CN114241141B - Smooth object three-dimensional reconstruction method and device, computer equipment and storage medium - Google Patents
Smooth object three-dimensional reconstruction method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN114241141B CN114241141B CN202210183551.0A CN202210183551A CN114241141B CN 114241141 B CN114241141 B CN 114241141B CN 202210183551 A CN202210183551 A CN 202210183551A CN 114241141 B CN114241141 B CN 114241141B
- Authority
- CN
- China
- Prior art keywords
- image
- matching
- smooth
- features
- dimensional model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a three-dimensional reconstruction method and device for a smooth object, computer equipment and a storage medium, which are applied to the technical field of three-dimensional reconstruction and used for improving the three-dimensional reconstruction effect of the smooth object. The method comprises the following steps: acquiring an object image acquired after surface material treatment is carried out on a smooth object; performing image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library; according to the depth image of the object image in the matching image library, performing three-dimensional reconstruction processing on the smooth object to obtain a three-dimensional model of the smooth object; and carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
Description
Technical Field
The present application relates to the field of three-dimensional reconstruction technologies, and in particular, to a smooth object three-dimensional reconstruction method, apparatus, computer device, and storage medium.
Background
With the gradual development of three-dimensional reconstruction technology of computer vision, people begin to use the three-dimensional reconstruction technology to record three-dimensional objective worlds digitally, and the three-dimensional reconstruction technology is also widely applied to scenes such as auxiliary driving, virtual reality, computer animation and the like.
The currently used three-dimensional reconstruction method of an object highly depends on the quality of the acquired two-dimensional image, and performs three-dimensional reconstruction of the object by identifying the object from the two-dimensional image. Because the surface of the smooth object can reflect the images of other objects in the environment, the reflected images can cause great interference to the identification of the smooth object, so that the smooth object cannot be normally identified, and the identification precision of the smooth object is lower and the three-dimensional reconstruction effect is poorer.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a smooth object three-dimensional reconstruction method, apparatus, computer device and computer readable storage medium capable of improving the smooth object three-dimensional reconstruction effect.
In a first aspect, the present application provides a smooth object three-dimensional reconstruction method. The method comprises the following steps:
acquiring an object image acquired after surface material treatment is carried out on a smooth object;
performing image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library;
according to the depth image of the object image in the matching image library, performing three-dimensional reconstruction processing on the smooth object to obtain a three-dimensional model of the smooth object;
and carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
In one embodiment, the image matching processing on the object image according to the target feature in the object image to obtain a matching image library includes:
clustering the target characteristics in the object image to obtain a clustering center;
and matching the object images according to the distance between the target features in the object images and the clustering center to obtain a matching image library.
In one embodiment, before performing image matching processing on the object image according to a target feature in the object image to obtain a matching image library, the method further includes:
according to the feature invariance, performing feature extraction processing on the object image to obtain candidate features;
screening the candidate features according to a preset stability metric to obtain screened features;
and performing characteristic filtering processing on the screened characteristics to obtain the target characteristics.
In one embodiment, after performing image matching processing on the object image according to a target feature in the object image to obtain a matching image library, the method further includes:
inquiring a preset number of object images from the matching image library to obtain an image set;
matching the feature descriptors of the images in the image set to obtain matched feature descriptors of the images in the image set;
according to the matched feature descriptors, removing features in the image set to obtain removed features;
and updating the target features in the image set by using the removed features.
In one embodiment, the three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object includes:
according to the octree, carrying out three-dimensional shape processing on the depth image to obtain a first three-dimensional model of the smooth object;
carrying out grid generation processing on the first three-dimensional model to obtain a second three-dimensional model;
and carrying out artifact processing on the second three-dimensional model to obtain the three-dimensional model.
In one embodiment, before performing three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object, the method further includes:
performing semi-global matching processing on the object image according to the camera parameter information of the object image in the matching image library to obtain the depth information of each pixel in the object image;
and generating a depth image of the object image according to the depth information.
In one embodiment, performing semi-global matching processing on an object image according to camera parameter information of the object image in the matching image library to obtain depth information of each pixel in the object image includes:
calculating depth candidate information between each pixel in the object image and a pixel in the parallax range according to the parallax range in the camera parameter information;
and generating the depth information according to the similarity between the depth candidate information.
In a second aspect, the application further provides a smooth object three-dimensional reconstruction device. The device comprises:
the image module is used for acquiring an object image acquired after surface material treatment is carried out on a smooth object;
the matching module is used for carrying out image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library;
the reconstruction module is used for performing three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object;
and the texture module is used for carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the following steps when executing the computer program:
acquiring an object image acquired after surface material treatment is carried out on a smooth object;
performing image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library;
according to the depth image of the object image in the matching image library, performing three-dimensional reconstruction processing on the smooth object to obtain a three-dimensional model of the smooth object;
and carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:
acquiring an object image acquired after surface material treatment is carried out on a smooth object;
performing image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library;
according to the depth image of the object image in the matching image library, performing three-dimensional reconstruction processing on the smooth object to obtain a three-dimensional model of the smooth object;
and carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
According to the smooth object three-dimensional reconstruction method, the smooth object three-dimensional reconstruction device, the computer equipment and the storage medium, the object image acquired after the surface material treatment is carried out on the smooth object is obtained, the image matching processing is carried out on the object image according to the target characteristics in the object image to obtain the matching image library, the smooth object is subjected to the three-dimensional reconstruction processing according to the depth image of the object image in the matching image library to obtain the three-dimensional model of the smooth object, the three-dimensional model is subjected to the texture processing, and then the target three-dimensional model of the smooth object is obtained. By adopting the method, the time required for retrieving the image in the three-dimensional reconstruction process of the smooth object is greatly shortened by matching the image library, so that the three-dimensional reconstruction efficiency of the smooth object is improved, and the influence of reflection information on the surface of the smooth object on the three-dimensional reconstruction effect is avoided by processing the surface material of the smooth object, so that the three-dimensional reconstruction effect of the smooth object is improved.
