Image labeling method and apparatus, computing device

Abstract

Embodiments of the present application provide an image labeling method and device, and a computing device. The method comprises: determining a first depth image and a second depth image; wherein the first depth image corresponds to first labeling information; extracting at least one first matching point pair of the first depth image and the second depth image in a first dimensional space; based on the at least one first matching point pair, extracting a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space; and determining second labeling information of the second depth image according to the at least one first matching point pair and the plurality of second matching point pairs, and in combination with the first labeling information of the first depth image. Embodiments of the present application improve the image labeling efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, and particularly relates to an image labeling method and device and a computing device. BACKGROUND

[0002] In the field of computer vision, the internal law and expression level of a sample image can be learned through a deep learning model, so that the model can have an analysis and learning ability like a human being, recognize the content in the image, and be applied to the fields of content recognition, target tracking or region segmentation.

[0003] A large number of training samples are needed in the deep learning process, and corresponding labeling information needs to be added for each training sample, such as the region of a target object, the key point of the target object, the category of the target object, the label or attribute and the like. However, the existing technical field mostly uses manual content labeling for training samples, and the labeling efficiency is low. SUMMARY

[0004] Therefore, the embodiments of the present application provide an image labeling method and device and a computing device to solve the technical problem of low labeling efficiency of manual image labeling in the prior art.

[0005] In a first aspect, the embodiments of the present application provide an image labeling method, comprising:

[0006] determining a first depth image and a second depth image; wherein the first depth image corresponds to first labeling information;

[0007] extracting at least one first matching point pair of the first depth image and the second depth image in a first dimensional space;

[0008] based on the at least one first matching point pair, extracting a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space;

[0009] determining second labeling information of the second depth image according to the at least one first matching point pair and the plurality of second matching point pairs, in combination with the first labeling information of the first depth image.

[0010] In a second aspect, the embodiments of the present application provide an image labeling method, comprising:

[0011] in response to a request of calling a target service, determining a processing resource corresponding to the target service;

[0012] using the processing resource corresponding to the target service to perform the following steps:

[0013] determine a first depth image and a second depth image; wherein the first depth image corresponds to first annotation information;

[0014] extract at least one first matching point pair of the first depth image and the second depth image in a first dimensional space;

[0015] extract a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space based on the at least one first matching point pair;

[0016] determine second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second matching point pairs, in combination with the first annotation information of the first depth image

[0017] In a third aspect, an embodiment of the present application provides an image annotation apparatus, comprising:

[0018] an image determination module configured to determine a first depth image and a second depth image; wherein the first depth image corresponds to first annotation information;

[0019] a first matching module configured to extract at least one first matching point pair of the first depth image and the second depth image in a first dimensional space;

[0020] a second matching module configured to extract a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space based on the at least one first matching point pair;

[0021] an image annotation module configured to determine second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second matching point pairs, in combination with the first annotation information of the first depth image.

[0022] In a fourth aspect, an embodiment of the present application provides a computing device, comprising: a storage component and a processing component; the storage component is configured to store one or more computer instructions; the one or more computer instructions are invoked by the processing component to execute the image annotation method provided by the embodiment of the present application.

[0023] The embodiment of the present application determines a first depth image and a second depth image, wherein the first depth image corresponds to first annotation information. Then, at least one first matching point pair of the first depth image and the second depth image in a first dimensional space can be extracted. And based on the at least one first matching point pair, multiple second matching point pairs of the first depth image and the second depth image in a second dimensional space can be extracted. By extracting the feature point pairs from both the first dimensional space and the second dimensional space, the matching information of the first depth image and the second depth image can be enriched, so that the second annotation information of the second depth image can be determined according to the at least one first matching point pair and the multiple second matching point pairs, in combination with the first annotation information of the first depth image. By extracting the rich feature point information, the matching information of the first depth image and the second depth image can be more detailed, so that the second annotation information of the second depth image can be accurately determined by using the first annotation information of the first depth image. The automatic and accurate image annotation is realized, and the annotation efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0025] Figure 1 A flowchart of one embodiment of an image annotation method provided by the embodiment of the present application;

[0026] Figure 2 A flowchart of another embodiment of an image annotation method provided by the embodiment of the present application;

[0027] Figure 3 A flowchart of another embodiment of an image annotation method provided by the embodiment of the present application;

[0028] Figure 4 A flowchart of another embodiment of an image annotation method provided by the embodiment of the present application;

[0029] Figure 5 An image annotation example diagram provided by the embodiment of the present application;

[0030] Figure 6 A flowchart of another embodiment of an image annotation method provided by the embodiment of the present application;

[0031] Figure 7 A structural schematic diagram of one embodiment of an image annotation device provided by the embodiment of the present application;

[0032] Figure 8 A structural schematic diagram of one embodiment of a computing device provided in the embodiments of the present application. DETAILED DESCRIPTION

[0033] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0034] The terms used in the embodiments of the present application are only for the purpose of describing particular embodiments and are not intended to limit the present application. The singular forms "a", "an" and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. "Multiple" generally includes at least two, but does not exclude the case of including at least one.

[0035] It should be understood that the term "and / or" used herein only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects.

[0036] Depending on the context, the word "if" as used herein can be interpreted as meaning "when" or "while" or "in response to determining" or "in response to identifying". Similarly, depending on the context, the phrase "if it is determined" or "if it is identified (a stated condition or event)" can be interpreted as meaning "when it is determined" or "in response to determining" or "when it is identified (a stated condition or event)" or "in response to identifying (a stated condition or event)".

[0037] It should also be noted that the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusion, so that a product or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such product or system. Without more limitations, the element defined by the sentence "including a" does not exclude the presence of other identical elements in the product or system including the element.

[0038] The technical scheme of the embodiment of the present application can be applied to the labeling scene of training samples, a depth image with labeling information is collected, and a matching point pair is extracted in two-dimensional spaces by matching the depth image with labeling information and an unlabeled depth image, so that the feature point pair is extracted, and the unlabeled image is automatically labeled by using the extracted feature point pair, and accurate and efficient labeling is realized.

