Waterproof adhesive tape detection method, device and apparatus

Abstract

The application provides a waterproof tape detection method, device and equipment, the method comprising: acquiring an image of a camera wiring position; inputting the image into a detection model, a detection module of the detection model outputting a detection result of waterproof tape at the wiring position, and an evaluation module of the detection model outputting a winding integrity evaluation result of the waterproof tape. The method of the application can not only output the detection result of the waterproof tape at the wiring position, realize accurate positioning of the waterproof tape, but also synchronously output the winding integrity evaluation result of the waterproof tape by inputting the acquired image of the camera wiring position into the detection model, thereby effectively breaking through the limitation of related technologies that can only detect the existence of a target, solving the problems of low efficiency, strong subjectivity and difficulty in quantitative tracing of traditional manual detection, realizing closed-loop management and control of waterproof tape installation quality, and effectively guaranteeing the waterproof reliability of outdoor cameras and other electronic devices.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, and equipment for detecting waterproof adhesive tape. Background Technology

[0002] When installing cameras, sensors, and other electronic equipment outdoors or in humid environments, industry standard operating procedures require that power cords and signal cable connectors be tightly wrapped with specialized waterproof tape to prevent moisture from seeping into the equipment along the cables and causing short circuits, corrosion, and damage. Traditional quality control relies on manual visual inspection, which is inefficient, subjective, difficult to quantify, and lacks traceability, seriously affecting project quality and operational reliability. Summary of the Invention

[0003] This invention provides a method, apparatus, and device for detecting waterproof tape. By inputting the image of the camera wiring section into the detection model, it can not only output the detection results of the waterproof tape at the wiring section and achieve precise positioning of the waterproof tape, but also simultaneously output the assessment results of the integrity of the waterproof tape wrapping. This effectively overcomes the limitation of related technologies that can only detect the existence of the target, and solves the problems of low efficiency, strong subjectivity, and difficulty in quantitative traceability of traditional manual inspection. It realizes closed-loop control of the installation quality of waterproof tape and effectively ensures the waterproof reliability of outdoor cameras and other electronic devices.

[0004] This invention provides a method for testing waterproof adhesive tape, comprising the following steps: Acquire an image of the camera's wiring connection point; The image is input into the detection model, the detection module of the detection model outputs the detection result of the waterproof tape at the connection point, and the evaluation module of the detection model outputs the evaluation result of the wrapping integrity of the waterproof tape.

[0005] According to a method for detecting waterproof adhesive tape provided by the present invention, the detection result of the waterproof adhesive tape includes the bounding box corresponding to the waterproof adhesive tape in the image; the evaluation module outputs the wrapping integrity evaluation result of the waterproof adhesive tape, including: Determine the probability distribution of different visual features within the image region corresponding to the bounding box; Based on the probability distribution, determine the visual feature distribution information entropy; Based on the information entropy, the integrity assessment result of the waterproof tape wrapping is output.

[0006] According to the waterproof adhesive tape testing method provided by the present invention, the testing model further includes: An adaptive preprocessing module is used to convert the image of the camera wiring section to the LAB color space, enhance the LAB three-channel information corresponding to the waterproof tape, and output the enhanced image of the camera wiring section.

[0007] According to the waterproof adhesive tape testing method provided by the present invention, the testing model further includes: The system includes a feature extraction module and a feature fusion module; wherein the feature extraction module is used to extract contour features from the image that are compatible with the wrapping shape of the waterproof tape. The feature fusion module is used to dynamically weight and fuse the texture features, context features, and contour features extracted from the image to obtain fused features; the fused features are used by the detection model to determine the detection result of the waterproof tape at the connection point and the evaluation result of the integrity of the waterproof tape wrapping.

[0008] According to the present invention, a method for detecting waterproof adhesive tape is provided, wherein the feature extraction module includes: Horizontal strip convolution kernel and vertical strip convolution kernel.

[0009] According to the present invention, a method for detecting waterproof adhesive tape is provided, wherein the detection model is obtained by training a target loss function, the target loss function comprising: Aspect Ratio Consistency Loss; The aspect ratio consistency loss is used to characterize the difference between the aspect ratio of the predicted bounding box corresponding to the waterproof tape in the detection results and the aspect ratio of the actual bounding box corresponding to the waterproof tape at the camera wiring location.