Drawings
FIG. 1 is a diagram of an embodiment of a method for three-dimensional reconstruction of a smooth object;
FIG. 2 is a schematic flow chart of a method for three-dimensional reconstruction of a smooth object according to an embodiment;
FIG. 3 is a schematic flow chart of a three-dimensional reconstruction method for a smooth object according to still another embodiment;
FIG. 4 is a schematic flow chart of a three-dimensional smooth object reconstruction method according to another embodiment;
FIG. 5 is a block diagram of an embodiment of an apparatus for three-dimensional reconstruction of smooth objects;
FIG. 6 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The smooth object three-dimensional reconstruction method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein the terminal 101 communicates with the server 102 via a network. The data storage system may store data that the server 102 needs to process. The data storage system may be integrated on the server 102, or may be located on the cloud or other network server. The method comprises the steps that a server 102 obtains an object image acquired after surface material processing is carried out on a smooth object, image matching processing is carried out on the object image according to target characteristics in the object image to obtain a matching image library, three-dimensional reconstruction processing is carried out on the smooth object according to depth images of the object image in the matching image library to obtain a three-dimensional model of the smooth object, texture processing is carried out on the three-dimensional model to further obtain a target three-dimensional model of the smooth object, the server 102 sends the target three-dimensional model to a terminal 101, and the terminal 101 displays the target three-dimensional model for a user to check. The terminal 101 may be but not limited to various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart car-mounted devices, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The server 102 may be implemented as a stand-alone server or as a server cluster comprised of multiple servers.
In one embodiment, as shown in fig. 2, a three-dimensional reconstruction method for a smooth object is provided, which is exemplified by the application of the method to the server in fig. 1, and includes the following steps:
step S201, acquiring an object image acquired after surface material processing is performed on a smooth object.
Wherein the surface material treatment comprises the step of treating the material of the surface of the smooth object by a physical method.
Specifically, the surface material of the smooth object is preprocessed to obtain a processed smooth object, the mirror reflection interference on the surface of the processed smooth object is reduced or disappears, an object image of the processed smooth object is collected, and the object image is sent to a server; wherein the pretreatment mode comprises the steps of carrying out pretreatment on the environment of the smooth object to realize surface material treatment and carrying out pretreatment on the smooth object to realize surface material treatment.
Practical application 1, the environment of the smooth object is pretreated to realize surface material treatment: the water mist or the smoke can be released to the storage environment of the smooth object, so that the mist with enough concentration is uniformly distributed in the storage environment of the smooth object to eliminate the mirror reflection on the surface of the smooth object, the object image of the smooth object is acquired and collected under the environment, and the object image is sent to the server.
Further, when the storage environment of the smooth object is a closed environment, the concentration of the fog in the environment is easier to control.
Practical application 2, the smooth object is pretreated to realize surface material treatment: the smooth object can be subjected to surface material treatment in an emergency cooling mode. For example, the smooth object is rapidly cooled to minus 20 ℃ or lower to obtain an emergency-cooled smooth object, and then the emergency-cooled smooth object is placed in a normal-temperature environment, wherein the normal-temperature environment can gradually desublimate water vapor on the surface of the emergency-cooled smooth object into a frost layer, and the frost layer can change the smooth state of the surface of the smooth object and eliminate the mirror reflection effect of the surface of the smooth object; when the fact that the surface of the smooth object is completely covered with the frost layer is confirmed, the smooth object is placed in a preset scene, an object image of the smooth object is collected, and the object image is sent to a server.
Practical application 3, the smooth object is pretreated to realize surface material treatment: the surface material treatment can be carried out on the smooth object by a slow cooling mode. For example, a smooth object is placed in a low-temperature environment, when the temperature of the smooth object is reduced to minus 20 ℃ or lower, a slow-cooling smooth object is obtained, the smooth object after slow cooling is placed in a normal-temperature environment, when it is confirmed that the surface of the smooth object is completely covered with a frost layer, the smooth object is placed in a preset scene, an object image of the smooth object is collected, and the object image is sent to a server.
Further, the cooling modes in practical application 2 and practical application 3 can be selected according to the influence of the temperature on the smooth object. If the influence of the rapid cooling mode on the smooth object is low, the surface material of the smooth object can be changed by adopting the rapid cooling mode so as to reduce the processing time of the smooth object; otherwise, the treatment can be carried out in a slow cooling mode so as to avoid damaging smooth objects.
Step S202, according to the target characteristics in the object image, image matching processing is carried out on the object image to obtain a matching image library.