[0039] In the prior art, before a mathematical processing model such as a machine learning model or a neural network model is used, the machine learning model needs to be trained by using training samples to obtain model parameters, and then the machine learning model with parameters is used. In the field of image processing, training images also need to be labeled to participate in training. However, in actual application, the labeling of training images is mostly performed in an artificial manner. For example, a labeling user can input an image to be labeled into a labeling program, and label the region of an object in the image and the class or attribute information of the object in the image in the labeling program. In this way, the labeling efficiency is low, and it is more difficult to label the images in a time-continuous video.

[0040] In the embodiment of the present application, a depth camera is first used to collect a video, so that a first depth image and a second depth image can be determined, wherein the first depth image corresponds to first labeling information. Then, at least one first matching point pair of the first depth image and the second depth image in a first-dimensional space can be extracted. Based on the at least one first matching point pair, a plurality of second matching point pairs of the first depth image and the second depth image in a second-dimensional space can be extracted. By extracting the feature point pairs from the first-dimensional space and the second-dimensional space, the matching information of the first depth image and the second depth image can be enriched, so that the second labeling information of the second depth image can be determined according to the at least one first matching point pair and the plurality of second matching point pairs in combination with the first labeling information of the first depth image. By extracting rich feature point information, the matching information of the first depth image and the second depth image can be more detailed, so that the second labeling information of the second depth image can be accurately determined by using the first labeling information of the first depth image. Automatic and accurate image labeling is realized, and the labeling efficiency of the image is improved.

[0041] The embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0042] As Figure 1 shown, a flowchart of one embodiment of an image labeling method provided by the embodiment of the present application is shown, which can include the following steps:

[0043] 101: Determine a first depth image and a second depth image.

[0044] The first depth image corresponds to first annotation information.

[0045] The image annotation method provided by the embodiments of the present application can be applied to a computing device or a server corresponding to the computing device. The computing device may, for example, include a computer, a notebook, an ultra-moblie personal computer (UMPC), a netbook, and the like, and the embodiments of the present application do not make excessive limitations on the specific type of the electronic device. The server corresponding to the computing device can communicate with the computing device based on a wired or wireless communication connection mode, and the server may, for example, include a general server or a cloud server, and the embodiments of the present application do not make excessive limitations on the specific type of the server.

[0046] The first annotation information of the first depth image can include object information of a first target object in the first depth image and a target position of the first target object.

[0047] The target position may, for example, include a region position of a region where the target object is located. The region position may, for example, include key point positions of an entire region where the object is located. For example, in the case of a rectangular region, the region position may include four coordinate points of the upper left, the lower left, the upper right, and the lower right of the rectangle. For example, in the case of an irregular region, the region position may include key coordinate points in the irregular region, such as angle conversion points, coordinate points at both ends of a circular arc, and vertex coordinate points.

[0048] The object information of the target object may, for example, include category information of the object, label information of the object, and / or attribute information of the object. The object information of the first target object may be obtained by pre-setting by a user.

[0049] 102: Extract at least one first matching point pair of the first depth image and the second depth image in a first dimensional space.

[0050] The first dimensional space may be a two-dimensional space corresponding to an image coordinate system or a pixel coordinate system. That is, the at least one first matching point pair of the first depth image in the first dimensional space may be at least one first matching point pair of the first depth image in the image coordinate system or the pixel coordinate system.

[0051] 103: Based on the at least one first matching point pair, extract a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space.

[0052] The number of first point pairs of the at least one first matching point pair is of an order smaller than the number of second point pairs of the plurality of second matching point pairs.

[0053] The second dimensional space is different from the first dimensional space. The second dimensional space can be a three-dimensional space corresponding to a camera coordinate system. The plurality of second matching point pairs can be matching point pairs belonging to the second dimensional space, so as to extract matching point pairs of different dimensional spaces.

[0054] 104: determining second annotation information of the second depth image based on the at least one first matching point pair and the plurality of second matching point pairs and in combination with the first annotation information of the first depth image.

[0055] The second annotation information is obtained based on the at least one first matching point pair and the plurality of second matching point pairs and in reference to the first annotation information.

[0056] The at least one first matching point pair of the first dimensional space and the plurality of second matching point pairs of the second dimensional space are used as matching information of the first depth image and the second depth image, so as to provide more detailed matching content and improve matching accuracy. When the at least one first matching point pair and the plurality of second matching point pairs are combined with the first annotation information, more accurate second annotation information of the second depth image can be obtained.

[0057] In some embodiments, after the second annotation information of the second depth image is determined, the step of determining the first depth image and the second depth image can be returned to continue to be executed, so as to realize continuous annotation of images.

[0058] In the embodiments of the present application, the first depth image and the second depth image can be determined, wherein the first depth image corresponds to the first annotation information. Then, the at least one first matching point pair of the first depth image and the second depth image in the first dimensional space can be extracted. Based on the at least one first matching point pair, the plurality of second matching point pairs of the first depth image and the second depth image in the second dimensional space can be extracted. By extracting feature point pairs from the first dimensional space and the second dimensional space, the matching information of the first depth image and the second depth image can be enriched, so that the second annotation information of the second depth image can be determined based on the at least one first matching point pair and the plurality of second matching point pairs and in combination with the first annotation information of the first depth image. By extracting rich feature point information, the matching information of the first depth image and the second depth image can be more detailed, so that the second annotation information of the second depth image can be accurately determined by using the first annotation information of the first depth image. Automatic and accurate image annotation is realized, and the annotation efficiency of images is improved.

[0059] As Figure 2As shown, a flowchart of another embodiment of an image labeling method provided by the present embodiment is shown, which can include the following steps:

[0060] 201: Determine a first depth image and a second depth image.

[0061] The first depth image corresponds to first labeling information.

[0062] 202: Determine whether to extract at least one first matching point pair of the first depth image and the second depth image in a first dimensional space, if yes, execute step 204; if no, execute step 203.

[0063] 203: Determine that the second depth image fails to be labeled.

[0064] If the second depth image matches the first depth image, the second depth image has the same or similar objects as the first depth image, and the first labeling information of the first depth image can be used as a generation basis of the second labeling information of the second depth image. On this basis, if the second depth image does not match the first depth image, it means that the objects in the second depth image are different from those in the first depth image, and the second depth image does not have a sharing basis for labeling information with the first depth image. At this time, the second depth image fails to be labeled. When the matching point pair is extracted for the first time, that is, the mapping condition of whether the depth image has labeling information is determined, and the extracted at least one matching point pair is used as the extraction basis of the subsequent second matching point pair, which can improve the extraction efficiency of the matching point pair.

