Data processing method and system

Abstract

This application discloses a data processing method and system. The method comprises: receiving a model training request; obtaining multiple sets of training samples and an initial model from the model training request, wherein at least one set of training samples includes multiple frames of training images and reference coordinates for the multiple frames of training images, the reference coordinates being determined based on segmented features of target objects contained in the multiple frames of training images; training the initial model using the multiple sets of training samples to obtain a positioning model; and outputting the positioning model. This application addresses the technical problem that data processing methods in related arts cannot achieve rapid retrieval of massive medical images.

Description

Technical Field

[0001] The present application relates to the field of medical image processing, and more specifically, to a data processing method and system. Background Art

[0002] In the medical field, it is necessary to search massive medical image databases for images of areas or lesions of interest. For CT (Computed Tomography) images, the "BodyPartExamined" field of DICOM (Digital Imaging and Communications in Medicine) can be used to filter out image sequences of the corresponding area, and then each sequence must be browsed one by one to find the segments of interest. This is a time-consuming and labor-intensive process. In addition, the image data stored in the image database comes from medical equipment from different manufacturers, and different lesions are often filmed using different methods. This results in the "SliceLocation" field of a single cross-sectional image in the CT image sequence being inconsistent. This method is not suitable for retrieving such large-scale image datasets with complex sources.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] The embodiments of the present application provide a data processing method and system to at least solve the technical problem that the data processing methods in the related art cannot achieve rapid retrieval of massive medical images.

[0005] According to one aspect of an embodiment of the present application, a data processing method is provided, including: receiving a model training request; obtaining multiple sets of training samples and an initial model from the model training request, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on segmented features of the target objects contained in the multiple frames of training images; training the initial model using the multiple sets of training samples to obtain a positioning model; and outputting the positioning model.

[0006] According to another aspect of an embodiment of the present application, a data processing method is also provided, including: obtaining multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; using the multiple groups of training samples to train the initial model to obtain a positioning model.

[0007] According to another aspect of an embodiment of the present application, a data processing method is also provided, including: receiving a sequence of images to be processed; processing the sequence of images to be processed using a positioning model to obtain positioning results of multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning results of at least one frame of images to be processed are determined based on segmented features of a target object contained in at least one frame of images to be processed; and outputting the positioning results of multiple frames of images to be processed.

[0008] According to another aspect of an embodiment of the present application, a data processing method is also provided, including: obtaining a sequence of images to be processed; processing the sequence of images to be processed using a positioning model to obtain positioning results of multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning result of at least one frame of images to be processed is determined based on the segmented features of the target object contained in at least one frame of images to be processed.

[0009] According to another aspect of an embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned data processing method.

[0010] According to another aspect of an embodiment of the present application, a computer terminal is further provided, comprising: a memory and a processor, wherein the processor is configured to run a program stored in the memory, wherein the above-mentioned data processing method is executed when the program is run.

[0011] According to another aspect of an embodiment of the present application, a data processing system is also provided, including: a processor; and a memory, connected to the processor, for providing the processor with instructions for processing the following processing steps: receiving a model training request; obtaining multiple sets of training samples and an initial model of the model training request, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; training the initial model using the multiple sets of training samples to obtain a positioning model; and outputting the positioning model.

[0012] In an embodiment of the present application, a trained positioning model can be used to obtain the positioning results of each frame of the image, and the purpose of rapid retrieval of massive medical images can be achieved based on the positioning results. Moreover, at least one group of training samples used in the model training contains reference coordinates of multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target object, so that the positioning model is independent of the device and imaging method. In addition, the segmented features are referenced during the training process, and supervised training of the positioning model is achieved, so that the model can obtain consistent prediction results for the same input sequence after each training, achieving the technical effect of improving the stability of the positioning model processing results, thereby solving the technical problem that the data processing methods in the related technology cannot achieve rapid retrieval of massive medical images. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0014] Figure 1 It is a schematic diagram of human body Z-direction positioning according to the prior art;

[0015] Figure 2 is a hardware structure block diagram of a computer terminal (or mobile device) for implementing a data processing method according to an embodiment of the present application;

[0016] Figure 3 is a schematic diagram of a computer terminal (or mobile device) serving as a server according to an embodiment of the present application;

[0017] Figure 4 is a flowchart of a first data processing method according to an embodiment of the present application;

[0018] Figure 5 is a schematic diagram of an optional human body part positioning model according to an embodiment of the present application;

[0019] Figure 6 This is a schematic diagram of an optional marking of the center points of each segment of the spine in the sagittal plane according to an embodiment of the present application;

[0020] Figure 7 is a flowchart of an optional data processing method according to an embodiment of the present application;

[0021] Figure 8 is a schematic diagram of an optional CT image sequence regression coordinate visualization result according to an embodiment of the present application;

[0022] Figure 9 is a flowchart of a second data processing method according to an embodiment of the present application;

[0023] Figure 10 is a flowchart of a third data processing method according to an embodiment of the present application;

[0024] Figure 11 is a flowchart of a fourth data processing method according to an embodiment of the present application;

[0025] Figure 12 is a schematic diagram of a first data processing device according to an embodiment of the present application;

[0026] Figure 13 is a schematic diagram of a second data processing device according to an embodiment of the present application;

[0027] Figure 14 is a schematic diagram of a third data processing device according to an embodiment of the present application;

[0028] Figure 15 is a schematic diagram of a fourth data processing device according to an embodiment of the present application; and

[0029] Figure 16 This is a structural block diagram of a computer terminal according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following interpretations:

[0033] CT imaging: Using precisely collimated X-ray beams, gamma rays, ultrasound, etc., together with highly sensitive detectors, one section after another is scanned around a certain part of the human body. Each section image is regarded as a frame of image, so that a complete CT image sequence can be obtained.

[0034] Batch Normalization: Batch Normalization, referred to as BN, can be used to normalize the corresponding network response through minni-batch during each stochastic gradient descent training of the model, so that the mean of the result (each dimension of the output signal) is 0 and the variance is 1.

[0035] S function (sigmoid): The curve of this function is similar to the letter S. Commonly used S functions include: logical function, hyperbolic sine function, inverse tangent function, Goodman function, error function, etc. This application does not make specific restrictions on this.

[0036] Distance regression loss function: It can be divided into L1 loss function (minimum absolute error), L2 loss function (minimum square error), and smooth L1 loss function. Among them, the smooth L1 loss function combines the L1 loss function and the L2 loss function. It is a piecewise function. When it is less than 1, it is the L2 loss function, and when it is greater than 1, it is the L1 loss function.

[0037] Adam (Adaptive Moment Estimation): The learning rate of each parameter can be dynamically adjusted using the first-order moment estimation and second-order moment estimation of the gradient.

[0038] In order to meet the purpose of rapid retrieval of massive medical images, the related art provides the following technical solutions: using image sequences to train a human body part positioning model, establish a unified Z-direction (i.e., inferior-superior direction) coordinate axis that is independent of the device and the shooting method, such as Figure 1 As shown, each frame of the sequence corresponds to a coordinate on the coordinate axis.

[0039] Based on this, two specific approaches can be adopted. The first uses a twin network to classify relationships between sequences. Two frames (with arbitrary spacing) are fed simultaneously, and the model automatically determines their coordinate size, outputting two categories: large and small. The second approach uses a self-supervised approach. Multiple frames are fed continuously (with equal spacing), and the model automatically regresses the coordinates of each frame. The regressed coordinates satisfy the spatial distribution and size relationship.

[0040] Although the twin network method is simple to operate, its modeling capability is not strong because it only compares the relationship between two frames of images each time. In order to ensure that the model can learn the relationship between the entire sequence, a better sampling method is needed as support. Using the self-supervised method, each time multiple frames of images are input, as long as there is overlap between the multiple frames during sampling, the coordinate relationship can be recursively inferred to the entire sequence, so the modeling capability is stronger; however, in order to learn the spatial relationship between multiple frames of images, it is necessary to find a more effective loss function to evaluate the regression results. In addition, the above two methods are completely unsupervised, and no constraints are added to the regression coordinate results, which will result in large differences between the prediction results obtained by inputting the same sequence after each model training.

[0041] In order to solve the above problems, the present application provides a method for localizing human body parts in CT images with reference to spinal segmentation. It can not only utilize the spatial relationship between sequences in an unsupervised manner, but also refer to the segmented features of the human spine to correct the regressed coordinates in a supervised manner, so that the model can obtain consistent prediction results for the same input sequence after each training, thereby improving the stability of the regression results.

[0042] Example 1

[0043] According to an embodiment of the present application, a data processing method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0044] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 2 FIG1 shows a hardware structure block diagram of a computer terminal (or mobile device) for implementing a data processing method. Figure 2 As shown, the computer terminal 10 (or mobile device 10) may include one or more (illustrated as 102a, 102b, ..., 102n) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 2 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 2 More or fewer components than shown, or with Figure 2 Different configurations shown.

[0045] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10 (or mobile device 10). As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0046] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device of the data processing method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above-mentioned data processing method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0047] The transmission device 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0048] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device 10).