The target features comprise features carrying feature invariance in the object image.
The matching image library is an image library for storing matched object images according to a preset mode. For example a library of matching images in the form of a lexical tree.
Specifically, according to the similarity (or distance) between the target features, the target features with higher similarity (or closer distance) are classified into one type in a clustering mode or a layering mode, and then the object images corresponding to the target features are matched together to obtain a matched image library.
And step S203, performing three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object.
Wherein the depth image is used to represent the geometry of the visible surface of a smooth object.
Specifically, according to the index value of the required depth image, a matching image library is inquired to obtain the required depth image, and incremental three-dimensional reconstruction is performed on the smooth object according to the depth image and the depth information corresponding to the depth image to obtain the three-dimensional Mesh model of the smooth object.
And step S204, carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
The texture of the three-dimensional model is the texture of a smooth object subjected to surface material treatment.
Specifically, on the basis of the three-dimensional model of the smooth object, texture reconstruction is performed on the surface texture of the three-dimensional model according to the original texture of the smooth object. And restoring the real texture of the smooth object on the three-dimensional model through the server, thereby obtaining a target three-dimensional model of the smooth object.
According to the smooth object three-dimensional reconstruction method, an object image acquired after surface material processing is carried out on a smooth object is acquired, image matching processing is carried out on the object image according to target characteristics in the object image to obtain a matching image library, three-dimensional reconstruction processing is carried out on the smooth object according to a depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object, texture processing is carried out on the three-dimensional model, and then the target three-dimensional model of the smooth object is obtained. By adopting the method, the time required for retrieving the image in the three-dimensional reconstruction process of the smooth object is greatly shortened by matching the image library, so that the three-dimensional reconstruction efficiency of the smooth object is improved, and the influence of reflection information on the surface of the smooth object on the three-dimensional reconstruction effect is avoided by processing the surface material of the smooth object, so that the three-dimensional reconstruction effect of the smooth object is improved.
In an embodiment, in step S202, according to the target feature in the object image, performing image matching processing on the object image to obtain a matching image library, which specifically includes the following contents:
clustering the target characteristics in the object image to obtain a clustering center; and matching the object images according to the distance between the target features in the object images and the clustering center to obtain a matching image library.
Specifically, all target features in the object image are obtained, clustering processing is performed on all the target features to obtain a clustering center, the distance between the target features in the object image and the clustering center is inquired, the clustering center with the minimum distance to the target features in the object image is obtained, the object image is matched to the category corresponding to the clustering center, the clustering process is repeated on the object images in each category respectively until the depth of the matched image library reaches the preset maximum depth.
For example, K-means clustering is performed on all target features, the object images are divided into K groups according to the distance between a clustering center and the object images, then the clustering process is repeated on the K groups of object images respectively until the depth of a matched image library is L, a vocabulary tree with the depth of L is constructed, the clustering center of the vocabulary tree is a feature descriptor, and common features and difference features between the object images are obtained by comparing the feature descriptors.
In the embodiment, the time required for retrieving the images in the three-dimensional reconstruction process of the smooth object is greatly shortened by matching the image library, so that the three-dimensional reconstruction efficiency of the smooth object is improved.
In one embodiment, before performing image matching processing on an object image according to a target feature in the object image to obtain a matching image library, the method further includes:
according to the feature invariance, performing feature extraction processing on the object image to obtain candidate features; screening the candidate features according to a preset stability metric to obtain screened features; and performing characteristic filtering processing on the screened characteristics to obtain target characteristics.
Wherein feature invariance means remaining invariant to both scale and rotation.
Where candidate features refer to a potentially unique group of pixels having characteristics that remain invariant to both scale and rotation.
Specifically, the positions of images with all scales in an object image are inquired through a Gaussian differential function to obtain candidate features with high probability and feature invariance, each candidate feature is fitted through a pre-constructed fitting model to obtain the position and scale corresponding to the candidate feature, the features with the stability degree higher than a preset stability measure are screened out according to the stability degree of the candidate features to serve as screened features, one or more directions are given to the position of each screened feature based on the local gradient direction of the image, the direction, scale and position of the screened features are changed, and the screened features which cannot be changed after the direction, scale and position are confirmed from the screened features are used as the invariance features; and carrying out grid filtering processing on the invariance characteristics to obtain target characteristics.
In this embodiment, the invariance features are filtered to obtain target features, so that the quantity difference of the target features between the object images is maintained at a corresponding quantity, and the efficiency and accuracy of constructing a matching image library by the object images are improved.
In one embodiment, after performing image matching processing on the object image according to the target feature in the object image to obtain a matching image library, the method further includes:
inquiring a preset number of object images from a matching image library to obtain an image set; matching the feature descriptors of the images in the image set to obtain the matched feature descriptors of the images in the image set; according to the matched feature descriptors, removing features in the image set to obtain removed features; and updating the target feature in the image set by using the removed feature.
Wherein the preset number is a multiple of 2, such as 2, 4, 6, 8, etc.
The image set includes at least one set of object images in the form of image pairs, that is, the image set includes at least two object images.
Wherein at least one same feature is included between two object images in the pair.
