect types, in which case the target defect type can be considered to be within the preset range of defect types).

[0074] Here, the second threshold is an empirical value, which can be set based on actual demand in practical applications, and the solution of the present disclosure does not impose any specific limitation on the second threshold.

[0075] In step S405, the flow ends.

[0076] In this way, the present disclosure can generate corresponding presentation information based on the obtained defect detection result when the preset defect requirements are not met, and in this way instruct the operator or robot to take corresponding measures for the spinning box, such as performing repair processing on the spinning box, so that the spinning production quality can reach the desired requirements, and thus ensure the production efficiency of the workshop.

[0077] In view of the above, the solution of the present disclosure has several advantages over the prior art, specifically including:

[0078] First, it realizes automatic processing: Compared with the traditional manual method, the method disclosed herein can efficiently detect defects in the spinning box without relying on human experience, thus realizing an automatic processing process, thus saving a large amount of labor and time costs, and further improving the efficiency of defect detection.

[0079] Second, improve the accuracy of inspection: the traditional manual inspection method is difficult to detect relatively small defects such as floating yarns and loose yarns in the spinning box, and is prone to overlooking the inspection, while the solution disclosed in the present disclosure uses a neural network model suitable for defect detection, which has a higher detection accuracy, especially for weak defects that are difficult to detect, and thus ensures the quality of the wound yarn package produced by the spinning box.

[0080] Third, improve production efficiency. The solution disclosed herein can perform automated and accurate defect detection on the spinning box, and can immediately take corresponding measures according to the obtained detection result (for example, the target detection result), such as generating suggestion information or warning information to suggest the operator or robot to carry out repair processing on the spinning box, thus further ensuring the normal production of spinning in the workshop.

[0081] The solution of the present disclosure also provides a detection device for the spinning box, as shown in FIG. 5, which comprises: an acquisition unit 501 for acquiring a detection target image of the spinning box when it is determined that the spinning box satisfies a preset defect detection condition; a detection unit 502 for inputting the detection target image into a target detection model to obtain a target detection result of the spinning box, the target detection model being used to detect whether there is a defect in the spinning box and obtain a target detection result, the target detection result including at least one of a total defect number, a defect location, and a defect type.

[0082] In one embodiment of the solution of the present disclosure, the acquisition unit specifically includes: the detection time of the spinning box reaches a preset detection timing; determining that a winding package to be downgraded is present among the plurality of winding packages obtained based on the spinning box; If it is determined that any one of the following conditions is met: the number of wound yarn packages to be downgraded among the multiple wound yarn packages obtained based on the spinning box is greater than a first threshold value, the detection target image of the spinning box is obtained.

[0083] In one embodiment of the solution of the present disclosure, the detection device of the spinning box comprises: The device further includes an abnormality display unit for generating abnormality display information corresponding to the target detection result when it is determined that the target detection result of the spinning box does not meet the preset defect requirements, and the abnormality display information is used to display the degree of the defect of the spinning box.

[0084] In one embodiment of the solution of the present disclosure, the anomaly presentation unit specifically includes: the total number of defects included in the target detection result is greater than a second threshold; the defect position included in the target detection result is within a preset range of defect positions; If it is determined that at least one of the following conditions is satisfied, the defect type included in the target detection result is within a predetermined range of defect types, the defect type is used to generate abnormality presentation information corresponding to the target detection result.

[0085] In one embodiment of the solution of the present disclosure, the target detection model includes at least a first network layer, a second network layer, and a third network layer; The first network layer is used to filter out background noise from an input detection target image to obtain a low-level feature map, and the first network layer includes at least an initial convolution module, a weight distribution module, a wavelet convolution module, and a feature fusion module, the initial convolution module is used to at least perform wavelet transform on the detection target image to obtain a first low-frequency feature map, the weight distribution module is used to obtain a target weight coefficient of the feature map of the input image, the wavelet convolution module is used to perform wavelet transform on the detection target image to obtain a global feature map representing low-frequency components and high-frequency components, and the feature fusion module is used to filter out background noise and fuse the first low-frequency feature map and the global feature map according to the target weight coefficient to obtain a low-level feature map, The second network layer is used for performing feature enhancement processing on the low-level feature map to obtain a high-level feature map; The third network layer is used to perform defect recognition based on the high-level feature map and obtain target detection results.

[0086] In one embodiment of the solution of the present disclosure, the initial convolution module is used to convolve the detection target image to obtain a first convolution feature map, and then wavelet transform the obtained first convolution feature map to obtain a first low-frequency feature map.

[0087] In one embodiment of the solution of the present disclosure, the weight distribution module is used for convolution processing the detection target image to obtain a second convolution feature map, wavelet transforming the obtained second convolution feature map to obtain a second low-frequency feature map, and nonlinearly transforming the obtained second low-frequency feature map to obtain a target weight coefficient.