[0049] Figure 2 The hardware structure block diagram shown can be used not only as an exemplary block diagram of the above-mentioned computer terminal 10 (or mobile device 10), but also as an exemplary block diagram of the above-mentioned server. In an optional embodiment, Figure 3 The block diagram shows the use of the above Figure 2 The computer terminal 10 (or mobile device 10) shown in FIG. 1 is an embodiment of a server. Figure 3 As shown, the computer terminal 10 (or mobile device 10) can be connected to one or more clients 20 via a data network connection or electronic connection. The data network connection can be a local area network connection, a wide area network connection, an Internet connection, or other types of data network connections. The computer terminal 10 (or mobile device) can provide network-based user services for the client, such as social networks, cloud resources, email, online payment or other online applications.

[0050] Under the above operating environment, this application provides Figure 4 The data processing method shown. Figure 4 This is a flow chart of the first data processing method according to an embodiment of the present application. Figure 4 As shown, the method may include the following steps:

[0051] Step S402: Receive a model training request.

[0052] In order to train a model with high processing accuracy, it is often necessary to use a large number of training samples for multiple trainings, and the amount of data and computing power of the entire training process is large. In order to reduce the resource consumption of user devices (such as smartphones, tablets, laptops, PDAs, personal computers, etc.), the model training can be performed by the server, and only the trained model is deployed in the user device for the convenience of user use. Furthermore, in order to significantly reduce the computing burden of the user device, the trained model can be directly deployed in the server. The user device connects to the server through a specific interface and sends the data to be processed to the server. The server uses the deployed model to process the data to be processed and feeds back the processing results to the user device.

[0053] The model training request in the above steps can be generated according to the user's model usage requirements. The request can carry the type of data that needs to be processed and the expected processing results, etc. For example, in the medical field, the type of data that needs to be processed may be CT images, and the expected processing results may be human body part positioning.

[0054] Because the server can provide model training services for different users, with varying model usage requirements and varying specific model structures and training samples, an optional embodiment can provide an interactive interface on the user device, allowing the user to enter a model training request in an input area, which allows the user device to send the model training request to the server via the network. For greater targeting, the server can provide different model training options based on the user's type, allowing the user to select a solution in the input area. The user device can then generate a model training request based on the user's selection and send it to the server via the network.

[0055] Step S404, obtaining multiple sets of training samples and initial models requested for model training, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on segmented features of the target objects contained in the multiple frames of training images.

[0056] The training samples in the above steps can be a large number of medical images used to train the initial model, for example, CT images, but not limited to this. Multiple frames of medical images of the same patient can be used as a set of training samples. The target object can refer to the human spine, which is composed of multiple segments. Moreover, the spinal segments will not change due to different equipment and imaging methods. Therefore, in an embodiment of the present application, a unified coordinate axis that is independent of the equipment and the photographing method can be constructed based on the segmented characteristics of the human spine, and each frame of the CT image can be mapped to the coordinate axis to obtain the above-mentioned reference coordinates.

[0057] To improve the accuracy of the positioning model, training samples can be constructed using a large number of CT image sequences from different patients. CT image sequences from different patients contain different body parts. For CT image sequences containing the spine, they can be pre-labeled to determine the reference coordinates of each frame in the CT image sequence. In this case, training samples containing multiple CT images and their reference coordinates can be obtained. For CT image sequences that do not contain the spine, they can be pre-labeled to set the positioning result label for each frame in the CT image sequence. In this case, training samples containing multiple CT images but without reference coordinates can be obtained.

[0058] It should be noted that a greater number of training image frames results in stronger modeling capabilities, but also requires greater computational effort and data. To improve modeling capabilities while reducing the computational effort and data required for training, a fixed number of training image frames can be determined experimentally in advance, for example, 5 frames. For greater specificity, the number of training image frames can be determined based on the model processing accuracy required by the user. This means that the number of frames can be dynamically updated based on model training requests and is no longer a fixed value.

[0059] The initial model in the above steps can be a model built using a deep convolutional neural network. The specific structure of the model can adopt an existing structure, and this application does not make specific restrictions on this. For example, Figure 5 As shown, the initial model can be a deep convolutional neural network including a convolutional layer, a batch normalization layer, an activation layer, a global average pooling layer and a fully connected layer, wherein multiple frames of training images are input into the deep convolutional neural network, the features of each frame of training image are extracted, and then the number of feature layers is fixed to m (i.e., the number of frames of the above-mentioned training image) through a convolutional layer, followed by a BN batch normalization layer and a ReLU activation layer, and then through a global average pooling layer and a fully connected layer to output m predicted coordinates, corresponding to each frame of training image.

[0060] Step S406: train the initial model using multiple sets of training samples to obtain a positioning model.

[0061] Optionally, an adaptive moment estimation algorithm can be used to train the initial model based on the training samples to obtain a positioning model. The initial learning rate and number of iterations during the training process can be dynamically adjusted according to actual training requirements, or can be set to a fixed value through experiments. For example, the initial learning rate of the initial model is 0.001, and the number of iterations of the initial model is 100, but not limited to this.

[0062] In an optional embodiment, during the training of the initial model, an Adam algorithm can be used for iterative optimization, with an initial learning rate of 0.001 and a total of 100 iterations. During each iteration, a training sample can be input into the initial model to obtain the predicted coordinates of each training frame. The model parameters are then updated based on the reference coordinates of each training frame, and this process continues until 100 iterations are reached.

[0063] Step S408: output the positioning model.

[0064] In an optional embodiment, if the positioning model needs to be deployed in the user device, the server can transmit the positioning model to the user device through the network; if the positioning model needs to be deployed in the server, the positioning model can be directly put online, so that the user can use the online positioning model to locate human body parts, thereby achieving the purpose of rapid retrieval of massive medical images.

[0065] Through the technical solution provided by the above-mentioned embodiment of the present application, after receiving a model training request, multiple sets of training samples and initial models corresponding to the model training request can be obtained, and the initial model can be further trained using multiple sets of training samples to achieve the purpose of positioning model training. It is easy to notice that the positioning result of each frame of the image can be obtained using the trained positioning model, and the purpose of rapid retrieval of massive medical images can be achieved based on the positioning result. Moreover, at least one set of training samples used in the model training contains the reference coordinates of multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target object, so that the positioning model is independent of the device and the imaging method. In addition, the segmented features are referenced during the training process, and supervised training of the positioning model is achieved, so that the model can obtain consistent prediction results for the same input sequence each time it is trained, achieving the technical effect of improving the stability of the positioning model processing results, thereby solving the technical problem that the data processing method in the related technology cannot achieve rapid retrieval of massive medical images.

[0066] In the above embodiment of the present application, before obtaining multiple groups of training samples requested by model training, the method also includes: obtaining a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; determining the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation features of the target object; randomly extracting a preset number of target images from the target image sequence; and generating at least one group of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0067] The image database in the above steps may be a massive medical image database, which stores a large number of image sequences. The devices and photographing methods for different image sequences may be the same or different.

[0068] To improve the processing accuracy of the positioning model, the training samples need to cover different parts of the human body. Therefore, in the embodiments of the present application, a preset condition for screening the training data set can be pre-set. By using this preset condition, a data set that meets the training requirements can be screened from the massive medical image database. For example, the preset condition can be that the sequence length is greater than 10 frames and the sequence contains the entire human spine, but the present invention is not limited to this.

[0069] The preset number of frames in the above steps can be a fixed number of frames, for example, 5 frames; or it can be a number of frames determined according to the model processing accuracy required by the user. In this case, the preset number of frames is a dynamically changing value.

[0070] For patients with different lesions, the image sequences they take contain different parts of the human body. Therefore, the number of image sequences containing the complete human spine in the massive medical image database is limited. In order to expand the data volume of training samples, the screened image sequences are not directly used as training samples. Instead, some images in the image sequence are randomly selected as training images.

[0071] For example, taking CT images as an example, all CT image sequences with a sequence length greater than 10 frames can be screened from the image database as the experimental data set. CT image sequences containing the complete human spine can be further screened as the annotation set. First, a coordinate system is established based on the human spine. Based on the correspondence between each frame of CT image and each segment of the spine, each frame of CT image is mapped to this coordinate system to determine the reference coordinates of each frame of CT image. Furthermore, multiple frames of CT images are randomly selected from each CT image sequence as training samples, and the training samples also contain the reference coordinates of the CT images.

[0072] In the above-mentioned embodiment of the present application, based on the segmentation characteristics of the target object, determining the reference coordinates of multiple frames of target images in the target image sequence includes: labeling multiple segments of the target object; determining the frame numbers of the target images where the center points of the multiple segments are located, and obtaining the frame numbers corresponding to the multiple segments; determining the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0073] For a target image sequence, each segment of the spine can be marked on the sagittal plane. Since each segment of the spine has a certain length, in the embodiment of the present application, the center point of each segment can be marked. Specifically, the existing spine segmentation method can be used for marking, such as Figure 6 As shown in the figure, the center points of each segment can be marked from S1 to C7, where C represents the cervical segment, T represents the thoracic segment, L represents the lumbar segment, and S represents the sacral segment. The smaller the number, the closer the segment is to the head.