[0065] In some embodiments, after determining that the second depth image fails to be labeled, the step of determining the first depth image and the second depth image can be returned to continue to be executed.

[0066] 204: Based on the at least one first matching point pair, extract a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space.

[0067] The order of magnitude of the first point pair number of the at least one first matching point pair is less than the second point pair number of the plurality of second matching point pairs.

[0068] 205: According to the at least one first matching point pair and the plurality of second matching point pairs, and in combination with the first labeling information of the first depth image, determine the second labeling information of the second depth image.

[0069] In some embodiments, after determining the second labeling information of the second depth image, the step of determining the first depth image and the second depth image can be returned to continue to be executed, so as to realize continuous labeling of images.

[0070] In some embodiments, if the first depth image and the second depth image fail to extract at least one first matching point pair in the first dimensional space, and the second depth image fails to be labeled, a prompt information of the labeling failure can be outputted.

[0071] After the at least one first matching point pair in the first dimensional space is extracted, a mapping relationship of the matching point pairs of the first depth image and the second depth image in the second dimensional space can be determined based on the at least one first matching point pair, so as to realize accurate acquisition of the second matching point pair.

[0072] As an embodiment, based on the at least one first matching point pair, extracting the plurality of second matching point pairs of the first depth image and the second depth image in the second dimensional space can include:

[0073] Mapping the at least one first matching point pair to the second dimensional space to obtain at least one first mapping point pair.

[0074] Based on the at least one first mapping point pair, determining a mapping relationship of the matching point pairs of the first depth image and the second depth image in the second dimensional space;

[0075] According to the mapping relationship of the matching point pairs, extracting the plurality of second matching point pairs of the first depth image and the second depth image in the second dimensional space.

[0076] Optionally, the mapping relationship of the matching point pairs can include a matching point pair mapping function. The input of the matching point pair mapping function is a coordinate point of the first depth image in the second dimensional space, and the output is a coordinate point in the second dimensional space which matches the input coordinate point of the first depth image.

[0077] As a possible implementation manner, according to the mapping relationship of the matching point pairs, extracting the plurality of second matching point pairs of the first depth image and the second depth image in the second dimensional space can include:

[0078] Determining a first depth sub-image of the first depth image and a second depth sub-image of the second depth image.

[0079] Based on the coordinate conversion formula, converting the first depth sub-image into a plurality of first point cloud data in the second dimensional space, and converting the second depth sub-image into a plurality of second point cloud data in the second dimensional space.

[0080] Based on the mapping relationship of the matching point pairs, determining the second point cloud data corresponding to the plurality of first point cloud data respectively from the plurality of second point cloud data;

[0081] Obtaining the plurality of second matching point pairs composed of the plurality of first point cloud data and the second point cloud data corresponding to the plurality of first point cloud data respectively.

[0082] In actual application, the first depth image and the second depth image can be RGB-D (Red, Green, Blue-Depth Map) images. The first depth image can include a first color sub-image and a first depth sub-image, and the second depth image can include a second color sub-image and a second depth sub-image. The first color sub-image and the second color sub-image belong to RGB images. The first depth sub-image and the second depth sub-image can belong to D (Depth) images.

[0083] The coordinate conversion formula can be pre-set for ready use. The coordinate conversion formula can be obtained by setting parameters such as camera intrinsic parameters. The coordinate conversion formula can convert a two-dimensional depth image into three-dimensional point cloud data, and the specific conversion manner is the same as that of existing depth images converted into point cloud data, which will not be described here. It should be noted that in the embodiments of the present application, after the first depth sub-image is converted into corresponding point cloud data, all data points in the obtained data cloud are referred to as a plurality of first cloud data points. After the second depth sub-image is converted into corresponding point cloud data, all data points in the obtained data cloud are referred to as a plurality of second point cloud data. That is, any first point cloud data represents a data point, and any second point cloud data also represents a data point.

[0084] Optionally, based on the mapping relationship of the matching points, for any second point cloud data, the matching degree of the second point cloud data and the plurality of first point cloud data can be calculated to determine the first point cloud data to which the second point cloud data is most matched, so as to obtain the first point cloud data corresponding to the plurality of second point cloud data respectively.

[0085] In a possible design, based on the mapping relationship of the matching points, determining the second point cloud data corresponding to the plurality of first point cloud data in the plurality of second point cloud data can include:

[0086] Based on the mapping relationship of the matching points, the plurality of first point cloud data are respectively mapped and calculated to obtain first mapping data corresponding to the plurality of first point cloud data respectively;

[0087] From the plurality of second point cloud data, second point cloud data respectively matched with the plurality of first mapping data are searched to obtain second point cloud data corresponding to the plurality of first mapping data respectively;

[0088] According to the second point cloud data corresponding to the plurality of first mapping data respectively, the plurality of first mapping data are replaced by corresponding first point cloud data to obtain second point cloud data corresponding to the plurality of first point cloud data respectively.

[0089] Optionally, the second point cloud data respectively corresponding to the plurality of first mapping data is obtained by searching the plurality of second point cloud data for the second point cloud data respectively matching the plurality of first mapping data.

[0090] In addition, the respective distance algorithm can be used to calculate and search the second point cloud data closest to the plurality of first mapping data to obtain the second point cloud data respectively corresponding to the plurality of first mapping data. The distance algorithm may, for example, include Hamming distance, Euclidean distance, Manhattan distance, Chebyshev distance, etc. The use of the distance algorithm in the embodiments of the present application is not limited too much.

[0091] As an embodiment, the at least one first matching point pair of the first depth image and the second depth image in the first dimension control includes:

[0092] The first color sub-image of the first depth image and the second color sub-image of the second depth image are determined.

[0093] The first color sub-image and the second color sub-image are input into a feature point matching algorithm to calculate the at least one first matching point pair in the first dimension space.

[0094] Optionally, the feature point matching algorithm can include a relatively common two-dimensional feature extraction algorithm such as ORB (Oriented Fast and Rotated Brief, feature point extraction and feature point description algorithm).