[0088] In one embodiment of the solution of the present disclosure, the feature fusion module is used for filtering out background noise, processing the first low-frequency feature map and the target weighting factor element-wise to obtain a low-level feature map, processing the global feature map and the target weighting factor element-wise, and adding the results obtained from the element-wise processing element-wise.

[0089] The specific functions and exemplary descriptions of each module and sub-module of the apparatus according to the embodiments of the present disclosure may refer to the relevant descriptions of the corresponding steps in the above-mentioned method embodiments, and will not be repeated here.

[0090] In the technical solution disclosed herein, the acquisition, storage, and application of users' personal information comply with the provisions of relevant laws and regulations and do not violate public order and morals.

[0091] FIG. 6 is a structural block diagram of an electronic device according to an embodiment of the present disclosure. As shown in FIG. 6, the electronic device includes a memory 610 and a processor 620, and the memory 610 stores a computer program executable by the processor 620. The number of the memory 610 and the processor 620 may be one or more. The memory 610 may store one or more computer programs, which, when executed by the electronic device, cause the electronic device to perform the method provided by the above method embodiment. The electronic device may further include: The communication interface 630 is used for communicating with an external device and performing data interaction and transmission.

[0092] When the memory 610, the processor 620, and the communication interface 630 are implemented independently, the memory 610, the processor 620, and the communication interface 630 can be connected to each other via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be classified into an address bus, a data bus, a control bus, etc. For ease of explanation, only one thick line is shown in FIG. 6, but this does not indicate only one bus or only one type of bus.

[0093] Optionally, in a specific implementation, when the memory 610, the processor 620, and the communication interface 630 are integrated on one chip, the memory 610, the processor 620, and the communication interface 630 can communicate with each other via an internal interface.

[0094] It should be understood that the processor may be a Central Processing Unit (CPU) or may be other general purpose processors, Digital Signal Processing (DSP), Application Specific Integrated Circuits (ASIC), Field Programmable Gate Arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware assemblies, etc. The general purpose processor may be a microprocessor or any conventional processor, etc. Additionally, the processor may be a processor supporting the Advanced RISC Machines (ARM) architecture.

[0095] Additionally, optionally, the memory may include read-only memory and random access memory, and may further include non-volatile random access memory. The memory may be either volatile or non-volatile memory, or may include both volatile and non-volatile memory. Here, non-volatile memory may include Read-Only Memory (ROM), Programmable ROM (PROM), Erasable PROM (EPROM), Electrically EPROM (EEPROM), or flash memory. Volatile memory may include Random Access Memory (RAM), which acts as an external cache. By way of example and not limitation, many forms of RAM are available. For example, static random access memory (Static RAM, SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (Enhanced SDRAM, ESDRAM), synchlink DRAM (Synchlink DRAM, SLDRAM), and direct RAMBUS RAM (DR RAM).

[0096] The above-mentioned embodiments may be implemented in whole or in part in software, hardware, firmware, or any combination thereof. When implemented in software, the whole or in part may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, a process or function according to the embodiments of the present disclosure is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center via wire (e.g., coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, Bluetooth, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device including a server, a data center, etc. integrated with one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a Digital Versatile Disc (DVD)), or a semiconductor medium (e.g., a Solid State Disk (SSD)). Note that the computer-readable storage medium referred to in this disclosure may be a non-volatile storage medium, in other words, a non-transitory storage medium.

[0097] Those skilled in the art can understand that all or part of the steps for realizing the above embodiments may be implemented by hardware, or may be implemented by instructing relevant hardware by a program, and the program may be stored in a computer-readable storage medium, and the storage medium may be a read-only memory, a magnetic disk, an optical disk, etc.

[0098] In describing the embodiments of the present disclosure, the reference words "one embodiment", "several embodiments", "examples", "particular examples", or "several examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present disclosure. And, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. Furthermore, a person skilled in the art may combine different embodiments or examples and features of different embodiments or examples described in the present disclosure to the extent that they are not mutually inconsistent.

[0099] In the description of the embodiments of the present disclosure, unless otherwise stated, " / " means "or", for example, A / B can mean either A or B. "And / or" in the present disclosure is only for describing the related relationship of related objects, and indicates that three types of relationships may exist, for example, A and / or B can indicate the following three types of situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0100] In the description of the embodiments of the present disclosure, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying a relative importance, nor should they be construed as implying a number of technical features shown. Thus, features defined as "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present disclosure, "plurality" means two or more, unless otherwise specified.

[0101] The above are merely illustrative examples of the present disclosure, and are not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure should be included within the scope of protection of the present disclosure.

