Artificial intelligence-based image detection method, device, equipment and storage medium

Abstract

The application belongs to the field of artificial intelligence and the field of digital medicine, and relates to an image detection method based on artificial intelligence, comprising: acquiring a joint image of a target user; calling a preset feature database; performing irrelevant image screening on the joint image based on the feature database to obtain a corresponding target image; performing compliance verification on the target image based on a preset multi-classification model to obtain a verification result corresponding to the target image; if the verification result is a verification pass, detecting the target image based on a preset detection model to generate a target detection result corresponding to the target image. The application also provides an image detection device based on artificial intelligence, a computer device and a storage medium. In addition, the application also relates to blockchain technology, and error positioning information can be stored in a blockchain. The application can be applied to a joint image detection scene in the field of digital medicine, improves the detection efficiency of joint images, and ensures the accuracy of the generated joint image detection result.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence development technology and digital healthcare, and in particular to image detection methods, devices, computer equipment and storage media based on artificial intelligence. Background Technology

[0002] In the medical field, limited active or passive range of motion in small joints such as the wrist and ankle is a very common clinical problem. Common assessments of joint movement function typically use pre-installed evaluation rules to evaluate input joint images, generating assessment results. These rules usually involve measuring the maximum range of motion of the joint using tools such as goniometers, combined with manual evaluation to produce the final assessment result for the joint images. This rule-based approach to joint movement function assessment suffers from low detection efficiency and high cost. Summary of the Invention

[0003] The purpose of this application is to propose an image detection method, apparatus, computer device, and storage medium based on artificial intelligence, in order to solve the technical problem that the existing common joint motion function assessment usually uses pre-installed evaluation rules in the device to evaluate the input joint image and generate the joint image evaluation result. Such evaluation rule-based joint motion function assessment methods have the problems of low detection efficiency and high detection cost.

[0004] To address the aforementioned technical problems, this application provides an image detection method based on artificial intelligence, employing the following technical solution:

[0005] Obtain joint images of the target user;

[0006] Call the preset feature database;

[0007] Irrelevant images are filtered out from the joint images based on the feature database to obtain the corresponding target images;

[0008] The target image is subjected to compliance verification based on a preset multi-classification model to obtain the verification result corresponding to the target image.

[0009] If the verification structure passes the verification, the target image is detected based on the preset detection model to generate a target detection result corresponding to the target image.

[0010] Furthermore, the step of filtering out irrelevant images from the joint images based on the feature database to obtain the corresponding target images specifically includes:

[0011] Construct a first feature vector corresponding to the joint image;

[0012] The similarity between the first feature vector and all the second feature vectors included in the feature database is calculated to generate the similarity between the first feature vector and each of the second feature vectors.

[0013] Based on the similarity, obtain the target similarity with the highest numerical value corresponding to the first feature vector;

[0014] Irrelevant images are filtered out from the joint images based on the target similarity.

[0015] The irrelevant images in the joint image are screened out to obtain the target image after screening.

[0016] Furthermore, the step of filtering irrelevant images from the joint images based on the target similarity specifically includes:

[0017] Obtain the specified similarity corresponding to the third feature vector; wherein, the third feature vector is any one of all the first feature vectors;

[0018] Determine whether the specified similarity is less than a preset similarity threshold;

[0019] If so, obtain the specified image corresponding to the third feature vector;

[0020] The specified image is used as an irrelevant image in the joint image.

[0021] Furthermore, the step of performing compliance verification on the target image based on a preset multi-classification model to obtain a verification result corresponding to the target image specifically includes:

[0022] Invoke the multi-classification model;

[0023] The target image is input into the multi-classification model;

[0024] The target image is classified using the multi-classification model to obtain a classification result corresponding to the target image.

[0025] Based on the classification results, a verification result corresponding to the target image is generated.

[0026] Furthermore, after the step of detecting the target image based on a preset detection model and generating a target detection result corresponding to the target image, the method further includes:

[0027] Obtain the preset confidence level assessment formula;

[0028] The confidence level corresponding to the target detection result is generated based on the confidence level evaluation formula;

[0029] Obtain the communication information of the target user;

[0030] Based on the communication information, the confidence level is pushed to the target user.

[0031] Furthermore, after the step of performing compliance verification on the target image based on a preset multi-classification model to obtain the verification result corresponding to the target image, the method further includes:

[0032] If the verification result is that the verification fails, obtain the specified classification result corresponding to the verification result;

[0033] Obtain the specified information template corresponding to the specified classification result;

[0034] Based on the specified classification result and the specified information template, generate corresponding error category reminder information;

[0035] Display the error category alert information.