[0095] As another embodiment, the second annotation information of the second depth image is determined according to the at least one first matching point pair and the plurality of second matching point pairs in combination with the first annotation information of the first depth image.

[0096] The plurality of second matching point pairs are respectively mapped to the first dimension space to obtain a plurality of second mapping point pairs.

[0097] The second annotation information of the second depth image is determined according to the at least one first matching point pair and the plurality of second mapping point pairs in combination with the first annotation information of the first depth image.

[0098] In the determination of the second annotation information of the second depth image by using the plurality of second matching point pairs, the plurality of second matching point pairs can be mapped to the two-dimensional first dimension space to ensure that the dimension of the matching point is the same as the space dimension of the at least one first matching point, the reference significance is stronger, and the matching effect is improved.

[0099] As a possible implementation, the determining, according to the at least one first matched point pair and the plurality of second mapped point pairs and in combination with the first annotation information of the first depth image, of the second annotation information of the second depth image comprises:

[0100] extracting a plurality of first key points belonging to the first depth image and a plurality of second key points belonging to the second depth image from the at least one first matched point pair and the plurality of second mapped point pairs;

[0101] determining, based on the first annotation information corresponding to the plurality of first key points, the second annotation information corresponding to the plurality of second key points.

[0102] In some embodiments, the determining, based on the first annotation information corresponding to the plurality of first key points, of the second annotation information corresponding to the plurality of second key points can comprise:

[0103] determining, based on the plurality of first key points, a first target object of the first depth image and a target location where the first target object is located.

[0104] determining, according to the first annotation information, object information of the first target object.

[0105] determining, based on the plurality of second key points, a second target object of the second depth image and a target location where the second target object is located;

[0106] determining, according to the object information of the first target object, object information of the second target object.

[0107] determining the second annotation information comprising the object of the plurality of second target objects and the target location where the plurality of second target objects are located.

[0108] Optionally, the first target object can comprise a plurality of first target objects. The first annotation information comprises object information of the first target object. That is, the determining, according to the first annotation information, of the object information of the first target object can comprise obtaining the object information of the first target object in the first annotation information. In a possible design, the object information of the first target object can comprise class information of the object, label information of the object and / or attribute information of the object, etc.

[0109] In actual applications, the second target object can be identical to the second target object, or the second target object has a high similarity to the object of the first target object, which is higher than a preset similarity threshold. In order to ensure that the first target object is very similar to the second target object, the similarity threshold can be set to 90% or even 95% or above. At this time, the determining, according to the object information of the first target object, of the object information of the second target object can comprise: taking the object information of the first target object as the object information of the second target object. The object information of the second target object is matched with the object information of the first target object, so as to realize the automatic annotation of the object information.

[0110] Reference Figure 3 A flowchart of still another embodiment of an image labeling method provided by an embodiment of the present application can include the following steps.

[0111] 301: Obtain a depth video clip captured by a depth camera.

[0112] 302: Extract a first depth image and a second depth image in the depth video clip.

[0113] The first depth image corresponds to first labeling information.

[0114] 303: Extract at least one first matching point pair of the first depth image and the second depth image in a first dimensional space.

[0115] 304: Based on the at least one first matching point pair, extract a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space.

[0116] 305: According to the at least one first matching point pair and the plurality of second matching point pairs, and in combination with the first labeling information of the first depth image, determine second labeling information of the second depth image.

[0117] The depth camera can continuously capture depth video clips and can continuously send the captured depth video clips to a computing device configured with the image labeling method provided by an embodiment of the present application to realize automatic labeling of the video clips. In addition, the depth video clips can be first sent to a user terminal, and after the user terminal selects a first depth image in the depth video clip and sets first labeling information of the first depth image, the user terminal sends the depth video clip, the first depth image in the depth video clip, and the first labeling information of the first depth image to a server, and the server realizes automatic labeling of the images in the video clip.

[0118] Using a depth camera to obtain a depth image can involve depth information of the image in the feature collection process of the image labeling point, and the extraction of the image feature point is an important extension to further deepen the content of the image feature point, to realize accurate extraction of the feature. Compared with the process of using only common two-dimensional images for feature matching to label images, there is more feature point information, which can improve the accuracy and effectiveness of image labeling.

[0119] Some steps in the embodiments of the present application are the same as some steps in the foregoing embodiments, which will not be described here.

[0120] In the embodiments of the present application, the first depth image and the second depth image in the depth video segment collected by the depth camera can be extracted, so as to realize the extraction of at least one first matching point pair in the first dimension space and a plurality of second matching point pairs in the second dimension space of the first depth image and the second depth image, and to determine the second annotation information of the second depth image based on the extracted at least one first matching point pair and the plurality of second matching point pairs, in combination with the first annotation information of the first depth image, so as to realize the automatic annotation of the second depth image in the depth video segment and improve the annotation efficiency of the images in the video segment.

[0121] As a possible implementation manner, the first depth image and the second depth image in the depth video segment can be extracted in the following manner:

[0122] detecting the first depth image selected by the first annotation user from the depth video segment;

[0123] determining the first timestamp of the first depth image in the depth video segment;

[0124] extracting the second depth image from the depth video segment based on the first timestamp and the preset collection frequency.

[0125] Optionally, the collection sequence of the second depth image can be recorded, and the second timestamp of the second depth image can be determined based on the first timestamp and the preset collection frequency, and the second depth image corresponding to the second timestamp can be extracted from the depth video segment. The second timestamp is the product of the difference between the collection sequence and 1 multiplied by the collection frequency based on the first timestamp. That is, T2=T1+(R-1)*F. Wherein, T2 is the second timestamp, T1 is the first timestamp, R is the collection sequence of the second depth image, and F is the collection frequency.

[0126] In the embodiments of the present application, by selecting the first depth image and setting the collection frequency, the automatic collection of the second depth image can be realized, so as to realize the automatic annotation of the depth image in the depth video segment, to solve the problem of automatic annotation of the depth video segment, and to improve the annotation efficiency of the video segment.

[0127] In a possible design, the first annotation information of the first depth image can be obtained in the following manner: obtaining the first annotation information set by the first annotation user for the first depth image.