[0010] The present invention also provides a waterproof adhesive tape testing device, comprising the following modules: The acquisition module is used to acquire images of the camera's wiring connections. The detection module is used to input the image into the detection model. The detection module of the detection model outputs the detection result of the waterproof tape at the connection point, and the evaluation module of the detection model outputs the evaluation result of the wrapping integrity of the waterproof tape.

[0011] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the waterproof tape detection method as described above.

[0012] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the waterproof tape detection method as described above.

[0013] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the waterproof tape detection method as described above.

[0014] This invention provides a method, apparatus, and device for detecting waterproof tape. By inputting the acquired image of the camera wiring location into the detection model, it can not only output the detection results of the waterproof tape at the wiring location, achieving precise positioning of the waterproof tape, but also simultaneously output the assessment results of the integrity of the waterproof tape wrapping. This effectively overcomes the limitation of related technologies that can only detect the existence of the target, and solves the problems of low efficiency, strong subjectivity, and difficulty in quantitative traceability of traditional manual inspection. It realizes closed-loop control of the installation quality of waterproof tape and effectively ensures the waterproof reliability of outdoor cameras and other electronic devices. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is one of the flowcharts of the waterproof adhesive tape testing method provided by the present invention.

[0017] Figure 2 This is the second flowchart of the waterproof adhesive tape testing method provided by the present invention.

[0018] Figure 3 This is a schematic diagram of the waterproof adhesive tape testing device provided by the present invention.

[0019] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] The following is combined with Figures 1 to 4 The present invention describes a method, apparatus, and equipment for testing waterproof adhesive tape.

[0022] To facilitate a clearer understanding of the technical solutions of the various embodiments of this application, some technical content related to the various embodiments of this application will be introduced first.

[0023] With the development of deep learning technology, object detection technologies, represented by the YOLO series algorithms, have been widely used in various visual recognition tasks. These technologies directly regress the location and category of objects from images through convolutional neural networks, achieving a balance between speed and accuracy. However, directly applying general object detection models to the specific scenario of waterproof tape detection still faces many challenges, such as small target size, varied shapes, similar colors to the background, complex lighting, and stringent performance requirements for mobile deployment.

[0024] Currently, to address the problem of target detection in complex scenarios, some solutions attempt to optimize the image at the input end or introduce general-purpose enhancement modules into the network structure. While these approaches are innovative, they still suffer from the following fundamental technical shortcomings when dealing with the specific, vertical, and demanding scenario of "detecting waterproof tape from a camera": Computational Redundancy and Deployment Paradox: The Bayesian-VAE-Kalman filter preprocessing stage involves complex probability integration, posterior inference, and iterative optimization, resulting in a massive computational load. This contradicts the core requirement of achieving real-time detection on resource-constrained Android mobile devices. Such a heavyweight preprocessing pipeline is virtually impossible to run efficiently on mobile devices, causing it to lose its practical value for real-time feedback and creating a deployment paradox of "theoretically feasible, but practically useless."

[0025] Information distortion and feature annihilation risks: Over- or inappropriate preprocessing is a double-edged sword. Waterproof tape (especially black tape wrapped around black cables) is inherently a low-contrast, low-texture target. Existing denoising and reconstruction models optimize the "sharpness" or "signal-to-noise ratio" of the global image. While smoothing noise, they may inadvertently "optimize" away key discriminative features such as the already weak edges between the tape and the cable, subtle reflections or textures on the tape surface, causing irreversible information loss and increasing the difficulty of subsequent detection.

[0026] The contradiction between the "broad-based" approach to feature extraction and the "precise and specialized" approach to scene analysis lies in the fact that the "parallel multi-attention module" integrated in YOLOv10 is a general design that is not specifically designed for the unique spatial morphology and topological relationship of waterproof tape, which is "long, slender, wrapped, and attached to cables." For targets like tape with a very large aspect ratio and a spiral wrapping shape, conventional attention mechanisms based on square receptive fields are inefficient at capturing its global morphological features.

[0027] The most critical issue is that existing technological solutions are essentially still just "target locators." They can at best answer the question, "Is there something like tape at this location in the image?", but cannot answer the more crucial questions in engineering practice: "Is this tape wrapped completely and properly?" They lack a mechanism for quality assessment of the target's internal state and cannot distinguish between fine-grained states such as "proper wrapping," "partial wrapping (exposed wires)," and "false wrapping (merely covering without tightening)," thus failing to achieve true closed-loop control of installation quality.