Specifically, acquiring a preset number of object images from a matching image library to obtain an image set; the obtaining mode may be a query mode, a random extraction mode, or an image obtaining instruction mode, where the query mode may be a name query, an image identifier query, or a feature query, and is not limited specifically herein. Performing luminosity matching processing on the feature descriptors of all object images in the image set to obtain a matching list matched with the feature descriptors of other object images; if any group of image pairs has only one effectively matched feature descriptor, inquiring at least two image descriptors which are most similar to the feature descriptor in the other image in the image pair according to the feature descriptor of one image in the image pair, setting a relative threshold value based on the Euclidean space distance of the at least two image descriptors, deleting repeated features in an image area containing repeated structures in the group of image pairs, and further obtaining the preliminarily removed features; inputting the preliminarily removed features into an outlier detection model according to the positions of the preliminarily removed features in the object image, performing filtering processing on the preliminarily removed features by using epipolar geometry in the outlier detection model to remove error features in the preliminarily removed features, iterating according to the number of the features after filtering processing to obtain the removed features, and updating the target features of the corresponding object image in the image set by using the removed features.
In practical application, any two object images (marked as an image A and an image B) are obtained from a matching image library in a vocabulary tree form to obtain an image set, luminosity matching is performed on the two object images in the image set to obtain a list of candidate features of each feature in the image A in the image B, wherein the luminosity matching comprises matching according to luminosity information of the object images; the method comprises the steps of assuming that only one feature in an image A or an image B can be effectively matched to judge whether the matching is effective in a nonlinear space; for each feature descriptor in the image A, two closest image descriptors are searched in the image B, and a relative threshold value is set based on the distance between the two image descriptors in Euclidean space, so that repeated features in a region containing repeated structures in the image A and the image B are eliminated, and the eliminated image A and/or the eliminated image B are obtained; and then according to the feature position in the image A and/or the image B after elimination, filtering the features in the image A and/or the image B after elimination by using epipolar geometry in a Random Sample Consensus (RANSAC) outlier detection model, iterating according to the number of the features after filtering to obtain the features after elimination, and updating the target features of the image A and/or the image B in the image set by using the features after elimination.
Further, after the step of updating the target feature in the image set using the culled feature, the method further includes:
carrying out feature tracking processing on the matched feature descriptors in the image set to obtain matched features including feature tracks in the matched feature descriptors and discrete matched features not including feature tracks in the matched feature descriptors;
deleting the discrete matching features to obtain deleted matching features;
screening out an image set with the maximum number of deleted matching features as a target image set according to the number of deleted matching features of each image set;
and acquiring camera parameter information according to the two-dimensional characteristics and the three-dimensional characteristics of the image pair in the target image set, and updating the image information file by using the camera parameter information.
Wherein the camera parameter information includes a position parameter of the camera (or the video camera).
The image information file is used for adding constraints to the position parameters of the camera in the three-dimensional reconstruction process of the smooth object.
Specifically, all matching feature descriptors in an image set are fused into tracks, each track represents a spatial point in reality, the point is visible in object images at multiple angles, and in the fusion process, if part of matching feature descriptors does not belong to any available track, the part of matching feature descriptors is deleted; acquiring a target image set; the target image set is the image set containing the maximum number of matched feature descriptors; if the same number of image sets exist, screening out an image set with the largest shooting angle difference from the shooting angles of the same number of image sets to serve as a target image set so as to improve the reliability of geometric information provided by the target object image; calculating to obtain a basic matrix between image pairs in the target image set, taking one object image in the image pair as an origin of a coordinate system, and triangularly converting two-dimensional features in the image pair in the target image set into three-dimensional features; according to the association between the two-dimensional features and the three-dimensional features, carrying out nonlinear minimization on the object image which is most associated with the three-dimensional features to obtain camera parameters; and repeating the steps until the new camera angle cannot be positioned.
In the embodiment, by performing feature matching and feature rejection on the target feature points of the object image in the matching image library, the target feature points which are mistakenly matched in the object image can be effectively reduced, and new features are appropriately added according to the number of the remaining rejected features to ensure the accuracy of the target features in the object image, so that the three-dimensional reconstruction accuracy of the smooth object is improved, and further the three-dimensional reconstruction effect of the smooth object is improved.
In an embodiment, in step S203, the three-dimensional reconstruction processing is performed on the smooth object according to the depth image of the object image in the matching image library, so as to obtain a three-dimensional model of the smooth object, which specifically includes the following contents:
according to the octree, carrying out three-dimensional shape processing on the depth image to obtain a first three-dimensional model of the smooth object; carrying out grid generation processing on the first three-dimensional model to obtain a second three-dimensional model; and carrying out artifact processing on the second three-dimensional model to obtain a three-dimensional model.
Specifically, depth images of object images in a matching image library are integrated into an octree, depth information corresponding to the depth images is stored to corresponding nodes in the octree to obtain a global octree, and three-dimensional body reconstruction is performed on the smooth object according to the global octree to obtain a first three-dimensional model of the smooth object; and (3) carrying out mesh generation processing on the first three-dimensional model according to a 3D-Delaunay tetrahedron algorithm (three-dimensional Delaunay algorithm) to obtain a triangular patch on the surface of the first three-dimensional model, and using the triangular patch as a second three-dimensional model to optimize the consistency of the boundary edge and the boundary surface of the specified region of the smooth object.