Claims

1. A method for detecting a spinning box, comprising the steps of: When it is determined that the spinning box satisfies a preset defect detection condition, a detection target image of the spinning box is acquired, and the detection target image includes at least image contents of a spinneret, a yarn, a yarn guide hook, and an oil nozzle in the spinning box; inputting the detection target image into a target detection model to obtain a target detection result of the spinning box, the target detection model is used to detect whether there is a defect in the spinning box to obtain a target detection result, the target detection result including at least one of a total defect number, a defect location, and a defect type; the target detection model includes at least a first network layer, a second network layer, and a third network layer; The first network layer is used to filter out background noise from an input detection target image to obtain a low-level feature map, and the first network layer includes at least an initial convolution module, a weight distribution module, a wavelet convolution module, and a feature fusion module, the initial convolution module is used to at least perform a wavelet transform on the detection target image to obtain a first low-frequency feature map, the weight distribution module is used to obtain a target weight coefficient of the feature map of the input image, the wavelet convolution module is used to perform a wavelet transform on the detection target image to obtain a global feature map representing low-frequency components and high-frequency components, and the feature fusion module is used to perform a fusion process between the first low-frequency feature map and the global feature map according to the target weight coefficient to filter out background noise and obtain a low-level feature map, The second network layer is used for performing feature enhancement processing on the low-level feature map to obtain a high-level feature map; the third network layer is used for performing defect recognition based on the high-level feature map and obtaining a target detection result; How to detect spinning box.

2. When it is determined that the spinning box satisfies a preset defect detection condition, acquiring a detection target image of the spinning box, the detection time of the spinning box reaches a preset detection timing; determining that a winding package to be downgraded is present among the plurality of winding packages obtained based on the spinning box; acquiring a detection target image of the spinning box when it is determined that any one of the conditions is satisfied. The method of claim 1 .

3. The method for detecting the spinning box comprises: When it is determined that the target detection result of the spinning box does not meet the preset defect requirement, generating abnormality display information corresponding to the target detection result, the abnormality display information is used to display the degree of the defect of the spinning box; The method of claim 1 .

4. When it is determined that the target detection result of the spinning box does not satisfy a preset defect requirement, generating abnormality display information corresponding to the target detection result, the total number of defects included in the target detection result is greater than a second threshold; the defect position included in the target detection result is within a preset range of defect positions; a defect type included in the target detection result is within a preset range of defect types. The method of claim 3, wherein the spinning box is a casing.

5. The initial convolution module is used for convolution processing the detection target image to obtain a first convolution feature map, and performing wavelet transform on the obtained first convolution feature map to obtain a first low-frequency feature map; The method of claim 1 .

6. The weight distribution module is used for convolution processing the detection target image to obtain a second convolution feature map, wavelet transforming the obtained second convolution feature map to obtain a second low-frequency feature map, and nonlinearly transforming the obtained second low-frequency feature map to obtain a target weight coefficient; The method of claim 1 .

7. the feature fusion module is used for filtering out background noise, element-wise processing the first low-frequency feature map and the target weighting factor to obtain a low-level feature map, element-wise processing the global feature map and the target weighting factor, and element-wise summing the results obtained from the element-wise processing; The method of claim 1 .

8. A detection device for a spinning box, comprising: an acquisition unit for acquiring a detection target image of the spinning box when it is determined that the spinning box satisfies a preset defect detection condition, the detection target image including at least image contents of a spinneret, a yarn, a yarn guide hook, and an oil nozzle in the spinning box; a detection unit for inputting the detection target image into a target detection model to obtain a target detection result of the spinning box, the target detection model is used to detect whether the spinning box has defects and obtain a target detection result, the target detection result including at least one of a total defect number, a defect position, and a defect type; the target detection model includes at least a first network layer, a second network layer, and a third network layer; The first network layer is used to filter out background noise from an input detection target image to obtain a low-level feature map, and the first network layer includes at least an initial convolution module, a weight distribution module, a wavelet convolution module, and a feature fusion module, the initial convolution module is used to at least perform a wavelet transform on the detection target image to obtain a first low-frequency feature map, the weight distribution module is used to obtain a target weight coefficient of the feature map of the input image, the wavelet convolution module is used to perform a wavelet transform on the detection target image to obtain a global feature map representing low-frequency components and high-frequency components, and the feature fusion module is used to perform a fusion process between the first low-frequency feature map and the global feature map according to the target weight coefficient to filter out background noise and obtain a low-level feature map, The second network layer is used for performing feature enhancement processing on the low-level feature map to obtain a high-level feature map; the third network layer is used for performing defect recognition based on the high-level feature map and obtaining a target detection result; Detection device for spinning box.

9. At least one processor; a memory in communication with the at least one processor; The memory stores instructions executable by the at least one processor, the instructions, when executed by the at least one processor, causing the at least one processor to perform a method according to any one of claims 1 to 7. Electronic devices.

10. A non-transitory computer readable storage medium having stored thereon instructions for causing a computer to perform the method of any one of claims 1 to 7.

11. A program for implementing the method according to any one of claims 1 to 7 when executed by a processor in a computer.

Images