[0028] Figure 1 This is one of the flowcharts illustrating the waterproof adhesive tape testing method provided by the present invention, such as... Figure 1 As shown, the method includes the following: Step 101: Obtain an image of the camera wiring connection point.

[0029] Specifically, in the implementation of this application, images of the camera wiring area can be efficiently obtained by taking pictures of the camera wiring area.

[0030] Step 102: Input the image into the detection model. The detection module of the detection model outputs the detection results of the waterproof tape at the connection point, and the evaluation module of the detection model outputs the evaluation results of the integrity of the waterproof tape wrapping.

[0031] Specifically, after acquiring an image of the camera wiring connection, the image can be input into the detection model to obtain the detection result of the waterproof tape at the connection and the evaluation result of the waterproof tape's wrapping integrity. Optionally, the detection model of this application includes a detection module and an evaluation module. The detection module is used to accurately output the detection result of the waterproof tape at the connection based on the image of the camera wiring connection. Optionally, the detection result of the waterproof tape at the connection includes the target category, confidence level, and bounding box coordinates of the waterproof tape. The evaluation module is used to output the evaluation result of the waterproof tape's wrapping integrity based on the visual features within the image area corresponding to the bounding box of the waterproof tape. This overcomes the limitation of related technologies that can only detect the existence of targets, and realizes an integrated and accurate judgment of whether the waterproof tape "exists" and "wraps properly." This effectively adapts to the needs of large-scale engineering deployment and subsequent inspection, and improves the closed-loop management efficiency of waterproof tape installation quality.

[0032] It should be noted that the waterproof tape detection method in this application embodiment can be applied to both mobile devices and servers. Optionally, when the method is applied to a mobile device, the mobile device acquires an image of the camera wiring connection to be detected using its own camera, and transmits the acquired image to the server. The server then inputs the acquired image into a detection model to obtain the waterproof tape detection result and the wrapping integrity assessment result of the wiring connection. Optionally, when the method is applied to a server, the server acquires the image of the camera wiring connection to be detected sent by the mobile device, inputs the acquired image into a detection model, and obtains the waterproof tape detection result and the wrapping integrity assessment result of the wiring connection.

[0033] The method described in the above embodiments, by inputting the image of the camera wiring section into the detection model, can not only output the detection result of the waterproof tape at the wiring section and achieve accurate positioning of the waterproof tape, but also simultaneously output the evaluation result of the wrapping integrity of the waterproof tape. This effectively overcomes the limitation of related technologies that can only detect the existence of the target, solves the problems of low efficiency, strong subjectivity, and difficulty in quantitative traceability of traditional manual inspection, realizes closed-loop control of the installation quality of waterproof tape, and effectively ensures the waterproof reliability of outdoor cameras and other electronic devices.

[0034] Optionally, in this embodiment, the detection module and the evaluation module can be two parallel regression task heads in the detection model. By setting up a parallel evaluation module on top of the detection module, it is possible to automatically detect whether the waterproof tape is installed in place, which is suitable for large-scale batch inspection scenarios and improves efficiency and accuracy.

[0035] In some embodiments, the detection result of the waterproof tape includes the bounding box corresponding to the waterproof tape in the image; the evaluation module outputs the wrapping integrity evaluation result of the waterproof tape, including: Determine the probability distribution of different visual features within the image region corresponding to the bounding box; Determine the information entropy of the visual feature distribution based on the probability distribution; Based on information entropy, output the assessment results of the wrapping integrity of the waterproof tape.

[0036] Specifically, in this embodiment, based on the waterproof tape bounding box output by the detection module, the region of interest (RoI) corresponding to the bounding box can be cropped from the image of the camera wiring location. After dimensionality compression and feature refinement processing through a convolutional network, it is then mapped to a multidimensional probability distribution vector P through a Softmax activation function. Optionally, the multidimensional probability distribution vector can be the probability distribution of different visual features within the image region corresponding to the bounding box. Visual features include texture consistency, color uniformity, and edge features of exposed cables, which are directly related to the wrapping quality.

[0037] Optionally, after determining the probability distribution of different visual features within the image region corresponding to the bounding box, the degree of disorder of this distribution can be calculated using the information entropy formula based on the probability distribution of different visual features within the image region corresponding to the bounding box. Optionally, if the waterproof tape is wrapped completely and neatly, the visual features are highly uniform, and the information entropy value is low. Optionally, if the waterproof tape is not wrapped completely, with exposed cables, foreign matter mixed in, or texture breaks caused by the tape not being stretched taut, then multiple heterogeneous visual features such as tape, cables, and air will coexist, and the information entropy H value will increase significantly.