[0036] Furthermore, prior to the step of invoking the preset feature database, the following steps are also included:

[0037] Collect a preset number of correctly captured images of the specified joints;

[0038] The specified joint image is subjected to feature extraction using a preset extraction model to obtain an image feature vector corresponding to the specified joint image;

[0039] The image feature vector is stored in the feature database.

[0040] To address the aforementioned technical problems, this application also provides an image detection device based on artificial intelligence, employing the following technical solution:

[0041] The first acquisition module is used to acquire joint images of the target user;

[0042] The calling module is used to call the preset feature database;

[0043] The processing module is used to filter out irrelevant images from the joint images based on the feature database to obtain the corresponding target images;

[0044] The verification module is used to perform compliance verification on the target image based on a preset multi-classification model, and obtain the verification result corresponding to the target image.

[0045] The first generation module is used to detect the target image based on a preset detection model and generate a target detection result corresponding to the target image if the verification structure passes the verification.

[0046] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0047] Obtain joint images of the target user;

[0048] Call the preset feature database;

[0049] Irrelevant images are filtered out from the joint images based on the feature database to obtain the corresponding target images;

[0050] The target image is subjected to compliance verification based on a preset multi-classification model to obtain the verification result corresponding to the target image.

[0051] If the verification structure passes the verification, the target image is detected based on the preset detection model to generate a target detection result corresponding to the target image.

[0052] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0053] Obtain joint images of the target user;

[0054] Call the preset feature database;

[0055] Irrelevant images are filtered out from the joint images based on the feature database to obtain the corresponding target images;

[0056] The target image is subjected to compliance verification based on a preset multi-classification model to obtain the verification result corresponding to the target image.

[0057] If the verification structure passes the verification, the target image is detected based on the preset detection model to generate a target detection result corresponding to the target image.

[0058] Compared with the prior art, the embodiments of this application have the following main advantages:

[0059] This application embodiment first loads a target page with pre-generated embedding points based on resource location embedding rules; then calls the target interface corresponding to the target page; subsequently, it collects business embedding data corresponding to the target resource location in the target page within a preset time period based on the target interface; then, it generates product conversion data corresponding to the target resource location based on the business embedding data; finally, it recommends and replaces products within the target resource location based on the product conversion data. This application embodiment uses a pre-built detection model to detect joint images, enabling rapid and accurate generation of detection results corresponding to the joint images. This effectively solves the problems of low detection efficiency and high detection cost in existing joint movement function assessment methods, improving the detection efficiency of joint images and ensuring the accuracy of the generated joint image detection results. Furthermore, before using the detection model to detect joint images, this application intelligently uses a feature database to filter irrelevant images from the joint images and uses a preset multi-classification model to perform compliance verification on the joint images. This quality control judgment of the joint images to be processed ensures the compliance of the subsequently processed joint images, which helps reduce the model error rate of subsequent image detection processing. Attached Figure Description

[0060] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0062] Figure 2 A flowchart of an embodiment of the AI-based image detection method according to this application;

[0063] Figure 3 This is a schematic diagram of a structure of an embodiment of the artificial intelligence-based image detection device according to this application;

[0064] Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0066] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0067] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0068] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0069] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0070] Terminal devices 101, 102, and 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers, etc.

[0071] Server 105 can be a server that provides various services, such as a backend server that supports the pages displayed on terminal devices 101, 102, and 103.

[0072] It should be noted that the AI-based image detection method provided in this application is generally executed by a server / terminal device, and correspondingly, the AI-based image detection device is generally located in the server / terminal device.

[0073] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0074] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0075] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0076] Continue to refer to Figure 2 This document illustrates a flowchart of an embodiment of the AI-based image detection method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different requirements. The AI-based image detection method provided in this application can be applied to any scenario requiring image detection, and thus can be applied to products in these scenarios, such as joint assessment in the digital healthcare field. The AI-based image detection method includes the following steps:

[0077] Step S201: Obtain the joint image of the target user.

[0078] In this embodiment, the artificial intelligence-based image detection method operates on an electronic device (e.g., Figure 1The server / terminal device shown can acquire joint images of the target user via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods. This solution can be specifically applied in medical applications within the medical field. The aforementioned joint images may refer to images of the wrist and ankle joints. These joint images can be images uploaded by the target user themselves, or they can be joint images extracted from the target user's medical data downloaded from a medical platform. The aforementioned medical data may include personal health records, prescriptions, examination reports, etc.

[0079] Step S202: Call the preset feature database.

[0080] In this embodiment, the aforementioned feature database is a pre-constructed database that stores feature vectors of a certain number of correctly captured images of specified joints.