And according to the maximum flow graph cutting algorithm, carrying out optimization processing on the whole volume of the second three-dimensional model, filtering preset nodes on the surface to obtain the optimized second three-dimensional model, and updating the second three-dimensional model by using the optimized second three-dimensional model. For example, there are multiple paths between any two points of the surface of the second three-dimensional model; and calculating the flow of all paths between any two points on the surface of the second three-dimensional model according to a maximum flow graph cutting algorithm, screening an edge with zero contribution degree in the second three-dimensional model according to the flow of all paths between any two points, and optimizing the edge with zero contribution degree to obtain the optimized second three-dimensional model.
And carrying out artifact filtering on the second three-dimensional model to obtain a three-dimensional model. In practical application, according to the laplacian filtering, grid smoothing is carried out on the second three-dimensional model, and the three-dimensional model is obtained.
In this embodiment, three-dimensional reconstruction is performed according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object in a state after surface treatment, and because the surface material of the smooth object is treated in advance, the influence of reflection information on the surface of the smooth object on the three-dimensional reconstruction effect is avoided, so that the three-dimensional reconstruction effect of the smooth object is improved.
In one embodiment, before performing three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object, the method further includes:
performing semi-global matching processing on the object image according to the camera parameter information of the object image in the matching image library to obtain the depth information of each pixel in the object image; and generating a depth image of the object image according to the depth information.
Specifically, as shown in fig. 3, after an object image acquired after surface texture processing is performed on a smooth object is acquired, camera information initialization is further included, that is, metadata of the image is extracted from the object image, and according to Exchangeable image file format (EXIF) data of the image, a focal length and sensor information of the image are extracted, so as to generate an image information file. According to the image information file, a camera parameter file corresponding to the object image under the visual angle is generated, according to visual angle information in the camera parameter file, semi-global matching processing is carried out on the corresponding object image, a three-dimensional matrix of the object image is obtained, depth information of each pixel in the object image is stored in the three-dimensional matrix, and a depth image of the object image is generated according to the depth information.
In practical applications, the image information file may be a camera information. The picture storage location, the width and height of the picture, whether the picture is compressed, the view angle of the picture, whether the picture uses white balance, the luminosity of the picture, and the like. The camera parameter file may be a viewpoint.
In this embodiment, three-dimensional reconstruction is performed according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object in a state after surface treatment, and since the surface material of the smooth object is treated in advance, influence of reflection information on the surface of the smooth object on the three-dimensional reconstruction effect is avoided, so that the three-dimensional reconstruction effect of the smooth object is improved.
In one embodiment, the obtaining depth information of each pixel in the object image by performing semi-global matching processing on the object image according to the camera parameter information of the object image in the matching image library includes:
calculating depth candidate information between each pixel in the object image and a pixel in the parallax range according to the parallax range in the camera parameter information; and generating depth information according to the similarity between the depth candidate information.
Specifically, a camera parameter file corresponding to the object image is acquired, and the parallax range (marked as "parallax range") of each pixel in the object image is specified according to the camera parameter information in the camera parameter fileD) According to the size of the matching image (label asW，H) Generating a size ofW×H×DMatrix of (labeled as)C) Calculating the matching cost value of each pixel in the object image and each image in the parallax range, namely each pixel comprises one or more depth candidate information, calculating the similarity between the depth candidate information according to a zero-mean normalization mutual tracking algorithm, and accumulating the similarity to a matrixCIn the method, a local minimum value is selected as depth information, and the object image is updated using the depth informationIndex values in the matching image library.
In this embodiment, a corresponding depth image is generated according to the depth information, and three-dimensional reconstruction is performed according to the depth image to obtain a three-dimensional model of the smooth object in a state after surface treatment.
In one embodiment, as shown in fig. 4, another smooth object three-dimensional reconstruction method is provided, which is exemplified by the application of the method to the server in fig. 1, and includes the following steps:
step S401, acquiring an object image acquired after surface material processing is performed on a smooth object.
Step S402, according to the characteristic invariance, carrying out characteristic extraction processing on the object image to obtain candidate characteristics; screening the candidate features according to a preset stability metric to obtain screened features; and carrying out feature filtering processing on the screened features to obtain target features.
Step S403, clustering the target characteristics in the object image to obtain a clustering center; and matching the object images according to the distance between the target features in the object images and the clustering center to obtain a matching image library.
Step S404, inquiring a preset number of object images from a matching image library to obtain an image set; and matching the feature descriptors of the images in the image set to obtain the matched feature descriptors of the images in the image set.
Step S405, removing the features in the image set according to the matched feature descriptors to obtain the removed features; and updating the target feature in the image set by using the removed feature.
Step S406, calculating, according to the parallax range in the camera parameter information, a parallax range between each pixel in the object image and the pixel in the parallax range, and calculating depth candidate information between each pixel in the object image and the pixel in the parallax range.
Step S407, generating depth information according to the similarity between the depth candidate information; and generating a depth image of the object image according to the depth information.
Step S408, according to the octree, three-dimensional shape processing is carried out on the depth image to obtain a first three-dimensional model of the smooth object; carrying out grid generation processing on the first three-dimensional model to obtain a second three-dimensional model; and carrying out artifact processing on the second three-dimensional model to obtain a three-dimensional model.