[0038] Optionally, after determining the visual feature distribution information entropy based on the probability distribution of different visual features within the image region corresponding to the bounding box, this application can compare the calculated information entropy H with a preset information entropy threshold T, and output the wrapping integrity assessment result of the waterproof tape. Optionally, if the calculated information entropy H is less than or equal to the information entropy threshold T, the wrapping integrity assessment result is output as complete and conforms to the specifications. Optionally, if the calculated information entropy H is greater than the information entropy threshold T, the wrapping integrity assessment result is output as incomplete and poses a waterproof risk.

[0039] For example, in this application embodiment, the wrapping integrity score of the waterproof tape is determined based on the following method: in, Visual features that strongly correlate with wrapping quality include texture consistency, color uniformity, and exposed cables within the area corresponding to the waterproof tape boundary frame. Indicates will A C-dimensional probability distribution vector is obtained through a small convolutional network and a softmax layer. This vector represents the probability distribution of different visual features such as texture consistency, color uniformity, and exposed cables within the area corresponding to the waterproof tape boundary frame. Represents the calculation of information entropy; This represents the final integrity score. Traditional object detection only focuses on the presence or absence of the object, while this application transforms the quality of waterproof tape wrapping into a quantifiable feature distribution problem by extracting the probability distribution of different visual features within the image region corresponding to the bounding box. This effectively overcomes the limitations of traditional detection methods and achieves closed-loop control of the waterproof tape installation quality.

[0040] The method described above extracts visual features strongly correlated with wrapping quality, such as texture consistency, color uniformity, and exposed cables, within the area corresponding to the waterproof tape boundary frame. These features are then processed by a small convolutional network and activated by Softmax to obtain a multidimensional probability distribution vector. The degree of feature distribution disorder is then quantified using information entropy, ultimately outputting an evaluation result of the waterproof tape wrapping integrity. This method overcomes the limitation of traditional target detection, which can only determine the existence of a target, and achieves a breakthrough from existence detection to quality evaluation. It effectively solves the problems of low efficiency, strong subjectivity, and inconsistent standards in manual detection, and realizes closed-loop control of waterproof tape installation quality, significantly improving the waterproof reliability and maintenance efficiency of outdoor cameras and other electronic devices.

[0041] In some embodiments, the detection model further includes: The adaptive preprocessing module is used to convert the image of the camera wiring area to the LAB color space, enhance the LAB three-channel information corresponding to the waterproof tape, and output the enhanced image of the camera wiring area.

[0042] Specifically, in this embodiment of the application, after acquiring the image of the camera wiring part, the original RGB format image of the camera wiring part can be mapped to the LAB three-channel through the adaptive preprocessing module to obtain the LAB format tensor, thereby effectively solving the problems of low contrast and similar color between the waterproof tape and the background cable in the RGB space, and highlighting the subtle color difference between the waterproof tape and the background.

[0043] Furthermore, the LAB tensor can be convolved using learnable 3x3 convolutional layers. Simultaneously, channel attention weight vectors are generated through global average pooling and dual fully connected layers. The convolved feature map is then element-wise multiplied with the weight vectors, followed by a sigmoid activation function to output an enhanced feature map. This automatically learns the optimal combination of the three LAB channels. For example, it enhances the reflective features of the L channel in bright light and amplifies the subtle color differences between the A and B channels in shadowy scenes. This effectively preserves the core features of the waterproof tape, such as its outline, texture, and color, while suppressing background noise, thus significantly improving the accuracy of waterproof tape detection.

[0044] For example, this application embeds a lightweight, learnable convolutional layer at the very front of the detection model. This layer is automatically optimized during training to adaptively enhance the input features most important to the current task, replacing the traditional fixed and cumbersome preprocessing process and achieving a balance between performance and efficiency. Optionally, the learnable convolutional layer is as follows: in, A tensor representing the conversion of the input image from RGB to LAB color space; This represents a 3x3 learnable convolutional layer that learns during training how to optimally combine and enhance the three-channel information of LAB to highlight the features of the adhesive tape. This represents a channel attention weight vector generated by simple global average pooling and two fully connected layers, used to dynamically adjust the importance of different channels; ⨀ represents element-wise multiplication; σ represents the Sigmoid activation function; This represents the output activation feature map.