[0074] In an optional embodiment, after acquiring the target image sequence, the center points of each spinal segment can be marked on the sagittal plane to obtain the marking results of each segment, and the frame number of the image where the center point is located is used as the frame number corresponding to the segment. A coordinate system is pre-established for the human spine, and the center point of each segment corresponds to a coordinate in the coordinate system, so that the frame number corresponding to each segment can be mapped to the corresponding reference coordinate. For the target image between two adjacent segments, the frame number of the target image can be combined with the frame numbers corresponding to the two adjacent segments to determine the reference coordinate of the target image based on the proportional relationship.

[0075] For example, still taking CT images as an example, after acquiring the CT image sequence, the center points of each segment can be marked as S1 to C7 on the sagittal plane, and the frame number of the CT image where the center point is located is: Z = (z S1 , z L5 ,...,z T1 , z C7 ), further combined with the frame number of each frame of CT image, the reference coordinates of each frame of CT image can be determined.

[0076] In the above embodiment of the present application, based on the annotation results of multiple segments, determining the reference coordinates corresponding to multiple segments includes: obtaining the reference coordinates corresponding to multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0077] In the embodiment of the present application, for ease of calculation, the segment L1 located in the middle of the spine can be selected as the origin to establish a coordinate system. In this case, the frame number corresponding to the segment L1 corresponds to the reference coordinate 0. The preset coordinate interval can be the change between the coordinates of two adjacent segments. For example, it can be 2, that is, starting from the target segment, the coordinates corresponding to each segment of the head increase by 2, and starting from this segment, the coordinates corresponding to each segment of the feet decrease by 2. However, the present invention is not limited to this and can be set according to actual needs.

[0078] For example, still taking CT images as an example, the annotation results of CT image sequences are as follows: Figure 6 As shown, a coordinate system is established with segment L1 as the origin, and the reference coordinate of the CT image where the center point of segment L1 is located is 0. On this basis, it can be obtained that: the reference coordinate of the CT image where the center point of segment L2 is located is -2, the reference coordinate of the CT image where the center point of segment L3 is located is -4, the reference coordinate of the CT image where the center point of segment L4 is located is -6, the reference coordinate of the CT image where the center point of segment L5 is located is -8, the reference coordinate of the CT image where the center point of segment S1 is located is -10, the reference coordinate of the CT image where the center point of segment T12 is located is 2, the reference coordinate of the CT image where the center point of segment T11 is located is 4, the reference coordinate of the CT image where the center point of segment T10 is located is 6, the reference coordinate of the CT image where the center point of segment T9 is located is 8, and the reference coordinate of the CT image where the center point of segment T8 is located is 10. The same applies to other segments and will not be repeated here.

[0079] In the above embodiment of the present application, based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments, determining the reference coordinates of the multiple frames of target images includes: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame of the target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and a first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0080] The first target image in the above steps may be a target image located between two adjacent segments, the first segment may be the segment with the smaller frame number of the two adjacent segments, and the second segment may be the segment with the larger frame number of the two adjacent segments.

[0081] In an optional embodiment, for other target images between target images where the center points of two adjacent segments are located, the reference coordinates s(i) may be calculated using the following formula:

[0082]

[0083] Among them, z s is the frame number corresponding to the first segment, z e is the frame number corresponding to the second segment, s(z s ) is the reference coordinate corresponding to the first segment.

[0084] In the above embodiment of the present application, randomly extracting a preset number of target images from a target image sequence includes: obtaining an initial frame number and an extraction interval; and extracting images from the target image sequence based on the initial frame number and the extraction interval to obtain a preset number of target images.

[0085] The initial frame number in the above steps can be the frame number of the image from which extraction begins, denoted by j. The extraction interval can be the frame number interval between two adjacent target frames, denoted by k. Different j and k values ​​can be set for different random samples, allowing for the acquisition of a large number of training samples with overlap between them.

[0086] In an optional embodiment, as Figure 5 As shown, the initial frame number j can be randomly selected from the target image sequence, and target images with a preset frame number m can be randomly extracted at a certain interval k. The frame numbers of the extracted target images are: j, j+k, j+2k, ..., j+(m-1)k.

[0087] In the above embodiment of the present application, generating at least one set of training samples based on the target image of a preset number of frames and the reference coordinates of the target image of the preset number of frames includes: flipping, scaling or rotating the target image of the preset number of frames to obtain the target image after the operation; generating at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0088] In an optional embodiment, in order to further expand the training samples, broaden the scope of application of the positioning model, and avoid model overfitting, data augmentation operations can be performed on the target image, such as random upside-down flipping, random scaling, and random rotation, where the probability of flipping is 0.5, the scaling size is 0.5-1.5 times, and the rotation angle is -90°-90°.

[0089] It should be noted that the above operation needs to process all randomly extracted target images at the same time, rather than processing part of the extracted target images.

[0090] In the above embodiment of the present application, after obtaining a target image sequence that meets preset conditions in an image database, the method also includes: adjusting the window width and window position of multiple frames of target images; adjusting the resolution of multiple frames of target images to the target resolution through a bilinear interpolation method; and adjusting the size of multiple frames of target images to the target size.

[0091] The target resolution in the above steps can be set according to actual training needs, for example, it can be 0.1mm / pixel, but not limited to this. The target size can also be set according to actual training needs, for example, it can be 224*224, but not limited to this.

[0092] In an optional embodiment, after screening the target image sequence, the window width and window position of different parts of the CT image sequence can be adjusted to the corresponding window width and window position; the resolution of all target images can be unified through bilinear interpolation; and the interpolated target image can be scaled to obtain the image data required for model training.

[0093] In the above embodiment of the present application, the initial model is trained using the training samples to obtain the positioning model, which includes: inputting the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; obtaining the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; determining the loss value of the initial model based on the coordinate difference, the reference coordinates and the predicted coordinates of the multiple frames of training images; and updating the parameters of the initial model based on the loss value to obtain the positioning model.

[0094] In an optional embodiment, in order to improve the stability of the positioning model processing results, a supervised loss function can be introduced in the training process of the initial model, so that the model training process is completed by combining the unsupervised loss function and the supervised loss function.

[0095] The coordinate difference can be obtained by the following formula: Δ i，j+1 =s i，j+1 -s i，j , where s i，j+1 and s i，j is the predicted coordinates of the target image in two adjacent frames.

[0096] In the above embodiment of the present application, based on the coordinate difference, the reference coordinates and the predicted coordinates of the multi-frame training images, determining the loss value of the initial model includes: determining the constant error based on the difference between the coordinate difference and the second preset value; inputting the coordinate difference into the S function to determine the sorting error; inputting the coordinate difference into the distance regression loss function to determine the interval error; inputting the reference coordinates and the predicted coordinates of the multi-frame training images into the distance regression loss function to determine the regression error; obtaining the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0097] In an optional embodiment, the constant error Loss can be calculated by the following formula: const , sorting error Loss order , Interval Error Loss dist and regression error Loss annot , where the constant error requirement Δ i，j+1 >b, the sorting error requires the predicted coordinates to satisfy s1<s2<s3<...<s m , the interval error requires the predicted coordinates to satisfy s2-s1=s3-s2=s m -s m-1 :

[0098]

[0099]

[0100]

[0101]

[0102] Wherein, N represents the data batch size of the training samples in each iteration, for example, it can be a constant of 8, but not limited to this; b represents the second preset value, for example, b can be a constant of 0.1, k can be a constant of 2, but not limited to this; A represents the data size of the annotation set during the training process, a i，j Indicates the reference coordinates.

[0103] The final loss value can be the sum of the above errors:

[0104] Loss=Loss const +Loss order +Loss dist +Loss annot .

[0105] The following combination Figure 7 and Figure 8 Taking CT images as an example, an optional embodiment of the present application is described in detail. Figure 7 As shown, the method may include the following steps:

[0106] Step S71, image collection;

[0107] Optionally, CT image sequences that meet the requirements are screened from a medical image database as a data set.

[0108] Step S72, marking the center points of the spinal segments;

[0109] Optionally, a CT image sequence containing a complete human spine is selected from the data set as a labeling set, the center point of each segment of the spine is labeled on the sagittal plane, and the frame number of the center point of each segment is obtained.

[0110] Step S73, data preprocessing;

[0111] Optionally, the window width and window level of the CT image sequence are adjusted, the resolution of all CT images is unified, and the size is scaled uniformly to obtain the target image for training. The annotation set is then parsed to generate the reference coordinates of each frame of the target image.

[0112] Step S74, building a model;

[0113] Optionally, a deep convolutional neural network is used to build a human body part positioning model, and m frames of images are randomly selected from the CT image sequence. The m frames of images are then input into the deep convolutional neural network for feature extraction, and m predicted coordinates are further obtained.

[0114] Step S75: model training.

[0115] Optionally, unsupervised loss function and supervised loss function are used in the training process, and the Adam algorithm is used for iterative optimization. After the training is completed, it is determined whether the model has converged. If not, the model training is continued. If converged, the training process is terminated to obtain a trained positioning model. The coordinate visualization results obtained by regression on a complete CT image sequence are as follows: Figure 8 shown.