[0081] Step S203: Based on the feature database, irrelevant images are filtered out from the joint images to obtain the corresponding target images.

[0082] In this embodiment, in a specific medical application scenario, the joint image comprises multiple images. Often, some images in the joint image are completely irrelevant interference images, or the patient is not correctly positioned within the image. Therefore, it is necessary to be able to identify and filter out these irrelevant image data from the joint image in advance. The specific implementation process of filtering out irrelevant images from the joint image based on the feature database to obtain the corresponding target image will be further described in detail in subsequent embodiments of this application, and will not be elaborated upon here.

[0083] Step S204: Perform compliance verification on the target image based on a preset multi-classification model to obtain the verification result corresponding to the target image.

[0084] In this embodiment, even after filtering out irrelevant images, some non-compliant shooting issues still exist in the target image, mainly including three problems: incomplete exposure of the exposed parts, shooting direction towards the thumb side, and shooting angle not being horizontal. Therefore, a multi-classification model is needed to further verify the compliance of the target image. Specifically, the multi-classification model can be a multi-classification model trained based on the DenseNet model. DenseNet is a deep convolutional neural network that enhances feature reuse and gradient flow by introducing dense connections into the network, thereby improving the model's performance and generalization ability. In DenseNet, each layer uses the outputs of all preceding layers as its input, forming a dense connection structure. DenseNet emphasizes feature reuse and information sharing, which may slightly reduce computational efficiency, but it typically performs excellently in terms of model accuracy and generalization ability. DenseNet is widely used in computer vision tasks such as image classification, object detection, and semantic segmentation. The specific implementation process of performing compliance verification on the target image based on the preset multi-classification model to obtain the verification result corresponding to the target image will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0085] Step S205: If the verification structure passes the verification, the target image is detected based on the preset detection model to generate a target detection result corresponding to the target image.

[0086] In this embodiment, a target image can be input into a detection model, which then detects the target image and generates a target detection result corresponding to the target image. The detection model is generated by training an initial Hr-Net-based model using pre-collected regular joint images. In a medical application scenario, the target image is a wrist joint image. Detecting the wrist joint image using the detection model outputs the detection results of the key points (joint monitoring points) of the wrist joint image. Specifically, a training set can be obtained by collecting a certain number of suitable sample joint images as sample image data and labeling the sample data with corresponding detection results (motor function disability assessment results). This training set is then used to train the initial Hr-Net-based model. Specifically, the sample image data in the training set is used as the model input, and the detection results corresponding to the sample image data in the training set are used as the model input to generate the corresponding detection model. Additionally, the key points of the target image can be connected to draw the joint angle to show the current range of motion.

[0087] Specifically, HRNet is a high-resolution feature pyramid network that improves detection accuracy by constructing a depthwise separable convolutional network and a high-resolution feature pyramid network. The main advantage of HRNet is that it can process both high-resolution and low-resolution feature maps simultaneously, thereby improving the accuracy of object detection. The basic unit of HRNet is a fully convolutional network consisting of two identical branches. Each branch contains a convolutional layer, a depthwise separable convolutional layer, and a resize module. These branches operate at different resolutions and combine them into a high-resolution feature pyramid. The resize module is used to enlarge the low-resolution feature map to the size of the original input image. Specifically, (1) the basic unit of HRNet consists of the following three sub-modules: Convolutional layer: This layer performs convolution operations on the input features to extract features of local regions. Usually, a 3×3 convolutional kernel is used. Depthwise separable convolutional layer: This layer consists of two parts: depthwise convolution and pointwise convolution. Depthwise convolution is used to extract information between channels, while pointwise convolution is used to strengthen the relationship between pixels. Compared with traditional convolutional layers, depthwise separable convolutional layers have fewer parameters, faster computation speed, and can improve feature representation ability. The Resize module: This module is used to resize low-resolution feature maps to the original size of the input image so that they can be merged with other branches. The high-resolution feature pyramid of HRNet consists of four branches, corresponding to four different resolutions of the input image. These branches are connected by global average pooling and bilinear interpolation. The highest resolution feature map receives contributions from all branches, while the lowest resolution feature map is provided only by its own branch. Specifically, (2) The high-resolution feature pyramid of HRNet consists of the following four branches: Branch with a resolution of 1 / 4: This branch is used to extract global context information and generate the lowest resolution feature map. Branch with a resolution of 1 / 2: This branch is used to extract coarser local structural information and generate a lower resolution feature map. Branch with a resolution of 3 / 4: This branch is used to extract finer local structural information and generate a higher resolution feature map. Branch with a resolution of 1: This branch is used to extract the most detailed local structural information and generate the highest resolution feature map. These branches are connected by global average pooling and bilinear interpolation. Specifically, the feature map of each branch is first subjected to global average pooling, and then bilinear interpolation is performed with the feature maps of other branches to obtain a feature map of the same resolution. Finally, the feature maps of all resolutions are combined in order from low to high to form a high-resolution feature pyramid. Specifically, (3) the architecture of HRNet: it consists of parallel high-to-low resolution subnetworks, and repeated information exchange (multi-scale fusion) is performed between the multi-resolution subnetworks. The horizontal and vertical directions correspond to the depth of the network and the scale of the feature map, respectively.HRNet can process both high-resolution and low-resolution feature maps simultaneously, thus enabling it to better capture object details; HRNet uses depthwise separable convolutions to reduce computation and achieve faster speeds; HRNet introduces a high-resolution feature pyramid network, which makes it perform exceptionally well in object detection accuracy.