[0128] In another possible design, the first depth image and the second depth image can also be obtained by being directly uploaded by a user. The step of determining the first depth image and the second depth image in the foregoing embodiments can include:

[0129] acquire a first depth image and a second depth image transmitted by a second labeling user.

[0130] The first labeling information is set by the second labeling user for the first depth image.

[0131] The first depth image and the second depth image and the first labeling information of the first depth image can be transmitted by the user to a computing device configured with the image labeling method provided in the embodiments of the present application.

[0132] As Figure 4 shown, a flowchart of another embodiment of the image labeling method provided in the embodiments of the present application, the method can include:

[0133] 401: determine a first depth image and a second depth image.

[0134] The first depth image corresponds to first labeling information.

[0135] 402: determine a first color sub-image of the first depth image and a second color sub-image of the second depth image.

[0136] 403: input the first color sub-image and the second color sub-image into a feature point matching algorithm to calculate at least one first matching point pair in a first dimensional space.

[0137] 404: map the at least one first matching point pair to a second dimensional space to obtain at least one first mapping point pair.

[0138] 405: determine a matching point pair mapping relationship of the first depth image and the second depth image in the second dimensional space based on the at least one first mapping point pair.

[0139] 406: determine a first depth sub-image of the first depth image and a second depth sub-image of the second depth image.

[0140] 407: convert the first depth sub-image to a plurality of first point cloud data in the second dimensional space and convert the second depth sub-image to a plurality of second point cloud data in the second dimensional space based on a coordinate conversion formula.

[0141] 408: based on the matching point pair mapping relationship, map and calculate the plurality of first point cloud data respectively to obtain first mapping data respectively corresponding to the plurality of first point cloud data.

[0142] 409: find second point cloud data respectively matched with the plurality of first mapping data from the plurality of second point cloud data to obtain second point cloud data respectively corresponding to the plurality of first mapping data.

[0143] 410: replacing the plurality of first mapping data with corresponding first point cloud data according to the second point cloud data corresponding to the plurality of first mapping data respectively, to obtain a plurality of second matching point pairs constituted by the plurality of first point cloud data and the second point cloud data corresponding to the plurality of first point cloud data respectively.

[0144] 411: mapping the plurality of second matching point pairs to the first dimensional space respectively to obtain a plurality of second mapping point pairs.

[0145] 412: determining second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second mapping point pairs, in combination with the first annotation information of the first depth image.

[0146] Some steps in the embodiments of the present application are the same as those in the foregoing embodiments, and for the sake of brevity of description, will not be described here.

[0147] In the embodiments of the present application, the first depth image and the second depth image can be determined first. The first annotation information of the first depth image is known. Then, the first color sub-image of the first depth image and the second color sub-image of the second depth image can be determined, and the first color sub-image and the second color sub-image can be input into a feature point matching algorithm to calculate at least one first matching point pair in the first dimensional space. The at least one first matching point pair is relatively sparse, and mapping the at least one first matching point pair to the second dimensional space can obtain at least one first mapping point pair, so as to complete the mapping of the first matching point pair from the first dimensional space to the second dimensional space.

[0148] Since the at least one first mapping point pair implies the feature point matching relationship between the first depth image and the second depth image, the mapping relationship of the matching point pairs of the first depth image and the second depth image in the second dimensional space can be obtained. After the first depth sub-image of the first depth image and the second depth sub-image of the second depth image are determined, the first depth sub-image can be converted into a plurality of first point cloud data in the second dimensional space and the second depth sub-image can be converted into a plurality of second point cloud data in the second dimensional space based on a coordinate conversion formula. The number of the plurality of first point cloud data is much larger than the number of pixel points of the first color sub-image, and the number of the plurality of second point cloud data is much larger than the number of pixel points of the second color sub-image.

[0149] Based on the mapping relationship corresponding to the matched point pairs, the plurality of first point cloud data is respectively mapped and calculated to obtain a plurality of first mapping data respectively corresponding to the plurality of first point cloud data. From the plurality of second point cloud data, the second point cloud data respectively corresponding to the plurality of first mapping data is obtained by searching the second point cloud data respectively matched with the plurality of first mapping data. The matching of the point cloud data in the second dimension space is realized. Therefore, according to the second point cloud data respectively corresponding to the plurality of first mapping data, the plurality of first mapping data is replaced by the corresponding first point cloud data, so as to obtain a plurality of second matched point pairs composed of the plurality of first point cloud data and the second point cloud data respectively corresponding to the plurality of first point cloud data. Since the number of the plurality of first point cloud data and the plurality of second point cloud data is very large, the number of the plurality of second matched point pairs obtained is much larger than the number of the at least one first matched point pair, the expansion from the small amount of sparse at least one first matched point pair to the large amount of dense plurality of second matched point pairs is realized, and the first depth image and the second depth image obtained have more precise and detailed feature point pairs, so that the annotation of the feature points is more rigorous.

[0150] The plurality of second matched point pairs is mapped to the first dimension space to obtain a plurality of second mapping point pairs. The unified space expression of the matched point pairs is completed. According to the at least one first matched point pair and the plurality of second mapping point pairs, and in combination with the first annotation information of the first depth image, the second annotation information of the second depth image can be determined. Through the second annotation information of the second depth image, through more detection and matching of feature points, accurate and detailed matching of image feature points can be obtained, and more accurate and detailed matched feature point pairs are used to realize automatic annotation of the image and improve the annotation efficiency of the image.

[0151] In order to facilitate understanding, the application scenario of image automatic annotation performed on the computer M2 by the depth camera M1 collecting a depth video segment is taken as an example to introduce the technical solutions of the embodiments of the present application in detail.

[0152] Reference Figure 5 The depth camera M1 can be configured at a certain intersection and collect a depth video segment. The collection process of the depth video segment is continuous, and the collected depth video segment is continuously sent 501 to the computer M2.

[0153] Then, the computer M2 can obtain any depth video segment and extract 502 the first depth image and the second depth image in the video segment. In addition, the computer M2 can also detect the first depth image selected by the annotation user from the depth video segment and obtain the first annotation information of the first depth image provided by the annotation user. According to the first timestamp of the first depth image and the preset collection frequency, the second depth image is extracted.