[0045] The method described in the above embodiments converts the original RGB image to the LAB color space through an adaptive preprocessing module, thereby effectively solving the limitations of low contrast and similar colors between waterproof tape and background cables in the RGB space, and accurately highlighting the subtle color differences between the two. Then, a learnable convolutional layer performs convolution operations on the LAB three-channel information, automatically learning the optimal combination of the LAB three channels. This not only efficiently preserves the core discrimination features such as the outline, texture, and color of the waterproof tape, but also effectively suppresses background noise, significantly improving the accuracy of waterproof tape detection.

[0046] In some embodiments, the detection model further includes: The system includes a feature extraction module and a feature fusion module; the feature extraction module is used to extract contour features from the image that are compatible with the shape of the waterproof tape wrapping. The feature fusion module is used to dynamically weight and fuse the texture features, context features, and contour features extracted from the image to obtain fused features. The fused features are used by the detection model to determine the detection results of the waterproof tape at the connection point and the evaluation results of the integrity of the waterproof tape wrapping.

[0047] Specifically, in this embodiment, horizontal and vertical strip convolution kernels can be used to perform convolution operations on the preprocessed enhanced image feature map, capturing the elongated contour information of the waterproof tape in the horizontal and vertical directions, and outputting a contour feature map. It should be noted that strip convolution can accurately match the slender, wrapped geometric shape of the waterproof tape, thereby accurately extracting contour features adapted to the shape of the waterproof tape, effectively solving the problems of low efficiency and large background interference in capturing elongated target features using traditional square convolution kernels.

[0048] Optionally, in this embodiment, detailed features such as material texture and wrapping patterns on the surface of the waterproof tape can also be extracted. Global average pooling and fully connected layers are used to generate gating weights corresponding to three types of features: texture, contour, and context. Multi-dimensional features are dynamically weighted and fused, effectively solving the problem of insufficient discrimination power of a single feature in complex environments and significantly improving the accuracy of detection results. Optionally, in the process of generating gating weights corresponding to the three types of features (texture, contour, and context) through global average pooling and fully connected layers, contour feature weights can be prioritized in cluttered backgrounds, texture feature weights can be strengthened in scenes with similar materials, and context feature weights can be emphasized in complex spatial layouts.

[0049] For example, considering the elongated and winding characteristics of waterproof tape, this application extracts contour features using a strip convolution kernel: in, This represents the feature map input from the backbone network; This represents horizontal strip convolution with a kernel size of 1 x k, for example, k=7; This represents a vertical strip convolution with a kernel size of k x 1. It should be noted that traditional square convolution kernels introduce a large amount of irrelevant background information when perceiving slender targets. The strip convolution proposed in this application forces the network to aggregate information along one dimension, greatly enhancing its ability to perceive elongated structures, thus perfectly matching the geometry of the tape.

[0050] For example, in this application, dynamic weighted fusion of texture features, context features, and contour features is performed in the following manner: in, , The output feature maps are for three parallel branches: texture, contour, and context; GAP represents global average pooling; FC represents a fully connected layer. This represents three learnable gating weight scalars whose sum is 1; This represents the feature map of the final fused output. In other words, this application introduces a dynamic, input-adaptive fusion mechanism that can adjust the fusion based on the current input features. The system determines whether the texture of the tape, its outline, or its context is more important, and assigns higher weights to the corresponding branches, thereby effectively solving the problem of insufficient discrimination power of single features in complex environments and significantly improving the accuracy of detection results.

[0051] The method described in the above embodiments accurately captures the elongated outline information of waterproof tape in the horizontal and vertical directions through horizontal and vertical strip convolution kernels. It accurately adapts to the slender and winding geometric shape of waterproof tape, effectively solving the problems of low efficiency and large background interference in capturing elongated target features by traditional square convolution kernels, and achieving accurate extraction of tape outline. At the same time, it simultaneously extracts detailed features such as surface material texture and winding pattern of tape, as well as contextual features related to tape and cable. Then, through the feature fusion module, it achieves dynamic weighted fusion of multi-dimensional features. The output fused features have morphological accuracy, rich details, and scene relevance, which can effectively solve the problem of insufficient discrimination power of single features in complex environments and significantly improve the accuracy of detection results.

[0052] In some embodiments, the detection model is trained using a target loss function, which includes: Aspect Ratio Consistency Loss: Aspect Ratio Consistency Loss is used to characterize the difference between the aspect ratio of the predicted bounding box corresponding to the waterproof tape in the detection results and the aspect ratio of the actual bounding box corresponding to the waterproof tape at the camera wiring location.