[0088] This application first loads a target page pre-generated with embedding rules based on resource location embedding rules; then it calls the target interface corresponding to the target page; subsequently, it collects business embedding data corresponding to the target resource location in the target page within a preset time period based on the target interface; then, it generates product conversion data corresponding to the target resource location based on the business embedding data; finally, it recommends and replaces products within the target resource location based on the product conversion data. This application uses a pre-built detection model to detect joint images, enabling rapid and accurate generation of detection results corresponding to the joint images. This effectively solves the problems of low detection efficiency and high detection cost in existing joint movement function assessment methods, improving the detection efficiency of joint images and ensuring the accuracy of the generated joint image detection results. Furthermore, before using the detection model to detect joint images, this application intelligently uses a feature database to filter out irrelevant images and uses a preset multi-classification model to perform compliance verification on the joint images. This quality control judgment ensures the compliance of the joint images processed subsequently, reducing the model error rate of the subsequent image detection processing.

[0089] In some alternative implementations, step S203 includes the following steps:

[0090] Construct a first feature vector corresponding to the joint image.

[0091] In this embodiment, a feature extraction model can be used to extract features from the joint image to construct a first feature vector corresponding to the joint image.

[0092] The similarity between the first feature vector and all the second feature vectors included in the feature database is calculated to generate the similarity between the first feature vector and each of the second feature vectors.

[0093] In this embodiment, the similarity between the first feature vector and all the second feature vectors included in the feature database can be calculated using a similarity algorithm to obtain the similarity between the first feature vector and each of the second feature vectors. The selection of the aforementioned similarity algorithm is not limited; for example, a cosine similarity algorithm can be used.

[0094] Based on the similarity, the target similarity with the highest value corresponding to the first feature vector is obtained.

[0095] In this embodiment, by calculating the similarity between the first feature vector and all the second feature vectors included in the feature database, a similarity score corresponding to each second feature vector is obtained. Then, all the obtained similarity scores are compared numerically, and the target similarity score with the highest value corresponding to the first feature vector is selected from all similarity scores.

[0096] Irrelevant images are filtered out from the joint images based on the target similarity.

[0097] In this embodiment, the specific implementation process of filtering irrelevant images from the joint images based on the target similarity will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0098] The irrelevant images in the joint image are screened out to obtain the target image after screening.

[0099] In this embodiment, the target image can be obtained by filtering out the irrelevant images in the joint image, that is, by deleting the irrelevant images from the joint image.

[0100] This application constructs a first feature vector corresponding to the joint image; then calculates the similarity between the first feature vector and all second feature vectors included in the feature database to generate a similarity score between the first feature vector and each of the second feature vectors; subsequently, based on the similarity score, it obtains the target similarity score with the highest numerical value corresponding to the first feature vector; subsequently, it filters irrelevant images from the joint image based on the target similarity score; finally, it removes the irrelevant images from the joint image to obtain the filtered target image. This application, by using a feature database, calculates the similarity between the constructed first feature vector corresponding to the joint image and all second feature vectors included in the feature database, thereby enabling fast and accurate filtering of irrelevant images from the joint image based on the obtained similarity data. This allows for the removal of the irrelevant images from the joint image, resulting in the filtered target image, thus obtaining the final target image that meets the requirements. This improves the efficiency of target image acquisition and ensures the accuracy of the target image data.

[0101] In some optional implementations of this embodiment, the step of filtering irrelevant images from the joint images based on the target similarity includes the following steps:

[0102] Obtain the specified similarity corresponding to the third feature vector.

[0103] In this embodiment, the third feature vector is any one of all the first feature vectors.