[0154] Then, the computer M2 can extract 503 the at least one first matching point pair of the first depth image and the second depth image in the first dimensional space, map 504 the at least one first matching point pair to the second dimensional space, and determine the mapping relationship of the matching point pairs of the first depth image and the second depth image in the second dimensional space. Thus, the first depth sub-image of the first depth image and the second depth sub-image of the second depth image are converted 505 to the second dimensional space to obtain the plurality of first point cloud data and the plurality of second point cloud data. Then, based on the mapping relationship of the matching point pairs, the plurality of second matching point pairs respectively corresponding to the plurality of first point cloud data and the plurality of first point cloud data are determined 506. The plurality of second matching point pairs are respectively mapped 507 to the first dimensional space to obtain the plurality of second mapping point pairs. Thus, the at least one first matching point pair in the first dimensional space and the plurality of second mapping point pairs are combined with the first annotation information of the first depth image to determine 508 the second annotation information of the second depth image. The automatic annotation of the second depth image in the depth video segment is realized. In addition, in actual application, the second depth image can be continuously extracted to automatically annotate the plurality of second depth images of the depth video segment, so as to complete the overall annotation of the depth video segment and improve the annotation efficiency of the image.

[0155] In a possible design, the technical solutions provided in the embodiments of the present application can be configured in a server to form a service that can be provided externally. For example, the server can be a cloud server. Figure 6 An embodiment of a flowchart of an image annotation method provided in the embodiments of the present application can include the following steps.

[0156] 601: In response to a request of invoking a target service, determining a processing resource corresponding to the target service.

[0157] The processing resource corresponding to the target service is used to perform the following steps:

[0158] 602: Determining a first depth image and a second depth image; wherein the first depth image corresponds to first annotation information.

[0159] 603: Extracting at least one first matching point pair of the first depth image and the second depth image in a first dimensional space.

[0160] 604: Based on the at least one first matching point pair, extracting a plurality of second matching point pairs of the first depth image and the second depth image in a second dimensional space.

[0161] 605: According to the at least one first matching point pair and the plurality of second matching point pairs, combining the first annotation information of the first depth image to determine second annotation information of the second depth image.

[0162] In this embodiment of the application, the processing resources corresponding to the target service perform each step of image annotation. Figure 1 The steps in the illustrated embodiment are the same, and for the sake of brevity, they will not be repeated here.

[0163] In practical applications, the server that provides external services can be a cloud server. The processing resources of the target service provided by the cloud server are configured in the cloud server so that the cloud server can call and execute the corresponding image annotation service at any time.

[0164] In this embodiment, the server can provide an external target service to receive requests from users to call the target service, determine the processing resources corresponding to the target service, and use the processing resources to perform the steps of determining a first depth image and a second depth image. It then uses at least one first matching point pair extracted from the first and second depth images in a first-dimensional space. Based on this at least one first matching point pair, multiple second matching point pairs can be extracted from the first and second depth images in a second-dimensional space. By extracting feature point pairs from both the first and second-dimensional spaces, the matching information of the first and second depth images can be enriched. Therefore, based on at least one first matching point pair and multiple second matching point pairs, combined with the first annotation information of the first depth image, the second annotation information of the second depth image can be determined. Through the extraction of rich feature point information, the matching information of the first and second depth images becomes more detailed, allowing for accurate determination of the second annotation information of the second depth image using the first annotation information of the first depth image. This achieves automatic and accurate image annotation, improving the annotation efficiency of depth images.

[0165] like Figure 7 The diagram shown is a structural schematic of one embodiment of an image annotation device provided in this application. The device may include:

[0166] Image determination module 701: used to determine the first depth image and the second depth image.

[0167] The first depth image corresponds to the first annotation information.

[0168] First matching module 702: used to extract at least one first matching point pair of the first depth image and the second depth image in the first dimension space.

[0169] Second matching module 703: for extracting multiple second matching point pairs of a first depth image and a second depth image in a second-dimensional space based on at least one first matching point pair.

[0170] The image labeling module 704 is configured to determine second labeling information of the second depth image according to the at least one first matching point pair and the plurality of second matching point pairs, in combination with the first labeling information of the first depth image.

[0171] In the embodiments of the present application, the first depth image and the second depth image can be determined, wherein the first depth image corresponds to the first labeling information. Then, at least one first matching point pair of the first depth image and the second depth image in the first dimensional space can be extracted. Based on the at least one first matching point pair, a plurality of second matching point pairs of the first depth image and the second depth image in the second dimensional space can be extracted. By extracting the feature point pairs from the first dimensional space and the second dimensional space, the matching information of the first depth image and the second depth image can be enriched, so that the second labeling information of the second depth image can be determined according to the at least one first matching point pair and the plurality of second matching point pairs, in combination with the first labeling information of the first depth image. By extracting the rich feature point information, the matching information of the first depth image and the second depth image can be more detailed, so that the second labeling information of the second depth image can be accurately determined by using the first labeling information of the first depth image. The automatic and accurate image labeling is realized, and the labeling efficiency of the image is improved.

[0172] As an embodiment, the apparatus can further include:

[0173] The third matching module is configured to determine that the second depth image labeling fails if the at least one first matching point pair of the first depth image and the second depth image in the first dimensional space is not extracted.

[0174] As another embodiment, the second matching module can include:

[0175] The first mapping unit is configured to map the at least one first matching point pair to the second dimensional space to obtain at least one first mapping point pair.

[0176] The relationship determining unit is configured to determine a matching point pair mapping relationship of the first depth image and the second depth image in the second dimensional space based on the at least one first mapping point pair.

[0177] The first extraction unit is configured to extract a plurality of second matching point pairs of the first depth image and the second depth image in the second dimensional space according to the matching point pair mapping relationship.

[0178] In some embodiments, the first extraction unit can include:

[0179] The first determining unit is configured to determine a first depth sub-image of the first depth image and a second depth sub-image of the second depth image.

[0180] The point cloud conversion unit is configured to convert the first depth sub-image into a plurality of first point cloud data in a second dimensional space and convert the second depth sub-image into a plurality of second point cloud data in the second dimensional space based on a coordinate conversion formula.

[0181] The second mapping unit is configured to determine, based on a mapping relationship of the matching point pairs, second point cloud data corresponding to the first point cloud data in the plurality of first point cloud data respectively from the plurality of second point cloud data.