And step S409, performing texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
The smooth object three-dimensional reconstruction method can provide the following beneficial effects:
(1) by processing the surface material of the smooth object, the influence of the reflection information on the surface of the smooth object on the three-dimensional reconstruction effect is avoided, and the three-dimensional reconstruction effect of the smooth object is improved.
(2) By matching the image library, the time required for retrieving the images in the three-dimensional reconstruction process of the smooth object is greatly shortened, and the three-dimensional reconstruction efficiency of the smooth object is improved.
(3) The target features are obtained after filtering processing, so that the quantity difference of the target features among the object images is maintained at a corresponding quantity, and the efficiency and the accuracy of constructing a matching image library by the object images are improved.
(4) By reducing the target characteristic points which are wrongly matched in the object image, the accuracy of the target characteristics in the object image can be ensured, so that the three-dimensional reconstruction accuracy of the smooth object is improved, and the three-dimensional reconstruction effect of the smooth object is further improved.
In order to more clearly illustrate the smooth object three-dimensional reconstruction method provided by the embodiments of the present disclosure, the following describes the smooth object three-dimensional reconstruction method in a specific embodiment. In one embodiment, as shown in fig. 3, the present disclosure further provides a smooth object three-dimensional reconstruction method, which specifically includes the following steps:
(1) pre-treating the object or environment: the smooth object or the environment of the smooth object is preprocessed to eliminate the mirror reflection on the surface of the smooth object, so that the preprocessed smooth object is obtained.
(2) Acquiring an object image: and taking the preprocessed smooth object as a center, and shooting the preprocessed smooth object according to various preset shooting heights and various preset shooting angles to obtain a clear image of the preprocessed smooth object as an object image.
(3) Initializing camera information: processing metadata of the object image, and acquiring the focal length and sensor information of the object image from EXIF data of the object image, wherein the acquired image information of the object image adds constraints for position parameters of a camera in the subsequent reconstruction process; generating a camera Init. Sfm files are generated for representing perspective information of the image of the object.
(4) Feature extraction: acquiring any two object images, searching image positions on all scales, and identifying interest points with high probability and feature invariance, namely candidate features, through Gaussian differential functions; determining the position and scale of the candidate feature by fitting a refined model; determining the screened features according to the stability degree of the candidate features; based on the local gradient direction of the image, distributing the local gradient direction to one or more directions of each screened feature, and determining the screened features which cannot be changed after the direction, the scale and the position are changed as invariance features; and carrying out grid filtering on the invariance characteristics to obtain target characteristics.
(5) Image matching: performing K-means clustering on all the extracted feature descriptors, wherein a clustering center becomes Visual Words; dividing the training data into K groups according to the clustering centers, wherein each group has the same clustering center, iterating according to the clustering process until the preset maximum depth L is reached to obtain a vocabulary tree with the depth of L, and taking the vocabulary tree as a matching image library.
(6) And (3) feature matching: any two object images are obtained by inquiring the vocabulary tree, repeated features in regions of repeated structures in the two object images are eliminated, primarily eliminated features are obtained, the primarily eliminated features are input into the RANSAC outlier detection model, filtering processing is carried out on the primarily eliminated features by using epipolar geometry so as to eliminate error features in the primarily eliminated features, iteration is carried out according to the number of the filtered features so as to obtain eliminated features, and the eliminated features are used for updating target features of corresponding object images in a new image set.
(7) Depth reconstruction: sfm file, the depth information of each pixel of the object image is restored using Semi-Global matching (Semi-Global Match).
(8) Incremental reconstruction: integrating the depth images of all object images into a global octree, simultaneously combining the depth images with compatible combined depth information into the same node of the octree, and performing three-dimensional reconstruction on the body of the smooth object; constructing a grid of a smooth object according to a 3D-Delaunay tetrahedron algorithm for reconstruction; performing a maximum flow graph cut algorithm to optimize the overall volume, filtering the bad nodes of the surface; according to the Laplace filtering, local artifact removing processing is carried out on the grids to obtain a three-dimensional model; and carrying out texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
In this embodiment, through matching the image library, the time required for retrieving images in the smooth object three-dimensional reconstruction process is greatly shortened, and then the three-dimensional reconstruction efficiency of the smooth object is improved.
It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.
Based on the same inventive concept, the embodiment of the present application further provides a smooth object three-dimensional reconstruction apparatus for implementing the above mentioned smooth object three-dimensional reconstruction method. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, so specific limitations in one or more embodiments of the smooth object three-dimensional reconstruction apparatus provided below can be referred to the limitations of the smooth object three-dimensional reconstruction method in the foregoing, and details are not repeated herein.
In one embodiment, as shown in fig. 5, there is provided a smooth object three-dimensional reconstruction apparatus 500, comprising: an image module 501, a matching module 502, a reconstruction module 503, and a texture module 504, wherein:
an image module 501, configured to obtain an object image acquired after surface material processing is performed on a smooth object;
a matching module 502, configured to perform image matching processing on the object image according to a target feature in the object image, to obtain a matching image library;
the reconstruction module 503 is configured to perform three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object;
and the texture module 504 is configured to perform texture processing on the three-dimensional model to obtain a target three-dimensional model of the smooth object.
In an embodiment, the matching module 502 is further configured to perform clustering processing on the target features in the object image to obtain a clustering center; and matching the object images according to the distance between the target features in the object images and the clustering center to obtain a matching image library.