[0053] Specifically, in the embodiments of this application, during the training process of the detection model, an aspect ratio consistency loss is added to the standard loss function, making the loss value sensitive to the proportional relationship of the bounding box. The detection model can then effectively learn the slender shape features of the waterproof tape, and the aspect ratio of the output bounding box of the waterproof tape can effectively approximate the true value, thereby ensuring that the predicted bounding box shape of the waterproof tape highly matches the actual tape outline and significantly improving the accuracy of the detection results.

[0054] For example, the aspect ratio consistency loss in this application embodiment is as follows: in, This is an indicator function; it is 1 if the i-th prediction box is responsible for predicting the j-th true target, and 0 otherwise. Let be the width and height of the i-th prediction box; Let be the width and height of the j-th real target bounding box; This is the angle corresponding to the aspect ratio of the calculation frame. In other words, this application maps the aspect ratio to angle space, using... As a loss, this makes the loss value sensitive to angular differences, i.e., the aspect ratio, but independent of the absolute size of the box. This strongly guides the network to learn and predict bounding boxes that conform to the slender characteristics of waterproof tape, effectively suppressing the tendency to generate short and wide boxes.

[0055] The method described above adds an aspect ratio consistency loss to the loss function of the detection model, making the loss value sensitive only to the aspect ratio and independent of the absolute size of the bounding box. This guides the model to actively learn the slender shape characteristics of the waterproof tape during training, ensuring that the predicted bounding box shape closely matches the actual tape outline. This solves the problem of uneven penalty for aspect ratio errors in traditional loss functions for bounding boxes of different sizes, thus improving the bounding box prediction accuracy for both small and large waterproof tapes. It significantly reduces the risk of false detections and missed detections caused by shape positioning deviations, effectively ensuring the accuracy of waterproof tape installation quality detection in complex scenarios.

[0056] In some embodiments, the test results of the waterproof tape include the detected target category and the confidence level of the target category, and the wrapping integrity assessment results include the wrapping integrity score; the waterproof tape test method further includes at least one of the following: When the target category is not waterproof tape, or the target category is waterproof tape but the confidence level is less than or equal to the first threshold, output a prompt message that no waterproof tape was detected; When the target category is waterproof tape and the confidence level is greater than the first threshold, if the wrapping integrity score is greater than the second threshold, an assessment report is generated indicating that the waterproof tape installation meets the specifications; if the wrapping integrity score is less than or equal to the second threshold, an alarm message is generated indicating that the waterproof tape is not fully wrapped and there is a risk of waterproofing.

[0057] Specifically, this application can accurately determine whether waterproof tape has been detected based on the target category and confidence level output by the detection module. Optionally, if the target category is another category, or if it is determined to be waterproof tape but the confidence level does not reach the first threshold (e.g., 0.85), then it is considered that waterproof tape has not been detected, and a prompt message indicating that waterproof tape has not been detected is output. Optionally, if the target category is waterproof tape and the confidence level is greater than the first threshold, then it is considered that waterproof tape has been detected, and further installation compliance assessment is performed based on the wrapping integrity score output by the evaluation module. Optionally, if the wrapping integrity score is higher than the second threshold (e.g., 0.90), it is determined that the waterproof tape is wrapped evenly and completely, and a compliance assessment report is generated; if the score is lower than or equal to the second threshold, it is determined that there are problems such as wrapping gaps, missing wrapping, or insufficient overlap, and an alarm message is generated to provide inspection personnel with accurate rectification information.

[0058] The method described in the above embodiments, based on the detected target category, target category confidence level, and wrapping integrity score, performs a graded judgment, which significantly improves the control accuracy of waterproof tape installation quality, effectively adapts to large-scale engineering inspection scenarios, reduces manual inspection costs, and reduces the risk of missed or false detections.

[0059] For example, such as Figure 2As shown in the embodiments of this application, a method for detecting waterproof adhesive tape is provided, as detailed below: (1) Users and inspectors start the smart inspection application on the Android terminal, point the camera at the camera wiring area to be inspected, and trigger the photo taking or real-time detection.

[0060] (2) The image acquisition module captures one or more frames of images and uses them as input tensors. It is passed to the Gaia Intelligent Evaluation Engine.