[0182] The first matching unit is configured to obtain a plurality of second matching point pairs from the plurality of first point cloud data and the second point cloud data corresponding to the first point cloud data respectively.

[0183] In some embodiments, the second mapping unit can be specifically configured to:

[0184] map the first point cloud data respectively based on the mapping relationship of the matching point pairs to obtain first mapping data corresponding to the first point cloud data respectively, search for second point cloud data matched with the first mapping data respectively from the plurality of second point cloud data to obtain second point cloud data corresponding to the first mapping data respectively, and replace the first mapping data with the corresponding first point cloud data according to the second point cloud data corresponding to the first mapping data respectively to obtain the second point cloud data corresponding to the first point cloud data respectively.

[0185] As a further example, the first matching module can include:

[0186] The second determining unit is configured to determine a first color sub-image of the first depth image and a second color sub-image of the second depth image.

[0187] The third matching unit is configured to input the first color sub-image and the second color sub-image into a feature point matching algorithm to obtain at least one first matching point pair in the first dimensional space.

[0188] In some embodiments, the image labeling module can include:

[0189] The third mapping unit is configured to map the plurality of second matching point pairs to the first dimensional space respectively to obtain a plurality of second mapping point pairs.

[0190] The image labeling unit is configured to determine second labeling information of the second depth image based on the at least one first matching point pair and the plurality of second mapping point pairs in combination with first labeling information of the first depth image.

[0191] As a possible implementation, the image labeling unit can include:

[0192] The key point extraction subunit is configured to extract a plurality of first key points belonging to the first depth image and a plurality of second key points belonging to the second depth image in the at least one first matching point pair and the plurality of second mapping point pairs.

[0193] The key point labeling subunit is configured to determine second labeling information corresponding to the plurality of second key points based on first labeling information corresponding to the plurality of first key points.

[0194] Further, the key point labeling subunit can be specifically configured to:

[0195] determine a first target object of the first depth image and a target position where the first target object is located based on the plurality of first key points, determine object information of the first target object according to the first labeling information, determine a second target object of the second depth image and a target position where the second target object is located based on the plurality of second key points, determine object information of the second target object according to the object information of the first target object, and determine the second labeling information composed of the object information of the plurality of second target objects and the target positions where the plurality of second target objects are located.

[0196] As a further embodiment, the image determination module can include:

[0197] The segment acquisition unit is configured to acquire a depth video segment captured by a depth camera.

[0198] The image acquisition unit is configured to extract a first depth image and a second depth image in the depth video segment.

[0199] The first depth image corresponds to first labeling information.

[0200] In some embodiments, the image acquisition unit can include:

[0201] The image detection subunit is configured to detect the first depth image selected by the first labeling user from the depth video segment.

[0202] The time determination subunit is configured to determine a first timestamp of the first depth image in the depth video segment.

[0203] The image determination subunit is configured to extract the second depth image from the depth video segment based on the first timestamp and a preset capture frequency.

[0204] In some embodiments, the first labeling information of the first depth image is obtained by:

[0205] The first labeling information set by the first labeling user for the first depth image is acquired.

[0206] As a further embodiment, the image determination module can include:

[0207] The image transmission unit is used to acquire a first depth image and a second depth image transmitted by the second annotation user. The first annotation information is obtained by the second annotation user setting information for the first depth image.

[0208] Figure 7 Image annotation devices can perform Figure 1 The implementation principle and technical effects of the image annotation method in the illustrated embodiment will not be repeated here. The specific implementation methods of the various modules, units, and sub-units executed by the processing components in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0209] In practical applications, Figure 7 The image annotation device shown can be configured as a computing device, see reference. Figure 8 This is a schematic diagram of the structure of one embodiment of a computing device provided in this application. The device may include: a storage component 801 and a processing component 802; the storage component 801 is used to store one or more computer instructions; one or more computer instructions are invoked by the processing component 802; the processing component 802 can be used to execute... Figures 1 to 4 Any of the image annotation methods shown.

[0210] The processing component 802 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the image annotation method provided in the above embodiments.

[0211] Storage component 801 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0212] exist Figure 8 When the computing device shown is a server, the processing component can be configured Figure 1The image labeling method of the embodiment shown corresponds to processing resources, and can be encapsulated as a service interface for providing a target service, so as to provide a corresponding image labeling service for users, so that the users can directly use the image labeling service, and the use efficiency of the target service is improved.

[0213] Of course, the computing device can also include other components, such as an input / output interface, a communication component, etc. The input / output interface provides an interface between the processing component and a peripheral interface module, which can be an output device, an input device, etc. The communication component is configured to facilitate wired or wireless communication between the computing device and other devices.

[0214] In addition, the embodiment of the present application further provides a computer readable storage medium, which can store one or more computer instructions, and the one or more computer instructions are executed to implement any one of the image labeling methods in the embodiments of the present application.

[0215] The device embodiments described above are only schematic, and units shown as separate components can or can not be physically separate, and components shown as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0216] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of a general hardware platform as necessary, and of course can also be realized by means of a combination of hardware and software. Based on such understanding, the above technical solutions can be embodied in the form of a computer product, and the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0217] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An image labeling method, characterized by, The method comprises: determining a first depth image and a second depth image; wherein the first depth image corresponds to first annotation information; extracting at least one first matching point pair of the first depth image and the second depth image in a first dimensional space; determining a matching point pair mapping relationship of the first depth image and the second depth image in a second dimensional space based on the at least one first matching point pair; determining a first depth sub-image of the first depth image and a second depth sub-image of the second depth image; converting the first depth sub-image into a plurality of first point cloud data in the second dimensional space and converting the second depth sub-image into a plurality of second point cloud data in the second dimensional space based on a coordinate conversion formula; determining second point cloud data corresponding to the plurality of first point cloud data respectively in the plurality of second point cloud data based on the matching point pair mapping relationship; obtaining a plurality of second matching point pairs formed by the plurality of first point cloud data and the second point cloud data corresponding to the plurality of first point cloud data respectively; determining second annotation information of the second depth image based on the at least one first matching point pair and the plurality of second matching point pairs and in combination with the first annotation information of the first depth image.

2. The method of claim 1, wherein, Further comprising: if no at least one first matching point pair of the first depth image and the second depth image in a first dimensional space is extracted, determining that the second depth image fails to be annotated.