[0104] Determine whether the specified similarity is less than a preset similarity threshold.

[0105] In this embodiment, the value of the aforementioned similarity threshold is not specifically limited and can be set according to actual usage needs. Furthermore, if the target user's joint image contains an image with a similarity lower than the preset similarity threshold, the target user can be further reminded to use the correct joint image for disability assessment.

[0106] If so, obtain the specified image corresponding to the third feature vector.

[0107] The specified image is used as an irrelevant image in the joint image.

[0108] This application obtains a specified similarity corresponding to a third feature vector; then determines whether the specified similarity is less than a preset similarity threshold; if so, it obtains a specified image corresponding to the third feature vector; subsequently, it uses the specified image as an irrelevant image in the joint image. This application uses a similarity threshold to numerically compare the similarity of feature vectors, thereby enabling rapid and accurate filtering of irrelevant images from joint images based on the obtained numerical comparison results, effectively improving the efficiency of filtering irrelevant images from joint images.

[0109] In some alternative implementations, step S204 includes the following steps:

[0110] Invoke the multi-classification model.

[0111] In this embodiment, the aforementioned multi-classification model can specifically be a four-classification model trained based on the DenseNet model. This can be achieved by collecting images of various different categories as sample data and classifying each sample data accordingly (e.g., compliance, incomplete exposure, incorrect orientation, incorrect angle) to obtain a training set. This training set can then be used to train the DenseNet model to generate the desired four-classification model.

[0112] The target image is input into the multi-classification model.

[0113] The target image is classified using the multi-classification model to obtain a classification result corresponding to the target image.

[0114] In this embodiment, the target image is classified using the multi-classification model, and the probability of the target image belonging to each category is output. The category with the highest probability is then used as the classification result corresponding to the target image.

[0115] Based on the classification results, a verification result corresponding to the target image is generated.

[0116] In this embodiment, if the classification result is compliant, a verification result of "verification passed" will be generated for the target image. If the classification result is not compliant, i.e., any one of incomplete exposure, incorrect orientation, or incorrect angle, a verification result of "verification failed" will be generated for the target image.

[0117] This application calls the multi-classification model; then inputs the target image into the multi-classification model; subsequently, the multi-classification model classifies the target image to obtain a classification result corresponding to the target image, and generates a verification result corresponding to the target image based on the classification result. This application, by using a preset multi-classification model to classify the target image, can quickly generate a classification result corresponding to the target image, and then accurately generate a verification result corresponding to the target image based on the classification result, improving the efficiency of verification result generation and ensuring the accuracy of the generated target image verification result.

[0118] In some alternative implementations, after step S205, the electronic device may further perform the following steps:

[0119] Obtain the preset confidence level assessment formula.

[0120] In this embodiment, the confidence assessment formula can specifically adopt a formula based on posterior probability. The confidence assessment formula can be: p(y|x), where x is the input of the model and y is the prediction result of the model.

[0121] The confidence level corresponding to the target detection result is generated based on the confidence evaluation formula.

[0122] In this embodiment, the confidence level can be obtained by substituting the target image and the target detection result into the confidence evaluation formula.

[0123] Obtain the communication information of the target user.

[0124] In this embodiment, the aforementioned communication information may include an email address or a mobile phone number.

[0125] Based on the communication information, the confidence level is pushed to the target user.

[0126] In this embodiment, by pushing the confidence level to the target user, the target user can be helped to determine whether the detection results need to be reviewed, thereby improving and reducing the measurement error of the detection model.

[0127] This application obtains a preset confidence assessment formula; then generates a confidence score corresponding to the target detection result based on the confidence assessment formula; subsequently, it obtains the communication information of the target user; and then, based on the communication information, it pushes the confidence score to the target user. After generating the target image detection result based on the detection model, this application further generates a confidence score corresponding to the target detection result based on the confidence assessment formula and pushes the generated confidence score to the target user to help the target user determine whether the detection result needs to be reviewed, thereby improving and reducing the measurement error of the detection model.

[0128] In some optional implementations of this embodiment, after step S204, the electronic device may further perform the following steps:

[0129] If the verification result is that the verification failed, obtain the specified classification result corresponding to the verification result.

[0130] In this embodiment, the specified classification result may include any one of incomplete exposure, incorrect orientation, and incorrect angle.

[0131] Obtain the specified information template corresponding to the specified classification result.

[0132] In this embodiment, for different classification results, information templates matching each classification result are pre-built. These information templates are pre-built based on business requirements for generating reminder information related to the classification results.

[0133] Based on the specified classification result and the specified information template, generate corresponding error category reminder information.