[0116] The solution provided by the above steps provides a method for localizing human body parts in CT images with reference to spinal segmentation. In addition to unsupervised utilization of the spatial relationship between sequences, the segmented features of the human spine are also referenced to perform supervised corrections to the regressed coordinates. This ensures that the model can obtain consistent prediction results for the same input sequence after each training, thereby improving the stability of the regression results.

[0117] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0118] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it 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 the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0119] Example 2

[0120] According to an embodiment of the present application, a data processing method is also provided.

[0121] Figure 9 Flowchart of the second data processing method according to the embodiment of the present application. Figure 9 As shown, the method may include the following steps:

[0122] Step S902 : obtaining multiple sets of training samples, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, where the reference coordinates are determined based on segmented features of target objects contained in the multiple frames of training images.

[0123] The training samples in the above steps can be a large number of medical images used to train the initial model, such as, but not limited to, CT images. Multiple frames of medical images from the same patient can serve as a set of training samples. The target object can be the human spine, which consists of multiple segments. Furthermore, spinal segmentation does not vary depending on the device or imaging method.

[0124] To improve the accuracy of the positioning model, training samples can be constructed using a large number of CT image sequences from different patients. CT image sequences from different patients contain different body parts. For CT image sequences containing the spine, they can be pre-labeled to determine the reference coordinates of each frame in the CT image sequence. In this case, training samples containing multiple CT images and their reference coordinates can be obtained. For CT image sequences that do not contain the spine, they can be pre-labeled to set the positioning result label for each frame in the CT image sequence. In this case, training samples containing multiple CT images but without reference coordinates can be obtained.

[0125] It should be noted that a greater number of training image frames results in stronger modeling capabilities, but also requires greater computational effort and data. To improve modeling capabilities while reducing the computational effort and data required for training, a fixed number of training image frames can be determined experimentally in advance, for example, 5 frames. For greater specificity, the number of training image frames can be determined based on the model processing accuracy required by the user. This means that the number of frames can be dynamically updated based on model training requests and is no longer a fixed value.

[0126] The initial model in the above steps can be a model built using a deep convolutional neural network. The specific structure of the model can adopt an existing structure, and this application does not make any specific limitations on this.

[0127] Step S904: train the initial model using multiple sets of training samples to obtain a positioning model.

[0128] Optionally, an adaptive moment estimation algorithm can be used to train the initial model based on the training samples to obtain a positioning model. The initial learning rate and number of iterations during the training process can be dynamically adjusted according to actual training requirements, or can be set to a fixed value through experiments. For example, the initial learning rate of the initial model is 0.001, and the number of iterations of the initial model is 100, but not limited to this.

[0129] In the above embodiment of the present application, before obtaining multiple groups of training samples, the method also includes: obtaining a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; determining the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation features of the target object; randomly extracting a preset number of target images from the target image sequence; and generating at least one group of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0130] In the above-mentioned embodiment of the present application, based on the segmentation characteristics of the target object, determining the reference coordinates of multiple frames of target images in the target image sequence includes: labeling multiple segments of the target object; determining the frame numbers of the target images where the center points of the multiple segments are located, and obtaining the frame numbers corresponding to the multiple segments; determining the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0131] In the above embodiment of the present application, based on the annotation results of multiple segments, determining the reference coordinates corresponding to multiple segments includes: obtaining the reference coordinates corresponding to multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0132] In the above embodiment of the present application, based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments, determining the reference coordinates of the multiple frames of target images includes: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame of the target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and a first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0133] In the above embodiment of the present application, randomly extracting a preset number of target images from a target image sequence includes: obtaining an initial frame number and an extraction interval; and extracting images from the target image sequence based on the initial frame number and the extraction interval to obtain a preset number of target images.

[0134] In the above embodiment of the present application, generating at least one set of training samples based on the target image of a preset number of frames and the reference coordinates of the target image of the preset number of frames includes: flipping, scaling or rotating the target image of the preset number of frames to obtain the target image after the operation; generating at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0135] In the above embodiment of the present application, after obtaining a target image sequence that meets preset conditions in an image database, the method also includes: adjusting the window width and window position of multiple frames of target images; adjusting the resolution of multiple frames of target images to the target resolution through a bilinear interpolation method; and adjusting the size of multiple frames of target images to the target size.

[0136] In the above embodiment of the present application, the initial model is trained using the training samples to obtain the positioning model, which includes: inputting the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; obtaining the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; determining the loss value of the initial model based on the coordinate difference, the reference coordinates and the predicted coordinates of the multiple frames of training images; and updating the parameters of the initial model based on the loss value to obtain the positioning model.

[0137] In the above embodiment of the present application, based on the coordinate difference, the reference coordinates and the predicted coordinates of the multi-frame training images, determining the loss value of the initial model includes: determining the constant error based on the difference between the coordinate difference and the second preset value; inputting the coordinate difference into the S function to determine the sorting error; inputting the coordinate difference into the distance regression loss function to determine the interval error; inputting the reference coordinates and the predicted coordinates of the multi-frame training images into the distance regression loss function to determine the regression error; obtaining the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0138] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0139] Example 3

[0140] According to an embodiment of the present application, a data processing method is also provided.

[0141] Figure 10 : is a flow chart of the third data processing method according to an embodiment of the present application. Figure 10 As shown, the method may include the following steps:

[0142] Step S1002: receiving an image sequence to be processed.

[0143] The image sequence to be processed in the above steps may be a medical image sequence, for example, a CT image sequence, but is not limited thereto.

[0144] In an optional embodiment, an interactive interface can be provided on the user device, allowing the user to upload the image sequence to be processed in an input area. The user can complete the upload process by directly dragging the image sequence to be processed into the area or by clicking the "Upload Image Sequence" button. If the positioning model needs to be deployed on the user device, the user device can directly process the image sequence to be processed; if the positioning model needs to be deployed on a server, the user device can send the image sequence to be processed to the server via the network, and the server will process it.

[0145] Step S1004 , using the positioning model to process the image sequence to be processed, and obtaining positioning results of multiple frames of the image sequence to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmentation features of the target object contained in the at least one frame of the image to be processed.

[0146] The target object in the above steps can be the human spine, which is composed of multiple segments. Furthermore, the segmentation of the spine does not change due to differences in equipment and imaging methods. The positioning model can be built using a deep convolutional neural network. The specific structure of the model can adopt existing structures and is not specifically limited in this application.

[0147] Step S1006: output the positioning results of multiple frames of images to be processed.

[0148] In an optional embodiment, if the positioning model needs to be deployed in the user device, the user device can directly display the positioning results in the display area of ​​the interactive interface; if the positioning model needs to be deployed in the server, the server can return the positioning results to the user device through the network, and the user device will display them in the display area of ​​the interactive interface. Furthermore, the user can retrieve images of interest from the massive medical image database based on the positioning results, thereby achieving the purpose of rapid retrieval of massive medical images.

[0149] In the above embodiment of the present application, the method also includes: obtaining multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; using the multiple groups of training samples to train the initial model to obtain a positioning model.

[0150] In the above embodiment of the present application, before obtaining multiple groups of training samples, the method also includes: obtaining a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; determining the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation features of the target object; randomly extracting a preset number of target images from the target image sequence; and generating at least one group of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0151] In the above-mentioned embodiment of the present application, based on the segmentation characteristics of the target object, determining the reference coordinates of multiple frames of target images in the target image sequence includes: labeling multiple segments of the target object; determining the frame numbers of the target images where the center points of the multiple segments are located, and obtaining the frame numbers corresponding to the multiple segments; determining the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0152] In the above embodiment of the present application, based on the annotation results of multiple segments, determining the reference coordinates corresponding to multiple segments includes: obtaining the reference coordinates corresponding to multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0153] In the above embodiment of the present application, based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments, determining the reference coordinates of the multiple frames of target images includes: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame of the target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and a first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0154] In the above embodiment of the present application, randomly extracting a preset number of target images from a target image sequence includes: obtaining an initial frame number and an extraction interval; and extracting images from the target image sequence based on the initial frame number and the extraction interval to obtain a preset number of target images.

[0155] In the above embodiment of the present application, generating at least one set of training samples based on the target image of a preset number of frames and the reference coordinates of the target image of the preset number of frames includes: flipping, scaling or rotating the target image of the preset number of frames to obtain the target image after the operation; generating at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0156] In the above embodiment of the present application, after obtaining a target image sequence that meets preset conditions in an image database, the method also includes: adjusting the window width and window position of multiple frames of target images; adjusting the resolution of multiple frames of target images to the target resolution through a bilinear interpolation method; and adjusting the size of multiple frames of target images to the target size.

[0157] In the above embodiment of the present application, the initial model is trained using the training samples to obtain the positioning model, which includes: inputting the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; obtaining the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; determining the loss value of the initial model based on the coordinate difference, the reference coordinates and the predicted coordinates of the multiple frames of training images; and updating the parameters of the initial model based on the loss value to obtain the positioning model.