In one embodiment, the smooth object three-dimensional reconstruction apparatus 500 further includes a feature extraction module, configured to perform feature extraction processing on the object image according to feature invariance to obtain candidate features; screening the candidate features according to a preset stability metric to obtain screened features; and performing characteristic filtering processing on the screened characteristics to obtain target characteristics.
In one embodiment, the smooth object three-dimensional reconstruction apparatus 500 further includes a feature processing module, configured to query a preset number of object images from the matching image library to obtain an image set; matching the feature descriptors of the images in the image set to obtain the matched feature descriptors of the images in the image set; according to the matched feature descriptors, removing features in the image set to obtain removed features; and updating the target feature in the image set by using the removed feature.
In an embodiment, the reconstructing module 503 is further configured to perform three-dimensional shape processing on the depth image according to an octree, so as to obtain a first three-dimensional model of the smooth object; carrying out grid generation processing on the first three-dimensional model to obtain a second three-dimensional model; and carrying out artifact processing on the second three-dimensional model to obtain a three-dimensional model.
In one embodiment, the smooth object three-dimensional reconstruction apparatus 500 further includes a depth image module, configured to perform a semi-global matching process on the object image according to the camera parameter information of the object image in the matching image library, so as to obtain depth information of each pixel in the object image; and generating a depth image of the object image according to the depth information.
In one embodiment, the smooth object three-dimensional reconstruction apparatus 500 further includes a depth information module, configured to calculate, according to the disparity range in the camera parameter information, a disparity range between each pixel in the object image and a pixel in the disparity range, and calculate depth candidate information between each pixel in the object image and a pixel in the disparity range; and generating depth information according to the similarity between the depth candidate information.
The modules in the smooth object three-dimensional reconstruction device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 6. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing relevant data of three-dimensional reconstruction of smooth objects, such as object images, depth images, three-dimensional models and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a smooth object three-dimensional reconstruction method.
Those skilled in the art will appreciate that the architecture shown in fig. 6 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is further provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method embodiments when executing the computer program.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), Magnetic Random Access Memory (MRAM), Ferroelectric Random Access Memory (FRAM), Phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims (10)
1. A method for three-dimensional reconstruction of smooth objects, the method comprising:
acquiring an object image acquired after surface material treatment is carried out on a smooth object; the mirror reflection interference of the surface of the smooth object after the surface material treatment is reduced or disappears;
according to the stability degree of the candidate features in the object image, screening out features with the stability degree higher than a preset stability measure from the candidate features as screened features;
determining invariance characteristics from the screened characteristics, and filtering the invariance characteristics to obtain target characteristics in the object image;
performing image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library;
according to the depth image of the object image in the matching image library, performing three-dimensional reconstruction processing on the smooth object to obtain a three-dimensional model of the smooth object;
and according to the original texture of the smooth object, performing texture reconstruction on the surface texture of the three-dimensional model to obtain a target three-dimensional model of the smooth object.
2. The method according to claim 1, wherein the performing image matching processing on the object image according to the target feature in the object image to obtain a matching image library comprises:
clustering the target characteristics in the object image to obtain a clustering center;
and matching the object images according to the distance between the target features in the object images and the clustering center to obtain a matching image library.
3. The method according to claim 1, before performing image matching processing on the object image according to a target feature in the object image to obtain a matching image library, further comprising:
according to the feature invariance, performing feature extraction processing on the object image to obtain candidate features;
screening the candidate features according to a preset stability metric to obtain screened features;
and performing characteristic filtering processing on the screened characteristics to obtain the target characteristics.
4. The method according to claim 1, wherein after performing image matching processing on the object image according to a target feature in the object image to obtain a matching image library, the method further comprises:
inquiring a preset number of object images from the matching image library to obtain an image set;
matching the feature descriptors of the images in the image set to obtain matched feature descriptors of the images in the image set;
according to the matched feature descriptors, removing features in the image set to obtain removed features;
and updating the target features in the image set by using the removed features.
5. The method according to claim 1, wherein the performing a three-dimensional reconstruction process on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object comprises:
according to the octree, carrying out three-dimensional shape processing on the depth image to obtain a first three-dimensional model of the smooth object;
carrying out grid generation processing on the first three-dimensional model to obtain a second three-dimensional model;
and carrying out artifact processing on the second three-dimensional model to obtain the three-dimensional model.
6. The method according to claim 1, before performing a three-dimensional reconstruction process on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object, further comprising:
performing semi-global matching processing on the object image according to the camera parameter information of the object image in the matching image library to obtain the depth information of each pixel in the object image;
and generating a depth image of the object image according to the depth information.
7. The method according to claim 6, wherein the performing a semi-global matching process on the object image according to the camera parameter information of the object image in the matching image library to obtain the depth information of each pixel in the object image comprises:
calculating depth candidate information between each pixel in the object image and a pixel in the parallax range according to the parallax range in the camera parameter information;
and generating the depth information according to the similarity between the depth candidate information.