[0134] In this embodiment, the corresponding error category reminder information can be generated by filling the specified classification result into the corresponding position in the specified information template.

[0135] Display the error category alert information.

[0136] In this embodiment, by displaying error category alerts, the target user is guided to retake correct and compliant joint images. Furthermore, the system can also retrieve and display normal, compliant images corresponding to the specified category, along with the error category alerts, to further assist the target user in retaking correct and compliant joint images.

[0137] When this application detects that the verification result is a failure, it obtains a specified classification result corresponding to the verification result; then it obtains a specified information template corresponding to the specified classification result; subsequently, it generates corresponding error category reminder information based on the specified classification result and the specified information template; and then displays the error category reminder information. When this application detects that the verification result of the target image is a failure, it further obtains the specified classification result corresponding to the verification result, and then generates and displays corresponding error category reminder information based on the specified classification result and the specified information template to guide the target user to retake a correct and compliant joint image, thereby completing intelligent reminder processing for the target user and improving the user experience.

[0138] In some optional implementations of this embodiment, before step S202, the electronic device may further perform the following steps:

[0139] Collect a preset number of correctly captured images of the specified joint.

[0140] In this embodiment, the value of the aforementioned preset quantity is not specifically limited and can be set according to actual usage needs. Correctly captured images of the specified joints can include postures that conform to the correct maximum range of motion, ensuring complete exposure of all joint monitoring points in an unobstructed state. For the ankle joint, the knee, lower leg, ankle, and foot must be exposed; for the wrist joint, the fingers, palm, wrist, and elbow must be exposed. Images should be captured from the side of the little toe or little finger, at a horizontal angle, and other suitable conditions.

[0141] The specified joint image is subjected to feature extraction using a preset extraction model to obtain an image feature vector corresponding to the specified joint image.

[0142] In this embodiment, the extraction model described above can specifically adopt the VGG-16 model. VGG-16 includes 13 convolutional layers, 3 fully connected layers, and 5 pooling layers. The convolutional and fully connected layers have weight coefficients, while the pooling layers do not involve weights.

[0143] The image feature vector is stored in the feature database.

[0144] This application acquires a preset number of correctly captured images of a specified joint; then, it extracts features from the specified joint images using a preset extraction model to obtain image feature vectors corresponding to the specified joint images; subsequently, it stores the image feature vectors in a feature database. After acquiring the specified joint images, this application extracts features from them using an extraction model, quickly obtaining image feature vectors corresponding to the specified joint images. Storing these image feature vectors in the feature database facilitates subsequent irrelevant image filtering of joint images, improving the efficiency of irrelevant image filtering for joint images.

[0145] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0146] It should be emphasized that, to further ensure the privacy and security of the target images, they can also be stored in a blockchain node.

[0147] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0148] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0149] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0150] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0151] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0152] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of an image detection device based on artificial intelligence, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0153] like Figure 3 As shown, the AI-based image detection device 300 described in this embodiment includes: a loading module 301, a calling module 302, a collection module 303, a generation module 304, and a processing module 305. Wherein:

[0154] The first acquisition module 301 is used to acquire joint images of the target user;

[0155] Module 302 is used to call a preset feature database;

[0156] Processing module 303 is used to perform irrelevant image filtering on the joint image based on the feature database to obtain the corresponding target image;

[0157] Verification module 304 is used to perform compliance verification on the target image based on a preset multi-classification model and obtain a verification result corresponding to the target image;

[0158] The first generation module 305 is used to detect the target image based on a preset detection model and generate a target detection result corresponding to the target image if the verification structure passes the verification.

[0159] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0160] In some optional implementations of this embodiment, the processing module 303 includes:

[0161] A submodule is constructed to construct a first feature vector corresponding to the joint image;

[0162] The calculation submodule is used to calculate the similarity between the first feature vector and all the second feature vectors included in the feature database, and generate the similarity between the first feature vector and each of the second feature vectors.

[0163] The first acquisition submodule is used to acquire the target similarity with the highest value corresponding to the first feature vector based on the similarity.

[0164] A filtering submodule is used to filter out irrelevant images from the joint images based on the target similarity;

[0165] The processing submodule is used to filter out the irrelevant images in the joint image to obtain the filtered target image.

[0166] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0167] In some optional implementations of this embodiment, the filtering submodule includes:

[0168] The first acquisition unit is used to acquire a specified similarity corresponding to the third feature vector; wherein the third feature vector is any one of all the first feature vectors;

[0169] The judgment unit is used to determine whether the specified similarity is less than a preset similarity threshold;

[0170] The second acquisition unit is used to acquire, if yes, a specified image corresponding to the third feature vector;

[0171] A determining unit is used to identify the specified image as an irrelevant image in the joint image.