[0158] In the above embodiment of the present application, based on the coordinate difference, the reference coordinates and the predicted coordinates of the multi-frame training images, determining the loss value of the initial model includes: determining the constant error based on the difference between the coordinate difference and the second preset value; inputting the coordinate difference into the S function to determine the sorting error; inputting the coordinate difference into the distance regression loss function to determine the interval error; inputting the reference coordinates and the predicted coordinates of the multi-frame training images into the distance regression loss function to determine the regression error; obtaining the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0159] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0160] Example 4

[0161] According to an embodiment of the present application, a data processing method is also provided.

[0162] Figure 11 Flowchart of the fourth data processing method according to the embodiment of the present application. Figure 11 As shown, the method may include the following steps:

[0163] Step S1102: Obtain an image sequence to be processed.

[0164] The image sequence to be processed in the above steps may be a medical image sequence, for example, a CT image sequence, but is not limited thereto.

[0165] Step S1104 , using the positioning model to process the image sequence to be processed, and obtaining positioning results of multiple frames of the image sequence to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmentation features of the target object contained in the at least one frame of the image to be processed.

[0166] The target object in the above steps can be the human spine, which is composed of multiple segments. Furthermore, the segmentation of the spine does not change due to differences in equipment and imaging methods. The positioning model can be built using a deep convolutional neural network. The specific structure of the model can adopt existing structures and is not specifically limited in this application.

[0167] In the above embodiment of the present application, the method also includes: obtaining multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; using the multiple groups of training samples to train the initial model to obtain a positioning model.

[0168] In the above embodiment of the present application, before obtaining multiple groups of training samples, the method also includes: obtaining a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; determining the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation features of the target object; randomly extracting a preset number of target images from the target image sequence; and generating at least one group of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0169] In the above-mentioned embodiment of the present application, based on the segmentation characteristics of the target object, determining the reference coordinates of multiple frames of target images in the target image sequence includes: labeling multiple segments of the target object; determining the frame numbers of the target images where the center points of the multiple segments are located, and obtaining the frame numbers corresponding to the multiple segments; determining the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0170] In the above embodiment of the present application, based on the annotation results of multiple segments, determining the reference coordinates corresponding to multiple segments includes: obtaining the reference coordinates corresponding to multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0171] In the above embodiment of the present application, based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments, determining the reference coordinates of the multiple frames of target images includes: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame of the target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and a first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0172] In the above embodiment of the present application, randomly extracting a preset number of target images from a target image sequence includes: obtaining an initial frame number and an extraction interval; and extracting images from the target image sequence based on the initial frame number and the extraction interval to obtain a preset number of target images.

[0173] In the above embodiment of the present application, generating at least one set of training samples based on the target image of a preset number of frames and the reference coordinates of the target image of the preset number of frames includes: flipping, scaling or rotating the target image of the preset number of frames to obtain the target image after the operation; generating at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0174] In the above embodiment of the present application, after obtaining a target image sequence that meets preset conditions in an image database, the method also includes: adjusting the window width and window position of multiple frames of target images; adjusting the resolution of multiple frames of target images to the target resolution through a bilinear interpolation method; and adjusting the size of multiple frames of target images to the target size.

[0175] In the above embodiment of the present application, the initial model is trained using the training samples to obtain the positioning model, which includes: inputting the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; obtaining the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; determining the loss value of the initial model based on the coordinate difference, the reference coordinates and the predicted coordinates of the multiple frames of training images; and updating the parameters of the initial model based on the loss value to obtain the positioning model.

[0176] In the above embodiment of the present application, based on the coordinate difference, the reference coordinates and the predicted coordinates of the multi-frame training images, determining the loss value of the initial model includes: determining the constant error based on the difference between the coordinate difference and the second preset value; inputting the coordinate difference into the S function to determine the sorting error; inputting the coordinate difference into the distance regression loss function to determine the interval error; inputting the reference coordinates and the predicted coordinates of the multi-frame training images into the distance regression loss function to determine the regression error; obtaining the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0177] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0178] Example 5

[0179] According to an embodiment of the present application, a data processing device for implementing the above data processing method is also provided. Figure 12 As shown, the device 1200 includes: a receiving module 1202 , an acquisition module 1204 , a training module 1206 and an output module 1208 .

[0180] Among them, the receiving module 1202 is used to receive a model training request; the acquisition module 1204 is used to obtain multiple groups of training samples and an initial model requested for model training, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target object contained in the multiple frames of training images; the training module 1206 is used to train the initial model using multiple groups of training samples to obtain a positioning model; the output module 1208 is used to output the positioning model.

[0181] It should be noted that the receiving module 1202, the acquiring module 1204, the training module 1206, and the output module 1208 correspond to steps S402 to S408 in Example 1. The examples and application scenarios implemented by the four modules and the steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0182] In the above embodiment of the present application, the device further includes: a determination module, an extraction module and a generation module.

[0183] Among them, the acquisition module is also used to obtain a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; the determination module is used to determine the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation characteristics of the target object; the extraction module is used to randomly extract a preset number of target images in the target image sequence; the generation module is used to generate at least one set of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0184] In the above embodiment of the present application, the determination module includes: a labeling unit, a first determination unit, a second determination unit and a third determination unit.

[0185] Among them, the labeling unit is used to label multiple segments of the target object; the first determination unit is used to determine the frame numbers of the target image where the center points of the multiple segments are located, and obtain the frame numbers corresponding to the multiple segments; the second determination unit is used to determine the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; the third determination unit is used to determine the reference coordinates of the multiple frame target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0186] In the above embodiment of the present application, the second determining unit is further configured to obtain reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0187] In the above embodiment of the present application, the third determination unit is also used to perform the following steps: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and the first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0188] In the above embodiment of the present application, the extraction module includes: a first acquisition unit and an extraction unit.

[0189] Among them, the first acquisition unit is used to obtain the extraction initial frame number and the extraction interval; the extraction unit is used to extract the target image sequence based on the extraction initial frame number and the extraction interval to obtain the target image with a preset number of frames.

[0190] In the above embodiment of the present application, the generation module includes: an operation unit and a generation unit.

[0191] Among them, the operation unit is used to flip, scale or rotate the target image of a preset number of frames to obtain the target image after the operation; the generation unit is used to generate at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0192] In the above embodiment of the present application, the device further includes: an adjustment module, an interpolation module and a scaling module.

[0193] Among them, the adjustment module is used to adjust the window width and window position of the multi-frame target image; the interpolation module is used to adjust the resolution of the multi-frame target image to the target resolution through the bilinear interpolation method; and the scaling module is used to adjust the size of the multi-frame target image to the target size.

[0194] In the above embodiment of the present application, the training module includes: an input unit, a second acquisition unit, a fourth determination unit and an update unit.

[0195] Among them, the input unit is used to input the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; the second acquisition unit is used to obtain the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; the fourth determination unit is used to determine the loss value of the initial model based on the coordinate difference, the reference coordinates and predicted coordinates of multiple frames of training images; the update unit is used to update the parameters of the initial model based on the loss value to obtain a positioning model.

[0196] In the above embodiment of the present application, the fourth determination unit is also used to perform the following steps: determine the constant error based on the difference between the coordinate difference and the second preset value; input the coordinate difference into the S function to determine the sorting error; input the coordinate difference into the distance regression loss function to determine the interval error; input the reference coordinates and predicted coordinates of multiple frames of training images into the distance regression loss function to determine the regression error; obtain the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0197] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0198] Example 6

[0199] According to an embodiment of the present application, a data processing device for implementing the above data processing method is also provided. Figure 13 As shown, the device 1300 includes: an acquisition module 1302 and a training module 1304.

[0200] Among them, the acquisition module 1302 is used to obtain multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; the training module 1304 is used to use multiple groups of training samples to train the initial model to obtain a positioning model.

[0201] It should be noted that the acquisition module 1302 and the training module 1304 correspond to steps S902 to S904 in Example 2. The examples and application scenarios implemented by the two modules and the steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0202] In the above embodiment of the present application, the device further includes: a determination module, an extraction module and a generation module.

[0203] Among them, the acquisition module is also used to obtain a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; the determination module is used to determine the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation characteristics of the target object; the extraction module is used to randomly extract a preset number of target images in the target image sequence; the generation module is used to generate at least one set of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0204] In the above embodiment of the present application, the determination module includes: a labeling unit, a first determination unit, a second determination unit and a third determination unit.

[0205] Among them, the labeling unit is used to label each segment of the target object; the first determination unit is used to determine the frame numbers of the target images where the center points of the multiple segments are located, and obtain the frame numbers corresponding to the multiple segments; the second determination unit is used to determine the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; the third determination unit is used to determine the reference coordinates of the multiple frame target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0206] In the above embodiment of the present application, the second determining unit is further configured to obtain reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0207] In the above embodiment of the present application, the third determination unit is also used to perform the following steps: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and the first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0208] In the above embodiment of the present application, the extraction module includes: a first acquisition unit and an extraction unit.

[0209] Among them, the first acquisition unit is used to obtain the extraction initial frame number and the extraction interval; the extraction unit is used to extract the target image sequence based on the extraction initial frame number and the extraction interval to obtain the target image with a preset number of frames.