8. An apparatus for three-dimensional reconstruction of smooth objects, the apparatus comprising:
the image module is used for acquiring an object image acquired after surface material treatment is carried out on a smooth object; the mirror reflection interference of the surface of the smooth object after the surface material treatment is reduced or disappears;
the characteristic extraction module is used for screening out the characteristics with the stability degree higher than the preset stability measure from the candidate characteristics according to the stability degree of the candidate characteristics in the object image and taking the characteristics as the screened characteristics; determining invariance characteristics from the screened characteristics, and filtering the invariance characteristics to obtain target characteristics in the object image;
the matching module is used for carrying out image matching processing on the object image according to the target characteristics in the object image to obtain a matching image library;
the reconstruction module is used for performing three-dimensional reconstruction processing on the smooth object according to the depth image of the object image in the matching image library to obtain a three-dimensional model of the smooth object;
and the texture module is used for reconstructing the surface texture of the three-dimensional model according to the original texture of the smooth object to obtain the target three-dimensional model of the smooth object.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210183551.0A CN114241141B (en) | 2022-02-28 | 2022-02-28 | Smooth object three-dimensional reconstruction method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210183551.0A CN114241141B (en) | 2022-02-28 | 2022-02-28 | Smooth object three-dimensional reconstruction method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114241141A CN114241141A (en) | 2022-03-25 |
| CN114241141B true CN114241141B (en) | 2022-05-24 |
Family
ID=80748279
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210183551.0A Active CN114241141B (en) | 2022-02-28 | 2022-02-28 | Smooth object three-dimensional reconstruction method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114241141B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115100383B (en) * | 2022-08-24 | 2022-11-15 | 深圳星坊科技有限公司 | Three-dimensional reconstruction method, device and equipment for mirror surface object based on common light source |
| CN115222896B (en) * | 2022-09-20 | 2023-05-23 | 荣耀终端有限公司 | Three-dimensional reconstruction method, three-dimensional reconstruction device, electronic equipment and computer readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109377524A (en) * | 2018-10-29 | 2019-02-22 | 山东师范大学 | A kind of single image depth recovery method and system |
| CN111344746A (en) * | 2017-11-22 | 2020-06-26 | 华为技术有限公司 | Three-dimensional (3D) reconstruction method for dynamic scene by using reconfigurable hybrid imaging system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10839542B2 (en) * | 2019-01-29 | 2020-11-17 | The Chinese University Of Hong Kong | Systems and methods for 3D laparoscopic surface reconstruction |
| US11671580B2 (en) * | 2019-05-14 | 2023-06-06 | Board Of Supervisors Of Louisiana State University And Agricultural And Mechanical College | System and method for reconstructing 3-D shapes of objects from reflection images |
-
2022
- 2022-02-28 CN CN202210183551.0A patent/CN114241141B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111344746A (en) * | 2017-11-22 | 2020-06-26 | 华为技术有限公司 | Three-dimensional (3D) reconstruction method for dynamic scene by using reconfigurable hybrid imaging system |
| CN109377524A (en) * | 2018-10-29 | 2019-02-22 | 山东师范大学 | A kind of single image depth recovery method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114241141A (en) | 2022-03-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111795704B (en) | Method and device for constructing visual point cloud map | |
| Pittaluga et al. | Revealing scenes by inverting structure from motion reconstructions | |
| CN114241141B (en) | Smooth object three-dimensional reconstruction method and device, computer equipment and storage medium | |
| Panek et al. | Meshloc: Mesh-based visual localization | |
| CN114332415B (en) | Three-dimensional reconstruction method and device of power transmission line corridor based on multi-view technology | |
| CN109842811B (en) | Method and device for implanting push information into video and electronic equipment | |
| CN110176079B (en) | Three-dimensional model deformation algorithm based on quasi-conformal mapping | |
| CN111680678A (en) | Target area identification method, device, equipment and readable storage medium | |
| CN116402978A (en) | Dense three-dimensional reconstruction method based on binocular vision structural characteristics | |
| Huo et al. | Three-dimensional mechanical parts reconstruction technology based on two-dimensional image | |
| CN110147460B (en) | Three-dimensional model retrieval method and device based on convolutional neural network and multi-view map | |
| CN113592015B (en) | Method and device for positioning and training feature matching network | |
| CN109857895B (en) | Stereo vision retrieval method and system based on multi-loop view convolutional neural network | |
| CN111260794A (en) | Outdoor augmented reality application method based on cross-source image matching | |
| CN115115825A (en) | Method and device for detecting object in image, computer equipment and storage medium | |
| CN117078790B (en) | Image generation method, device, computer equipment and storage medium | |
| Guo et al. | Improved 3D point cloud segmentation for accurate phenotypic analysis of cabbage plants using deep learning and clustering algorithms | |
| CN115630660B (en) | Barcode positioning method and device based on convolutional neural network | |
| CN111414802B (en) | Protein data characteristic extraction method | |
| Budianti et al. | Background blurring and removal for 3d modelling of cultural heritage objects | |
| CN114863201A (en) | Training method and device of three-dimensional detection model, computer equipment and storage medium | |
| Qiao et al. | Point clouds segmentation of rapeseed siliques based on sparse-dense point clouds mapping | |
| Singh et al. | A generative modelling technique for 3D reconstruction from a single 2D image | |
| Sattler et al. | Exploiting spatial and co-visibility relations for image-based localization | |
| CN117351246B (en) | Mismatching pair removing method, system and readable medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |