Three-dimensional object reconstruction method and system based on edge computing power

Abstract

The invention provides a three-dimensional object reconstruction method and system based on edge computing power, and the method comprises the steps: collecting a plurality of pieces of original image data correspondingly generated by a target object under multiple visual angles in real time through a preset distributed edge device, and carrying out the real-time alignment processing of the plurality of pieces of original image data through a preset dynamic timestamp synchronization algorithm, correspondingly generating a plurality of standard image data; detecting a plurality of adjacent edge nodes correspondingly contained in the distributed edge equipment in real time, and acquiring a spatial topological relation correspondingly generated by the plurality of adjacent edge nodes in real time; creating a corresponding drift compensation model in real time according to the spatial topological relation, and correcting the offset error of the plurality of standard image data in real time through the drift compensation model to correspondingly generate a plurality of target image data; and correspondingly inputting the plurality of target image data into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time. The reconstruction efficiency of the object can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional modeling technology, and in particular to a three-dimensional object reconstruction method and system based on edge computing power. Background Art

[0002] With the advancement of science and technology and the rapid development of productivity, people have developed various types of edge devices, such as smart cameras, edge servers, and smart gateways, which have been deeply applied in many fields.

[0003] Among them, in the existing industrial production process, in order to detect whether the surface of the products produced in real time has defects, most of the existing technologies will set up a detection area in the production line, and will set up several sensors in the detection area to collect the surface data of the products in real time and complete subsequent judgments.

[0004] Furthermore, in actual applications, since the data formats and acquisition frequencies generated by different existing sensors are not completely consistent with each other, precise time and space calibration is required during the data fusion process. Otherwise, the constructed three-dimensional model will be misaligned or distorted, which will reduce the product detection efficiency. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a three-dimensional object reconstruction method and system based on edge computing power to solve the problem of misalignment or distortion that easily occurs in the process of real-time construction of product three-dimensional models in the existing technology.

[0006] The first aspect of the embodiment of the present invention proposes: A method for reconstructing a three-dimensional object based on edge computing power, wherein the method comprises: Collecting a plurality of raw image data corresponding to a target object under multiple perspectives in real time through a preset distributed edge device, and performing real-time alignment processing on the plurality of raw image data through a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data; Detecting in real time a number of adjacent edge nodes corresponding to the distributed edge device, and obtaining in real time a spatial topological relationship generated by the number of adjacent edge nodes; Creating a corresponding drift compensation model in real time according to the spatial topological relationship, and correcting the offset errors of the plurality of standard image data in real time by using the drift compensation model to correspondingly generate a plurality of target image data; The target image data are input into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time.

[0007] The beneficial effect of the present invention is that by collecting the original image data of the target object under multiple perspectives in real time, the object information corresponding to the current target object can be obtained. Based on this, in order to facilitate subsequent modeling, the present invention can perform real-time alignment processing on several current original image data, thereby generating several standard image data, thereby eliminating offset errors. Based on this, the present invention can also obtain the spatial topological relationship corresponding to the edge device in real time, thereby obtaining the required drift compensation model, and finally obtain the required target image data through the model, and finally complete the construction of the target three-dimensional model, thereby effectively avoiding the occurrence of dislocation and distortion, and correspondingly improving the efficiency of model creation.

[0008] Furthermore, the step of performing real-time alignment processing on the plurality of original image data by using a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data includes: When a plurality of the original image data are acquired in real time, the acquisition time corresponding to each of the original image data is detected in sequence; The corresponding time base is determined in real time according to the acquisition time through a preset PTP protocol, and a plurality of the original image data are aligned in real time according to the time base to generate a plurality of the standard image data accordingly.

[0009] Furthermore, the step of performing real-time alignment processing on the plurality of original image data according to the time reference to correspondingly generate the plurality of standard image data includes: When the time reference is acquired in real time, a target time difference between the acquisition time of each of the original image data and the time reference is detected in real time; A time compensation factor corresponding to each of the original image data is determined in real time according to the time reference and the target time difference through a preset algorithm, and the acquisition time of each of the original image data is corrected to be synchronized with the time reference according to the time compensation factor.

[0010] Furthermore, the expression of the preset algorithm is:

[0011] Among them, Δ t i represents the time reference, Δ t targe represents the target time difference, k represents the adjustment parameter, α i Represents the time compensation factor of the i-th original image data.

[0012] Furthermore, the step of creating a corresponding drift compensation model in real time according to the spatial topological relationship includes: When the spatial topological relationship is acquired in real time, a preset neural network is called out in real time from a preset database; Performing a full scan of the spatial topological relationship to detect in real time a number of topological nodes corresponding to the spatial topological relationship; Detecting in real time the target topology value contained in each of the topological nodes, and creating in real time a corresponding topology matrix according to each of the target topological values; The drift compensation model is created in real time according to the topology matrix and the preset neural network, and the topology matrix is ​​unique.

[0013] Furthermore, the step of creating the drift compensation model in real time according to the topology matrix and the preset neural network includes: When the topology matrix is ​​acquired in real time, the encoding layer, the training layer, and the output layer contained in the preset neural network are detected in real time; Performing real-time encoding processing on the topology matrix through a preset encoder in the encoding layer to generate corresponding target training codes in real time, and detecting a number of corresponding training nodes included in the training layer in real time; The drift compensation model is created in real time according to the target training code and a plurality of the training nodes, and the target training code is unique.

[0014] Furthermore, the step of creating the drift compensation model in real time according to the target training code and the plurality of training nodes includes: When the target training code is acquired in real time, the target training code is split into training sub-codes respectively adapted to each of the training nodes; Loading each of the training subcodes into each of the training nodes to perform iterative training on each of the training nodes, and determining in real time whether a highest training value is generated during the iterative training; If it is determined in real time that the highest training value is generated in the iterative training, several target nodes are generated accordingly, and the target nodes are integrated to generate the drift compensation model in real time, and the drift compensation model is output accordingly through the output module.

[0015] The second aspect of the embodiment of the present invention proposes: A three-dimensional object reconstruction system based on edge computing power, wherein the system includes: An acquisition module is configured to acquire a plurality of raw image data corresponding to a target object from multiple perspectives in real time through a preset distributed edge device, and to perform real-time alignment processing on the plurality of raw image data using a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data; A detection module is used to detect in real time a number of adjacent edge nodes corresponding to the distributed edge device, and obtain in real time the spatial topological relationship generated by the number of adjacent edge nodes; a correction module, configured to create a corresponding drift compensation model in real time according to the spatial topological relationship, and to correct the offset errors of the plurality of standard image data in real time using the drift compensation model to correspondingly generate a plurality of target image data; The execution model is used to input the target image data into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time.

[0016] Furthermore, the acquisition module is specifically used to: When a plurality of the original image data are acquired in real time, the acquisition time corresponding to each of the original image data is detected in sequence; The corresponding time base is determined in real time according to the acquisition time through a preset PTP protocol, and a plurality of the original image data are aligned in real time according to the time base to generate a plurality of the standard image data accordingly.

[0017] Furthermore, the acquisition module is specifically used to: When the time reference is acquired in real time, a target time difference between the acquisition time of each of the original image data and the time reference is detected in real time; A time compensation factor corresponding to each of the original image data is determined in real time according to the time reference and the target time difference through a preset algorithm, and the acquisition time of each of the original image data is corrected to be synchronized with the time reference according to the time compensation factor.

[0018] Furthermore, the expression of the preset algorithm is:

[0019] Among them, Δ t i represents the time reference, Δ t targe represents the target time difference, k represents the adjustment parameter, α i Represents the time compensation factor of the i-th original image data.

[0020] Furthermore, the correction module is specifically used to: When the spatial topological relationship is acquired in real time, a preset neural network is called out in real time from a preset database; Performing a full scan of the spatial topological relationship to detect in real time a number of topological nodes corresponding to the spatial topological relationship; Detecting in real time the target topology value contained in each of the topological nodes, and creating in real time a corresponding topology matrix according to each of the target topological values; The drift compensation model is created in real time according to the topology matrix and the preset neural network, and the topology matrix is ​​unique.