[0210] In the above embodiment of the present application, the generation module includes: an operation unit and a generation unit.

[0211] Among them, the operation unit is used to flip, scale or rotate the target image of a preset number of frames to obtain the target image after the operation; the generation unit is used to generate at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0212] In the above embodiment of the present application, the device further includes: an adjustment module, an interpolation module and a scaling module.

[0213] Among them, the adjustment module is used to adjust the window width and window position of the multi-frame target image; the interpolation module is used to adjust the resolution of the multi-frame target image to the target resolution through the bilinear interpolation method; and the scaling module is used to adjust the size of the multi-frame target image to the target size.

[0214] In the above embodiment of the present application, the training module includes: an input unit, a second acquisition unit, a fourth determination unit and an update unit.

[0215] Among them, the input unit is used to input the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; the second acquisition unit is used to obtain the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; the fourth determination unit is used to determine the loss value of the initial model based on the coordinate difference, the reference coordinates and predicted coordinates of multiple frames of training images; the update unit is used to update the parameters of the initial model based on the loss value to obtain a positioning model.

[0216] In the above embodiment of the present application, the fourth determination unit is also used to perform the following steps: determine the constant error based on the difference between the coordinate difference and the second preset value; input the coordinate difference into the S function to determine the sorting error; input the coordinate difference into the distance regression loss function to determine the interval error; input the reference coordinates and predicted coordinates of multiple frames of training images into the distance regression loss function to determine the regression error; obtain the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0217] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0218] Example 7

[0219] According to an embodiment of the present application, a data processing device for implementing the above data processing method is also provided. Figure 14 As shown, the device 1400 includes: a receiving module 1402 , a processing module 1404 and an output module 1406 .

[0220] Among them, the receiving module 1402 is used to receive the image sequence to be processed; the processing module 1404 is used to process the image sequence to be processed using the positioning model to obtain the positioning results of multiple frames of images to be processed in the image sequence to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmented features of the target object contained in at least one frame of the image to be processed; the output module 1406 is used to output the positioning results of multiple frames of images to be processed.

[0221] It should be noted that the receiving module 1402, processing module 1404, and output module 1406 correspond to steps S1002 to S1006 in Example 3. The examples and application scenarios implemented by the three modules and the steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0222] In the above embodiment of the present application, the device further includes: an acquisition module and a training module.

[0223] Among them, the acquisition module is used to obtain multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; the training module is used to use the multiple groups of training samples to train the initial model to obtain a positioning model.

[0224] In the above embodiment of the present application, the device further includes: a determination module, an extraction module and a generation module.

[0225] Among them, the acquisition module is also used to obtain a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; the determination module is used to determine the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation characteristics of the target object; the extraction module is used to randomly extract a preset number of target images in the target image sequence; the generation module is used to generate at least one set of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0226] In the above embodiment of the present application, the determination module includes: a labeling unit, a first determination unit, a second determination unit and a third determination unit.

[0227] Among them, the labeling unit is used to label multiple segments of the target object; the first determination unit is used to determine the frame numbers of the target image where the center points of the multiple segments are located, and obtain the frame numbers corresponding to the multiple segments; the second determination unit is used to determine the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; the third determination unit is used to determine the reference coordinates of the multiple frame target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0228] In the above embodiment of the present application, the second determining unit is further configured to obtain reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0229] In the above embodiment of the present application, the third determination unit is also used to perform the following steps: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and the first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0230] In the above embodiment of the present application, the extraction module includes: a first acquisition unit and an extraction unit.

[0231] Among them, the first acquisition unit is used to obtain the extraction initial frame number and the extraction interval; the extraction unit is used to extract the target image sequence based on the extraction initial frame number and the extraction interval to obtain the target image with a preset number of frames.

[0232] In the above embodiment of the present application, the generation module includes: an operation unit and a generation unit.

[0233] Among them, the operation unit is used to flip, scale or rotate the target image of a preset number of frames to obtain the target image after the operation; the generation unit is used to generate at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0234] In the above embodiment of the present application, the device further includes: an adjustment module, an interpolation module and a scaling module.

[0235] Among them, the adjustment module is used to adjust the window width and window position of the multi-frame target image; the interpolation module is used to adjust the resolution of the multi-frame target image to the target resolution through the bilinear interpolation method; and the scaling module is used to adjust the size of the multi-frame target image to the target size.

[0236] In the above embodiment of the present application, the training module includes: an input unit, a second acquisition unit, a fourth determination unit and an update unit.

[0237] Among them, the input unit is used to input the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; the second acquisition unit is used to obtain the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; the fourth determination unit is used to determine the loss value of the initial model based on the coordinate difference, the reference coordinates and predicted coordinates of multiple frames of training images; the update unit is used to update the parameters of the initial model based on the loss value to obtain a positioning model.

[0238] In the above embodiment of the present application, the fourth determination unit is also used to perform the following steps: determine the constant error based on the difference between the coordinate difference and the second preset value; input the coordinate difference into the S function to determine the sorting error; input the coordinate difference into the distance regression loss function to determine the interval error; input the reference coordinates and predicted coordinates of multiple frames of training images into the distance regression loss function to determine the regression error; obtain the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0239] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0240] Example 8

[0241] According to an embodiment of the present application, a data processing device for implementing the above data processing method is also provided. Figure 15 As shown, the device 1500 includes: an acquisition module 1502 and a processing module 1504.

[0242] Among them, the acquisition module 1502 is used to obtain the image sequence to be processed; the processing module 1504 is used to process the image sequence to be processed using the positioning model to obtain the positioning results of multiple frames of images to be processed in the image sequence to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmented features of the target object contained in at least one frame of the image to be processed.

[0243] It should be noted that the acquisition module 1502 and processing module 1504 correspond to steps S1102 to S1104 in Example 4. The examples and application scenarios implemented by the two modules and the steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0244] In the above embodiment of the present application, the device further includes: a training module.

[0245] Among them, the acquisition module is also used to obtain multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; the training module is used to use the multiple groups of training samples to train the initial model to obtain a positioning model.

[0246] In the above embodiment of the present application, the device further includes: a determination module, an extraction module and a generation module.

[0247] Among them, the acquisition module is also used to obtain a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; the determination module is used to determine the reference coordinates of multiple frames of target images in the target image sequence based on the segmentation characteristics of the target object; the extraction module is used to randomly extract a preset number of target images in the target image sequence; the generation module is used to generate at least one set of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0248] In the above embodiment of the present application, the determination module includes: a labeling unit, a first determination unit, a second determination unit and a third determination unit.

[0249] Among them, the labeling unit is used to label multiple segments of the target object; the first determination unit is used to determine the frame numbers of the target image where the center points of the multiple segments are located, and obtain the frame numbers corresponding to the multiple segments; the second determination unit is used to determine the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; the third determination unit is used to determine the reference coordinates of the multiple frame target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0250] In the above embodiment of the present application, the first determining unit is further configured to obtain reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0251] In the above embodiment of the present application, the third determination unit is also used to perform the following steps: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and the first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0252] In the above embodiment of the present application, the extraction module includes: a first acquisition unit and an extraction unit.

[0253] Among them, the first acquisition unit is used to obtain the extraction initial frame number and the extraction interval; the extraction unit is used to extract the target image sequence based on the extraction initial frame number and the extraction interval to obtain the target image with a preset number of frames.

[0254] In the above embodiment of the present application, the generation module includes: an operation unit and a generation unit.

[0255] Among them, the operation unit is used to flip, scale or rotate the target image of a preset number of frames to obtain the target image after the operation; the generation unit is used to generate at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0256] In the above embodiment of the present application, the device further includes: an adjustment module, an interpolation module and a scaling module.

[0257] Among them, the adjustment module is used to adjust the window width and window position of the multi-frame target image; the interpolation module is used to adjust the resolution of the multi-frame target image to the target resolution through the bilinear interpolation method; and the scaling module is used to adjust the size of the multi-frame target image to the target size.

[0258] In the above embodiment of the present application, the training module includes: an input unit, a second acquisition unit, a fourth determination unit and an update unit.

[0259] Among them, the input unit is used to input the training samples into the initial model to obtain the predicted coordinates of multiple frames of training images; the second acquisition unit is used to obtain the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; the fourth determination unit is used to determine the loss value of the initial model based on the coordinate difference, the reference coordinates and predicted coordinates of multiple frames of training images; the update unit is used to update the parameters of the initial model based on the loss value to obtain a positioning model.

[0260] In the above embodiment of the present application, the fourth determination unit is also used to perform the following steps: determine the constant error based on the difference between the coordinate difference and the second preset value; input the coordinate difference into the S function to determine the sorting error; input the coordinate difference into the distance regression loss function to determine the interval error; input the reference coordinates and predicted coordinates of multiple frames of training images into the distance regression loss function to determine the regression error; obtain the sum of the constant error, sorting error, interval error and regression error to obtain the loss value.