3. The method of claim 1, wherein, The determination of the matching point pair mapping relationship of the first depth image and the second depth image in the second dimensional space based on the at least one first matching point pair comprises: mapping the at least one first matching point pair to the second dimensional space to obtain at least one first mapping point pair; determining the matching point pair mapping relationship of the first depth image and the second depth image in the second dimensional space based on the at least one first mapping point pair.

4. The method of claim 1, wherein, The determination of the second point cloud data corresponding to the plurality of first point cloud data respectively in the plurality of second point cloud data based on the matching point pair mapping relationship comprises: performing mapping calculation on the plurality of first point cloud data respectively based on the matching point pair mapping relationship to obtain first mapping data corresponding to the plurality of first point cloud data respectively; finding second point cloud data respectively matched with the plurality of first mapping data from the plurality of second point cloud data to obtain second point cloud data corresponding to the plurality of first mapping data respectively; replacing the plurality of first mapping data with corresponding first point cloud data according to the second point cloud data corresponding to the plurality of first mapping data respectively to obtain the second point cloud data corresponding to the plurality of first point cloud data respectively.

5. The method of claim 1, wherein, The extraction of the at least one first matching point pair of the first depth image and the second depth image in the first dimensional space comprises: determining a first color sub-image of the first depth image and a second color sub-image of the second depth image; inputting the first color sub-image and the second color sub-image into a feature point matching algorithm to calculate the at least one first matching point pair in the first dimensional space.

6. The method of claim 1, wherein, The determining the second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second matching point pairs, and in combination with the first annotation information of the first depth image comprises: mapping the plurality of second matching point pairs to the first dimensional space respectively to obtain a plurality of second mapping point pairs; determining the second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second mapping point pairs, and in combination with the first annotation information of the first depth image.

7. The method of claim 6, wherein, The determining the second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second mapping point pairs, and in combination with the first annotation information of the first depth image comprises: extracting a plurality of first key points belonging to the first depth image and a plurality of second key points belonging to the second depth image from the at least one first matching point pair and the plurality of second mapping point pairs; determining second annotation information corresponding to the plurality of second key points based on first annotation information corresponding to the plurality of first key points.

8. The method of claim 7, wherein, The determining the second annotation information of the second depth image according to the at least one first matching point pair and the plurality of second mapping point pairs, and in combination with the first annotation information of the first depth image comprises: determining a first target object of the first depth image and a target position where the first target object is located based on the plurality of first key points; determining object information of the first target object according to the first annotation information; determining a second target object of the second depth image and a target position where the second target object is located based on the plurality of second key points; determining object information of the second target object according to the object information of the first target object; determining second annotation information composed of the object information of the plurality of second target objects and the target positions where the plurality of second target objects are located.

9. The method of claim 1, wherein, The determining the first depth image and the second depth image; wherein the first depth image corresponds to first annotation information comprises: obtaining a depth video segment collected by a depth camera; extracting a first depth image and a second depth image in the depth video segment; wherein the first depth image corresponds to first annotation information.

10. The method of claim 9, wherein, The extracting the first depth image and the second depth image in the depth video segment comprises: detecting a first depth image selected by a first annotation user from the depth video segment; determining a first timestamp of the first depth image in the depth video segment; extracting the second depth image from the depth video segment based on the first timestamp and a preset collection frequency.

11. The method of claim 10, wherein, The first annotation information of the first depth image is obtained by: obtaining first annotation information set by the first annotation user for the first depth image.

12. The method of claim 1, wherein, The determining the first depth image and the second depth image; wherein the first depth image corresponds to first annotation information comprises: obtaining the first depth image and the second depth image transmitted by a second annotation user; wherein the first annotation information is obtained by the second annotation user setting for the first depth image.

13. An image labeling method characterized by comprising: comprises: In response to a request for invoking a target service, a processing resource corresponding to the target service is determined; The processing resource corresponding to the target service is used to perform the following steps: A first depth image and a second depth image are determined; the first depth image corresponds to first annotation information; At least one first matching point pair in a first dimension space of the first depth image and the second depth image is extracted; Based on the at least one first matching point pair, a matching point pair mapping relationship between the first depth image and the second depth image in a second dimension space is determined; A first depth sub-image of the first depth image and a second depth sub-image of the second depth image are determined; Based on a coordinate conversion formula, the first depth sub-image is converted into a plurality of first point cloud data in the second dimension space, and the second depth sub-image is converted into a plurality of second point cloud data in the second dimension space; Based on the matching point pair mapping relationship, second point cloud data corresponding to the plurality of first point cloud data in the plurality of second point cloud data is determined; A plurality of second matching point pairs formed by the plurality of first point cloud data and the second point cloud data corresponding to the plurality of first point cloud data are obtained; Based on the at least one first matching point pair and the plurality of second matching point pairs, the second annotation information of the second depth image is determined in combination with the first annotation information of the first depth image.

14. An image labeling apparatus characterized by comprising: Comprise: An image determination module is configured to determine a first depth image and a second depth image; the first depth image corresponds to first annotation information; A first matching module is configured to extract at least one first matching point pair in a first dimension space of the first depth image and the second depth image; A second matching module is configured to determine a matching point pair mapping relationship between the first depth image and the second depth image in a second dimension space based on the at least one first matching point pair; determine a first depth sub-image of the first depth image and a second depth sub-image of the second depth image; based on a coordinate conversion formula, convert the first depth sub-image into a plurality of first point cloud data in the second dimension space, and convert the second depth sub-image into a plurality of second point cloud data in the second dimension space; based on the matching point pair mapping relationship, determine second point cloud data corresponding to the plurality of first point cloud data in the plurality of second point cloud data; obtain a plurality of second matching point pairs formed by the plurality of first point cloud data and the second point cloud data corresponding to the plurality of first point cloud data; An image annotation module is configured to determine second annotation information of the second depth image in combination with the first annotation information of the first depth image based on the at least one first matching point pair and the plurality of second matching point pairs.

15. A computing device, comprising: Comprise: storage component and processing component; the storage component is used to store one or more computer instructions; the one or more computer instructions are called by the processing component to execute the image annotation method in any one of claims 1-12.

Images