[0172] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0173] In some optional implementations of this embodiment, the verification module 304 includes:

[0174] Call the submodule to invoke the multi-classification model;

[0175] An input submodule is used to input the target image into the multi-classification model;

[0176] The classification submodule is used to classify the target image using the multi-classification model to obtain a classification result corresponding to the target image;

[0177] A generation submodule is used to generate a verification result corresponding to the target image based on the classification result.

[0178] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0179] In some optional implementations of this embodiment, the AI-based image detection device further includes:

[0180] The second acquisition module is used to acquire the preset confidence assessment formula;

[0181] The second generation module is used to generate a confidence level corresponding to the target detection result based on the confidence level evaluation formula;

[0182] The third acquisition module is used to acquire the communication information of the target user;

[0183] The push module is used to push the confidence level to the target user based on the communication information.

[0184] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0185] In some optional implementations of this embodiment, the AI-based image detection device further includes:

[0186] The fourth acquisition module is used to acquire the specified classification result corresponding to the verification result if the verification result is that the verification failed.

[0187] The fifth acquisition module is used to acquire a specified information template corresponding to the specified classification result;

[0188] The third generation module is used to generate corresponding error category reminder information based on the specified classification result and the specified information template;

[0189] The display module is used to display the error category alert information.

[0190] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0191] In some optional implementations of this embodiment, the AI-based image detection device further includes:

[0192] The acquisition module is used to acquire a preset number of correctly captured images of a specified joint;

[0193] The extraction module is used to extract features from the specified joint image using a preset extraction model to obtain an image feature vector corresponding to the specified joint image.

[0194] A storage module is used to store the image feature vectors into the feature database.

[0195] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based image detection method in the aforementioned embodiments, and will not be repeated here.

[0196] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0197] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0198] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0199] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for image detection methods based on artificial intelligence. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0200] In some embodiments, the processor 42 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions for the artificial intelligence-based image detection method.

[0201] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0202] Compared with the prior art, the embodiments of this application have the following main advantages:

[0203] In this embodiment, a target page with pre-generated embeddings based on resource location embedding rules is first loaded; then, a target interface corresponding to the target page is called; subsequently, business embedding data corresponding to the target resource location in the target page is collected from the user within a preset time period based on the target interface; subsequently, product conversion data corresponding to the target resource location is generated based on the business embedding data; finally, product replacement processing is performed on the products within the target resource location based on the product conversion data. This embodiment uses a pre-built detection model to detect joint images, enabling rapid and accurate generation of detection results corresponding to the joint images. This effectively solves the problems of low detection efficiency and high detection cost in existing joint movement function assessment methods, improving the detection efficiency of joint images and ensuring the accuracy of the generated joint image detection results. Furthermore, before using the detection model to detect joint images, this application intelligently uses a feature database to filter irrelevant images from the joint images and uses a preset multi-classification model to perform compliance verification on the joint images. This quality control judgment of the joint images to be processed ensures the compliance of the subsequently processed joint images, which helps reduce the model error rate of subsequent image detection processing.

[0204] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the artificial intelligence-based image detection method described above.

[0205] Compared with the prior art, the embodiments of this application have the following main advantages:

[0206] In this embodiment, a target page with pre-generated embeddings based on resource location embedding rules is first loaded; then, a target interface corresponding to the target page is called; subsequently, business embedding data corresponding to the target resource location in the target page is collected from the user within a preset time period based on the target interface; subsequently, product conversion data corresponding to the target resource location is generated based on the business embedding data; finally, product replacement processing is performed on the products within the target resource location based on the product conversion data. This embodiment uses a pre-built detection model to detect joint images, enabling rapid and accurate generation of detection results corresponding to the joint images. This effectively solves the problems of low detection efficiency and high detection cost in existing joint movement function assessment methods, improving the detection efficiency of joint images and ensuring the accuracy of the generated joint image detection results. Furthermore, before using the detection model to detect joint images, this application intelligently uses a feature database to filter irrelevant images from the joint images and uses a preset multi-classification model to perform compliance verification on the joint images. This quality control judgment of the joint images to be processed ensures the compliance of the subsequently processed joint images, which helps reduce the model error rate of subsequent image detection processing.