[0261] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0262] Example 9

[0263] According to an embodiment of the present application, a data processing system is further provided, including:

[0264] Processor. And

[0265] The memory is connected to the processor and is used to provide the processor with instructions for processing the following processing steps: receiving a model training request; obtaining multiple sets of training samples and an initial model in the model training request, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on segmented features of the target object contained in the multiple frames of training images; using the multiple sets of training samples to train the initial model to obtain a positioning model; and outputting the positioning model.

[0266] It should be noted that the preferred implementation scheme involved in the above embodiments of this application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0267] Example 10

[0268] The embodiment of the present application can provide a computer terminal, which can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal can also be replaced by a terminal device such as a mobile terminal.

[0269] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of a computer network.

[0270] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the data processing method: receiving a model training request; obtaining multiple sets of training samples and an initial model in the model training request, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; using the multiple sets of training samples to train the initial model to obtain a positioning model; and outputting the positioning model.

[0271] Optionally, Figure 16 This is a structural block diagram of a computer terminal according to an embodiment of the present application. Figure 16 As shown, the computer terminal A may include: one or more (only one is shown in the figure) processors 1602 and a memory 1604.

[0272] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules of the data processing method and device in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned data processing method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal A via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0273] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: receive a model training request; obtain multiple sets of training samples and an initial model in the model training request, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target object contained in the multiple frames of training images; use the multiple sets of training samples to train the initial model to obtain a positioning model; and output the positioning model.

[0274] Optionally, the processor may also execute the program code of the following steps: obtaining a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; determining the reference coordinates of multiple frames of target images in the target image sequence based on segmentation features of the target object; randomly extracting a preset number of target images from the target image sequence; and generating at least one set of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

[0275] Optionally, the processor may also execute the program code of the following steps: labeling multiple segments of the target object; determining the frame numbers of the target image where the center points of the multiple segments are located, and obtaining the frame numbers corresponding to the multiple segments; determining the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0276] Optionally, the processor may further execute program code of the following steps: obtaining reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0277] Optionally, the processor may also execute the program code of the following steps: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and a first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0278] Optionally, the processor may also execute the program code of the following steps: obtaining the extraction initial frame number and the extraction interval; performing image extraction on the target image sequence based on the extraction initial frame number and the extraction interval to obtain a target image with a preset number of frames.

[0279] Optionally, the processor may also execute the program code of the following steps: flipping, scaling or rotating a preset number of target frames to obtain an operated target image; and generating at least one set of training samples based on the operated target image and the reference coordinates of the operated target image.

[0280] Optionally, the processor may further execute program codes of the following steps: adjusting the window width and window position of the multi-frame target images; adjusting the resolution of the multi-frame target images to the target resolution by bilinear interpolation method; and adjusting the size of the multi-frame target images to the target size.

[0281] Optionally, the processor may also execute the program code of the following steps: input the training sample into the initial model to obtain the predicted coordinates of multiple frames of training images; obtain the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; determine the loss value of the initial model based on the coordinate difference, the reference coordinates and the predicted coordinates of multiple frames of training images; and update the parameters of the initial model based on the loss value to obtain a positioning model.

[0282] Optionally, the processor may also execute the program code of the following steps: determining a constant error based on the difference between the coordinate difference and a second preset value; inputting the coordinate difference into the S function to determine a sorting error; inputting the coordinate difference into the distance regression loss function to determine an interval error; inputting the reference coordinates and predicted coordinates of multiple frames of training images into the distance regression loss function to determine a regression error; obtaining the sum of the constant error, sorting error, interval error, and regression error to obtain a loss value.

[0283] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; use the multiple groups of training samples to train the initial model to obtain a positioning model.

[0284] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: receiving a sequence of images to be processed; using a positioning model to process the sequence of images to be processed to obtain positioning results of multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmented features of the target object contained in at least one frame of the image to be processed; and outputting the positioning results of the multiple frames of images to be processed.

[0285] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtaining a sequence of images to be processed; using a positioning model to process the sequence of images to be processed to obtain positioning results of multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmented features of the target object contained in at least one frame of the image to be processed.

[0286] By using an embodiment of the present application, a solution for locating human body parts is provided. Using a trained positioning model, a positioning result for each frame of an image can be obtained, and the purpose of rapid retrieval of massive medical images can be achieved based on the positioning result. Moreover, at least one set of training samples used in model training contains reference coordinates of multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target object, making the positioning model independent of the device and imaging method. In addition, the segmented features are referenced during the training process, realizing supervised training of the positioning model, so that the model can obtain consistent prediction results for the same input sequence after each training, achieving the technical effect of improving the stability of the positioning model processing results, and thus solving the technical problem that the data processing method in the related art cannot achieve rapid retrieval of massive medical images.

[0287] It can be understood by those skilled in the art that Figure 16 The structure shown is for illustration only, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Figure 16 It does not limit the structure of the above electronic device. For example, the computer terminal A may also include Figure 16 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 16 Different configurations shown.

[0288] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0289] Example 11

[0290] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the data processing method provided in the above embodiment.

[0291] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0292] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: receiving a model training request; obtaining multiple sets of training samples and an initial model of the model training request, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; training the initial model using the multiple sets of training samples to obtain a positioning model; and outputting the positioning model.

[0293] Optionally, the above-mentioned storage medium is also configured to store program code for executing the following steps: obtaining a target image sequence that meets preset conditions in an image database, wherein the target image sequence contains a target object; determining the reference coordinates of multiple frames of target images in the target image sequence based on segmentation features of the target object; randomly extracting a preset number of target images from the target image sequence; and generating at least one set of training samples based on the target images with a preset number of frames and the reference coordinates of the target images with a preset number of frames.

[0294] Optionally, the storage medium is further configured to store program codes for executing the following steps: labeling multiple segments of the target object; determining the frame numbers of the target image where the center points of the multiple segments are located, and obtaining the frame numbers corresponding to the multiple segments; determining the reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

[0295] Optionally, the storage medium is further configured to store program codes for executing the following steps: obtaining reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments.

[0296] Optionally, the storage medium is further configured to store program code for executing the following steps: obtaining the difference between the frame number of the first target image and the frame number corresponding to the first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; obtaining the difference between the frame number corresponding to the second segment and the frame number corresponding to the first segment to obtain a second difference; obtaining the ratio of the first difference to the second difference; obtaining the product of the ratio and a first preset value; obtaining the sum of the product and the reference coordinate corresponding to the first segment to obtain the reference coordinate of the first target image.

[0297] Optionally, the storage medium is further configured to store program codes for executing the following steps: obtaining the initial frame number and the extraction interval; performing image extraction on the target image sequence based on the initial frame number and the extraction interval to obtain a target image with a preset number of frames.

[0298] Optionally, the above-mentioned storage medium is also configured to store program code for executing the following steps: flipping, scaling or rotating the target image of a preset number of frames to obtain the target image after the operation; generating at least one set of training samples based on the target image after the operation and the reference coordinates of the target image after the operation.

[0299] Optionally, the above-mentioned storage medium is also configured to store program codes for executing the following steps: adjusting the window width and window position of multiple frames of target images; adjusting the resolution of multiple frames of target images to the target resolution through a bilinear interpolation method; and adjusting the size of multiple frames of target images to the target size.

[0300] Optionally, the storage medium is also configured to store program codes for executing the following steps: inputting the training sample into the initial model to obtain the predicted coordinates of multiple frames of training images; obtaining the difference between the predicted coordinates of two adjacent frames of training images to obtain the coordinate difference; determining the loss value of the initial model based on the coordinate difference, the reference coordinates and the predicted coordinates of multiple frames of training images; and updating the parameters of the initial model based on the loss value to obtain a positioning model.

[0301] Optionally, the storage medium is also configured to store program codes for executing the following steps: determining a constant error based on a difference between a coordinate difference and a second preset value; inputting the coordinate difference into an S function to determine a sorting error; inputting the coordinate difference into a distance regression loss function to determine an interval error; inputting the reference coordinates and predicted coordinates of multiple frames of training images into a distance regression loss function to determine a regression error; obtaining the sum of the constant error, sorting error, interval error, and regression error to obtain a loss value.

[0302] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: obtaining multiple groups of training samples, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, and the reference coordinates are determined based on the segmented features of the target objects contained in the multiple frames of training images; using the multiple groups of training samples to train the initial model to obtain a positioning model.

[0303] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: receiving a sequence of images to be processed; processing the sequence of images to be processed using a positioning model to obtain positioning results of multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning results of at least one frame of images to be processed are determined based on segmented features of a target object contained in at least one frame of images to be processed; and outputting the positioning results of multiple frames of images to be processed.

[0304] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: obtaining a sequence of images to be processed; processing the sequence of images to be processed using a positioning model to obtain positioning results of multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning results of at least one frame of images to be processed are determined based on segmented features of the target object contained in at least one frame of images to be processed.

[0305] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0306] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0307] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0308] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0309] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0310] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0311] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A data processing method, comprising: Receive model training requests; Obtaining multiple sets of training samples and an initial model requested for model training, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, wherein the reference coordinates are obtained by mapping the multiple frames of training images to preset coordinate axes based on segmented features of target objects contained in the multiple frames of training images, wherein the coordinate axes are constructed based on segmented features of the human spine; Training the initial model using the multiple sets of training samples to obtain a positioning model; The positioning model is output.