[0061] (3) "The beginning of chaos" CGL layer reception It is processed by learnable convolutional kernels to output an enhanced feature map. Optionally, the "Chaos-Genesis Layer (CGL)" is a learnable adaptive preprocessing layer embedded within the network, serving as a lightweight map enhancement module to replace traditional fixed preprocessing. For example, the "Chaos-Genesis" CGL layer can be used as an adaptive preprocessing module in a detection model to perform learnable adaptive preprocessing enhancement on the input raw image, outputting an enhanced feature map that replaces traditional fixed preprocessing and improves the robustness of subsequent feature extraction.

[0062] (4) Backbone network Multi-layer convolution and downsampling are performed to extract basic feature pyramids {P3, P4, P5} at different scales. For example, the backbone network can serve as a feature extraction module in the detection model, performing multi-layer convolution and downsampling on the enhanced feature map to extract multi-scale basic feature pyramids {P3, P4, P5}, providing unified feature support for the detection and evaluation modules.

[0063] (5) The "Nine-Revolutions Exquisite Neck" receives {P3, P4, P5} and processes and fuses them through parallel branches of texture, contour, and context within it, outputting a set of more targeted multi-scale fusion features {N3, N4, N5}. Optionally, the "Nine-Revolutions Exquisite Neck" (NREN) multi-axis attention neck is a novel network neck designed specifically for capturing slender, entangled targets, incorporating multi-dimensional feature fusion of texture, contour, and context. For example, the "Nine-Revolutions Exquisite Neck" can be used as a feature fusion module in a detection model, performing multi-dimensional feature fusion through parallel branches of texture, contour, and context, outputting multi-scale fusion features {N3, N4, N5}, thereby effectively enhancing the feature representation of the slender, entangled shape of waterproof tape.

[0064] (6) The "Tailored Anchor Generation" TAG and the detection head dynamically predict anchor boxes that match the target shape based on {N3, N4, N5}, and regress the target's category cls, confidence conf, and bounding box coordinates bbox. Optionally, the "Tailored Anchor Generation" dynamic anchor box matching mechanism is a strategy that dynamically generates high aspect ratio anchor boxes based on feature maps, and is supplemented by a dedicated loss function for supervision. For example, the "Tailored Anchor Generation" TAG and the detection head can be used as a detection module in the detection model, taking multi-scale fused features as input to complete the target category determination and confidence calculation.

[0065] (7) The “Mirror Flower Water Moon” IISH evaluation head uses bounding boxes to crop the features of the region of interest (RoI) from the feature map {N3, N4, N5}. The system is then analyzed to derive an integrity score within the range of [0, 1]. Optionally, the "Illusory Integrity Scoring Head" (IISH) is a novel task head that runs parallel to the detection head and is used to quantify and score the integrity of tape wrapping. For example, the IISH evaluation head can serve as an evaluation module within the detection model. Based on the bounding box output by the detection module, it extracts the corresponding image region and combines the texture and contour features from the multi-scale fusion features output by the feature fusion module to calculate the wrapping integrity score.

[0066] (8) Output detection results .

[0067] (9) Determine whether cls is "waterproof tape" and conf is higher than the preset threshold. If yes, evaluate the integrity of the waterproof tape wrapping; if no, report "no waterproof tape detected".

[0068] (10) Determine the completeness score Is it higher than a preset standard threshold (e.g., 0.9)? If yes, generate a "Compliant with Standards" report; otherwise, generate a "Incomplete Wrapping, Risk Present" warning, and you can... As a basis for quantification.

[0069] (11) Display the results with annotation boxes, ratings and text suggestions on the screen.

[0070] The method described above seamlessly integrates adaptive preprocessing, morphology-aware feature extraction, and application-level quality assessment into a unified, lightweight neural network, achieving end-to-end intelligent judgment from the original image to whether the installation is compliant.

[0071] like Figure 3 As shown, the waterproof adhesive tape testing device provided by the present invention will be described below. The waterproof adhesive tape testing device described below can be referred to in correspondence with the waterproof adhesive tape testing device method described above. The waterproof adhesive tape testing device includes: The acquisition module 310 is used to acquire images of the camera wiring connection point; The detection module 320 is used to input images into the detection model. The detection module of the detection model outputs the detection results of the waterproof tape at the connection point, and the evaluation module of the detection model outputs the evaluation results of the wrapping integrity of the waterproof tape.