[0207] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, 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 is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0208] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. An image detection method based on artificial intelligence, characterized in that, Includes the following steps: Obtain joint images of the target user; Call the preset feature database; Irrelevant images are filtered out from the joint images based on the feature database to obtain the corresponding target images; The target image is subjected to compliance verification based on a preset multi-classification model to obtain the verification result corresponding to the target image. If the verification result is successful, the target image is detected based on the preset detection model to generate a target detection result corresponding to the target image; The step of filtering out irrelevant images from the joint images based on the feature database to obtain the corresponding target images specifically includes: Construct a first feature vector corresponding to the joint image; The similarity between the first feature vector and all the second feature vectors included in the feature database is calculated to generate the similarity between the first feature vector and each of the second feature vectors. Based on the similarity, obtain the target similarity with the highest numerical value corresponding to the first feature vector; Irrelevant images are filtered out from the joint images based on the target similarity. The irrelevant images in the joint images are filtered out to obtain the target image after filtering. Specifically, the multi-classification model is a four-classification model trained based on the DenseNet model. It collects images of various different categories as sample data and performs corresponding classification and labeling for each sample data, such as compliance, incomplete exposure, incorrect orientation, and incorrect angle, to obtain a training set. Then, the DenseNet model is trained using this training set to generate the required four-classification model.

2. The image detection method based on artificial intelligence according to claim 1, characterized in that, The step of filtering irrelevant images from the joint images based on the target similarity specifically includes: Obtain the specified similarity corresponding to the third feature vector; wherein, the third feature vector is any one of all the first feature vectors; Determine whether the specified similarity is less than a preset similarity threshold; If so, obtain the specified image corresponding to the third feature vector; The specified image is used as an irrelevant image in the joint image.

3. The image detection method based on artificial intelligence according to claim 1, characterized in that, The step of performing compliance verification on the target image based on a preset multi-classification model to obtain the verification result corresponding to the target image specifically includes: Invoke the multi-classification model; The target image is input into the multi-classification model; The target image is classified using the multi-classification model to obtain a classification result corresponding to the target image. Based on the classification results, a verification result corresponding to the target image is generated.

4. The image detection method based on artificial intelligence according to claim 1, characterized in that, After the step of detecting the target image based on a preset detection model and generating a target detection result corresponding to the target image, the method further includes: Obtain the preset confidence level assessment formula; The confidence level corresponding to the target detection result is generated based on the confidence level evaluation formula; Obtain the communication information of the target user; Based on the communication information, the confidence level is pushed to the target user.

5. The image detection method based on artificial intelligence according to claim 1, characterized in that, After the step of performing compliance verification on the target image based on a preset multi-classification model to obtain the verification result corresponding to the target image, the method further includes: If the verification result is that the verification fails, obtain the specified classification result corresponding to the verification result; Obtain the specified information template corresponding to the specified classification result; Based on the specified classification result and the specified information template, generate corresponding error category reminder information; Display the error category alert information.

6. The image detection method based on artificial intelligence according to claim 1, characterized in that, Before the step of calling the preset feature database, the following is also included: Collect a preset number of correctly captured images of the specified joints; The specified joint image is subjected to feature extraction using a preset extraction model to obtain an image feature vector corresponding to the specified joint image; The image feature vector is stored in the feature database.

7. An image detection device based on artificial intelligence, characterized in that, include: The first acquisition module is used to acquire joint images of the target user; The calling module is used to call the preset feature database; The processing module is used to filter out irrelevant images from the joint images based on the feature database to obtain the corresponding target images; The verification module is used to perform compliance verification on the target image based on a preset multi-classification model, and obtain the verification result corresponding to the target image. The first generation module is used to detect the target image based on a preset detection model and generate a target detection result corresponding to the target image if the verification result is a successful verification. The processing module includes: A submodule is constructed to construct a first feature vector corresponding to the joint image; The calculation submodule is used to perform similarity calculation between the first feature vector and all the second feature vectors included in the feature database to generate the similarity between the first feature vector and each of the second feature vectors. The first acquisition submodule is used to acquire the target similarity with the highest value corresponding to the first feature vector based on the similarity. A filtering submodule is used to filter out irrelevant images from the joint images based on the target similarity; The processing submodule is used to filter out the irrelevant images in the joint image to obtain the target image after filtering. Specifically, the multi-classification model is a four-classification model trained based on the DenseNet model. It collects images of various different categories as sample data and performs corresponding classification and labeling for each sample data, such as compliance, incomplete exposure, incorrect orientation, and incorrect angle, to obtain a training set. Then, the DenseNet model is trained using this training set to generate the required four-classification model.

8. A computer device comprising a memory and a processor, the memory storing computer-readable instructions, wherein the processor, when executing the computer-readable instructions, implements the steps of the artificial intelligence-based image detection method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based image detection method as described in any one of claims 1 to 6.

Images