2. The method according to claim 1, wherein Before obtaining the multiple sets of training samples requested for model training, the method further includes: Acquire a target image sequence that meets a preset condition in an image database, wherein the target image sequence contains the target object; Determining reference coordinates of multiple frames of target images in the target image sequence based on segmentation features of the target object; Randomly extracting a preset number of target images from the target image sequence; The at least one set of training samples is generated based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

3. The method according to claim 2, wherein: Determining reference coordinates of multiple frames of target images in the target image sequence based on the segmented features of the target object includes: labeling multiple segments of the target object; Determine the frame numbers of the target image where the center points of the multiple segments are located, and obtain the frame numbers corresponding to the multiple segments; Determining reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments; The reference coordinates of the multiple frames of target images are determined based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

4. The method according to claim 3, wherein: Determining reference coordinates corresponding to the multiple segments based on the labeling results of the multiple segments includes: Based on the annotation results corresponding to the multiple segments and the preset coordinate intervals corresponding to two adjacent segments, the reference coordinates corresponding to the multiple segments are obtained.

5. The method according to claim 3, wherein: Determining the reference coordinates of the multiple frames of target images based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments includes: Obtaining a difference between a frame number of a first target image and a frame number corresponding to a first segment to obtain a first difference, wherein the first target image is any frame target image in the target image sequence, the frame number of the first target image is greater than the frame number corresponding to the first segment and less than the frame number corresponding to the second segment, and the first segment is adjacent to the second segment; Obtaining a difference between a frame sequence number corresponding to the second segment and a frame sequence number corresponding to the first segment to obtain a second difference; Obtaining a ratio of the first difference to the second difference; Obtaining a product of the ratio and a first preset value; The sum of the product and the reference coordinates corresponding to the first segment is obtained to obtain the reference coordinates of the first target image.

6. The method according to claim 2, wherein: Randomly extracting a preset number of target images from the target image sequence includes: Get the initial frame number and extraction interval; Image extraction is performed on the target image sequence based on the extracted initial frame number and the extraction interval to obtain the target image with the preset frame number.

7. The method according to claim 2, wherein: Generating the at least one set of training samples based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames includes: Performing a flipping, scaling, or rotating operation on the target image of the preset number of frames to obtain a target image after the operation; The at least one set of training samples is generated based on the target image after the operation and the reference coordinates of the target image after the operation.

8. The method according to claim 2, wherein: After obtaining a target image sequence that meets a preset condition in an image database, the method further includes: Adjusting the window width and window position of the multi-frame target images; Adjusting the resolution of the multi-frame target images to a target resolution by a bilinear interpolation method; The sizes of the multiple frames of target images are adjusted to a target size.

9. The method according to claim 1, wherein: Training the initial model using the at least one set of training samples to obtain a positioning model includes: Inputting the training samples into the initial model to obtain predicted coordinates of the multiple frames of training images; Obtain the difference in predicted coordinates between two adjacent frames of training images to obtain the coordinate difference; Determining a loss value of the initial model based on the coordinate difference, the reference coordinates of the multiple frames of training images, and the predicted coordinates; Parameters of the initial model are updated based on the loss value to obtain the positioning model.

10. The method according to claim 9, wherein: Determining the loss value of the initial model based on the coordinate difference, the reference coordinates of the multiple frames of training images, and the predicted coordinates includes: Determining a constant error based on a difference between the coordinate difference and a second preset value; Inputting the coordinate difference into an S function to determine a sorting error; Inputting the coordinate difference into a distance regression loss function to determine the interval error; Inputting the reference coordinates and predicted coordinates of the multiple frames of training images into a distance regression loss function to determine the regression error; The sum of the constant error, the sorting error, the interval error, and the regression error is obtained to obtain the loss value.

11. The method according to claim 1, wherein The initial model is trained based on the training samples using an adaptive moment estimation algorithm to obtain the positioning model, wherein an initial learning rate of the initial model is 0.001, and the number of iterations of the initial model is 100.

12. A data processing method, comprising: Acquire multiple sets of training samples, wherein at least one set of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, wherein the reference coordinates are obtained by mapping the multiple frames of training images to preset coordinate axes based on segmented features of target objects contained in the multiple frames of training images, wherein the coordinate axes are constructed based on segmented features of the human spine; The initial model is trained using the multiple sets of training samples to obtain a positioning model.

13. The method according to claim 12, wherein: Before obtaining multiple sets of training samples, the method further includes: Acquire a target image sequence that meets a preset condition in an image database, wherein the target image sequence contains the target object; Determining reference coordinates of multiple frames of target images in the target image sequence based on segmentation features of the target object; Randomly extracting a preset number of target images from the target image sequence; The at least one set of training samples is generated based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

14. The method according to claim 13, wherein Determining reference coordinates of multiple frames of target images in the target image sequence based on the segmented features of the target object includes: labeling multiple segments of the target object; Determine the frame numbers of the target image where the center points of the multiple segments are located, and obtain the frame numbers corresponding to the multiple segments; Determining reference coordinates corresponding to the multiple segments based on the annotation results corresponding to the multiple segments; The reference coordinates of the multiple frames of target images are determined based on the frame numbers of the multiple frames of target images, the frame numbers corresponding to the multiple segments, and the reference coordinates corresponding to the multiple segments.

15. The method according to claim 12, wherein: After the initial model is trained using the multiple sets of training samples to obtain a positioning model, the method further includes: Acquire a sequence of images to be processed, wherein the sequence of images to be processed contains the target object; The image sequence to be processed is processed using the positioning model to obtain positioning results of multiple frames of images to be processed in the image sequence to be processed.

16. A data processing method comprising: receiving an image sequence to be processed; Processing the sequence of images to be processed using a positioning model to obtain positioning results for multiple frames of images to be processed in the sequence of images to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on segmentation features of a target object contained in the at least one frame of the image to be processed, and the positioning model is obtained by training multiple sets of training samples, at least one set of training samples including: multiple frames of training images, and reference coordinates of the multiple frames of training images, the reference coordinates being obtained by mapping the multiple frames of training images to preset coordinate axes based on the segmentation features of the target objects contained in the multiple frames of training images, and the coordinate axes being constructed based on segmentation features of the human spine; Output the positioning results of the multiple frames of images to be processed.

17. The method according to claim 16, wherein The method further comprises: Obtain multiple sets of training samples; The initial model is trained using the multiple groups of training samples to obtain the positioning model.

18. The method according to claim 17, wherein Before obtaining multiple sets of training samples, the method further includes: Acquire a target image sequence that meets a preset condition in an image database, wherein the target image sequence contains the target object; Determining reference coordinates of multiple frames of target images in the target image sequence based on segmentation features of the target object; Randomly extracting a preset number of target images from the target image sequence; The at least one set of training samples is generated based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

19. A data processing method comprising: Obtain the image sequence to be processed; The image sequence to be processed is processed using a positioning model to obtain positioning results of multiple frames of images to be processed in the image sequence to be processed, wherein the positioning result of at least one frame of the image to be processed is determined based on the segmentation features of the target object contained in the at least one frame of the image to be processed, and the positioning model is obtained by training multiple groups of training samples, and at least one group of training samples includes: a preset number of frames of target images, and reference coordinates of the preset number of frames of target images, and the reference coordinates are obtained by mapping the multiple frames of training images to preset coordinate axes based on the segmentation features of the target objects contained in the multiple frames of training images, and the coordinate axes are constructed by segmentation features of the human spine.

20. The method according to claim 19, wherein The method further comprises: Obtain multiple sets of training samples; The initial model is trained using the multiple groups of training samples to obtain the positioning model.

21. The method according to claim 20, wherein Before obtaining multiple sets of training samples, the method further includes: Acquire a target image sequence that meets a preset condition in an image database, wherein the target image sequence contains the target object; Determining reference coordinates of multiple frames of target images in the target image sequence based on segmentation features of the target object; Randomly extracting a preset number of target images from the target image sequence; The at least one set of training samples is generated based on the target images of the preset number of frames and the reference coordinates of the target images of the preset number of frames.

22. A computer-readable storage medium comprising a stored program, wherein: When the program is running, the device where the computer-readable storage medium is located is controlled to execute the data processing method according to any one of claims 1 to 21.

23. A computer terminal comprising: A memory and a processor, wherein the processor is used to run a program stored in the memory, wherein the program executes the data processing method according to any one of claims 1 to 21 when running.

24. A data processing system comprising: processor; as well as A memory, connected to the processor, is configured to provide the processor with instructions for processing the following processing steps: receiving a model training request; Acquire multiple groups of training samples and initial models requested for model training, wherein at least one group of training samples includes: multiple frames of training images, and reference coordinates of the multiple frames of training images, wherein the reference coordinates are obtained by mapping the multiple frames of training images to preset coordinate axes based on segmented features of target objects contained in the multiple frames of training images, and the coordinate axes are constructed by segmented features of the human spine; train the initial model using the training samples to obtain a positioning model; and output the positioning model.

Images