[0072] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, communications interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a waterproof tape detection method, which includes: acquiring an image of the camera wiring connection; inputting the image into a detection model; the detection module of the detection model outputting the detection result of the waterproof tape at the wiring connection; and the evaluation module of the detection model outputting an evaluation result of the waterproof tape's wrapping integrity. Optionally, the electronic device can be a mobile device or a server.

[0073] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the waterproof tape detection method provided by the above methods. The method includes: acquiring an image of the camera wiring part; inputting the image into a detection model; the detection module of the detection model outputting the detection result of the waterproof tape at the wiring part; and the evaluation module of the detection model outputting the evaluation result of the wrapping integrity of the waterproof tape.

[0075] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program is implemented to perform the waterproof tape detection method provided by the above methods. The method includes: acquiring an image of the camera wiring portion; inputting the image into a detection model; the detection module of the detection model outputting the detection result of the waterproof tape at the wiring portion; and the evaluation module of the detection model outputting the evaluation result of the wrapping integrity of the waterproof tape.

[0076] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0077] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for testing waterproof adhesive tape, characterized in that, include: Acquire an image of the camera's wiring connection point; The image is input into the detection model, the detection module of the detection model outputs the detection result of the waterproof tape at the connection point, and the evaluation module of the detection model outputs the evaluation result of the wrapping integrity of the waterproof tape.

2. The method for testing waterproof adhesive tape according to claim 1, characterized in that, The detection result of the waterproof tape includes the bounding box corresponding to the waterproof tape in the image; the evaluation module outputs the wrapping integrity evaluation result of the waterproof tape, including: Determine the probability distribution of different visual features within the image region corresponding to the bounding box; Based on the probability distribution, determine the visual feature distribution information entropy; Based on the information entropy, the integrity assessment result of the waterproof tape wrapping is output.

3. The method for testing waterproof adhesive tape according to claim 1 or 2, characterized in that, The detection model also includes: An adaptive preprocessing module is used to convert the image of the camera wiring section to the LAB color space, enhance the LAB three-channel information corresponding to the waterproof tape, and output the enhanced image of the camera wiring section.

4. The method for testing waterproof adhesive tape according to claim 1 or 2, characterized in that, The detection model also includes: The system includes a feature extraction module and a feature fusion module; wherein the feature extraction module is used to extract contour features from the image that are compatible with the wrapping shape of the waterproof tape. The feature fusion module is used to dynamically weight and fuse the texture features, context features, and contour features extracted from the image to obtain fused features; the fused features are used by the detection model to determine the detection result of the waterproof tape at the connection point and the evaluation result of the wrapping integrity of the waterproof tape.

5. The method for testing waterproof adhesive tape according to claim 1 or 2, characterized in that, The feature extraction module includes: Horizontal strip convolution kernel and vertical strip convolution kernel.

6. The method for testing waterproof adhesive tape according to claim 1 or 2, characterized in that, The detection model is trained using a target loss function, which includes: Aspect Ratio Consistency Loss; The aspect ratio consistency loss is used to characterize the difference between the aspect ratio of the predicted bounding box corresponding to the waterproof tape in the detection results and the aspect ratio of the actual bounding box corresponding to the waterproof tape at the camera wiring location.

7. The method for testing waterproof adhesive tape according to claim 1 or 2, characterized in that, The detection results of the waterproof tape include the detected target category and the confidence level of the target category; the winding integrity assessment results include a winding integrity score; the method further includes at least one of the following: When the target category is not waterproof tape, or when the target category is waterproof tape but the confidence level is less than or equal to the first threshold, a prompt message indicating that waterproof tape was not detected is output. When the target category is waterproof tape and the confidence level is greater than the first threshold, if the wrapping integrity score is greater than the second threshold, an evaluation report showing that the waterproof tape installation meets the specifications is generated; if the wrapping integrity score is less than or equal to the second threshold, an alarm message indicating that the waterproof tape is not fully wrapped and there is a risk of waterproofing is generated.

8. A waterproof adhesive tape testing device, characterized in that, include: The acquisition module is used to acquire images of the camera's wiring connections. The detection module is used to input the image into the detection model. The detection module of the detection model outputs the detection result of the waterproof tape at the connection point, and the evaluation module of the detection model outputs the evaluation result of the wrapping integrity of the waterproof tape.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the waterproof adhesive tape detection method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the waterproof adhesive tape testing method as described in any one of claims 1 to 7.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the waterproof adhesive tape testing method as described in any one of claims 1 to 7.

Images