[0021] Furthermore, the correction module is specifically used to: When the topology matrix is ​​acquired in real time, the encoding layer, the training layer, and the output layer contained in the preset neural network are detected in real time; Performing real-time encoding processing on the topology matrix through a preset encoder in the encoding layer to generate corresponding target training codes in real time, and detecting a number of corresponding training nodes included in the training layer in real time; The drift compensation model is created in real time according to the target training code and a plurality of the training nodes, and the target training code is unique.

[0022] Furthermore, the correction module is specifically used to: When the target training code is acquired in real time, the target training code is split into training sub-codes respectively adapted to each of the training nodes; Loading each of the training subcodes into each of the training nodes to perform iterative training on each of the training nodes, and determining in real time whether a highest training value is generated during the iterative training; If it is determined in real time that the highest training value is generated in the iterative training, several target nodes are generated accordingly, and the target nodes are integrated to generate the drift compensation model in real time, and the drift compensation model is output accordingly through the output module.

[0023] The third aspect of the embodiment of the present invention proposes: A computer comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the three-dimensional object reconstruction method based on edge computing power as described above is implemented.

[0024] The fourth aspect of the embodiments of the present invention proposes: A readable storage medium stores a computer program thereon, wherein when the program is executed by a processor, the three-dimensional object reconstruction method based on edge computing power as described above is implemented.

[0025] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A flowchart of a three-dimensional object reconstruction method based on edge computing provided by the first embodiment of the present invention; Figure 2 This is a structural block diagram of a three-dimensional object reconstruction system based on edge computing power provided in the third embodiment of the present invention.

[0027] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0028] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate several embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present invention.

[0029] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. The terms "vertical," "horizontal," "left," "right," and similar expressions used herein are for illustrative purposes only.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0031] See also Figure 1 , shown is the three-dimensional object reconstruction method based on edge computing provided by the first embodiment of the present invention. The three-dimensional object reconstruction method based on edge computing provided by this embodiment can avoid the phenomenon of dislocation in the process of creating the model, and can quickly and effectively construct the required three-dimensional model, thereby correspondingly improving the reconstruction efficiency of the model.

[0032] Specifically, this embodiment provides: A three-dimensional object reconstruction method based on edge computing power specifically includes the following steps: Step S10: collecting a plurality of raw image data corresponding to the target object under multiple viewing angles in real time through a preset distributed edge device, and performing real-time alignment processing on the plurality of raw image data through a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data; It should be noted that in order to accurately construct a three-dimensional model of each object in the production line, the present invention needs to accurately collect object information corresponding to each object. It should be noted that in order to detect whether the surface of the object has defects, the present invention will use a pre-set distributed edge device to collect images of each object in the production line. Specifically, the edge device can be set as an intelligent camera, and can collect the original image data generated by each object at different perspectives in real time through the edge device, so that the complete image data corresponding to the current surface of each object can be obtained. It should be noted that since there may be a certain time difference in the image acquisition process of each edge device, the current original image data may not be generated at the same time node. Based on this, the present invention will immediately align each current original image data in real time through the existing dynamic timestamp synchronization algorithm, so that the acquisition time node of each current original image data can be synchronized to the same time node, that is, it can generate the required standard image data in real time for subsequent processing.

[0033] Step S20: detecting in real time a number of adjacent edge nodes corresponding to the distributed edge device, and obtaining in real time a spatial topological relationship generated by the number of adjacent edge nodes; Among them, it should be noted that, in order to facilitate the subsequent model creation, it is also necessary to judge in real time whether the current standard image data needs data compensation. Based on this, the present invention also needs to detect in real time the corresponding adjacent edge nodes contained in the current distributed edge device. At the same time, it will also obtain in real time the corresponding spatial topological relationship between each adjacent edge node, that is, the communication connection generated between each adjacent edge node is detected in real time for subsequent processing.

[0034] Step S30, creating a corresponding drift compensation model in real time according to the spatial topological relationship, and correcting the offset errors of the plurality of standard image data in real time by using the drift compensation model to correspondingly generate a plurality of target image data; Among them, it should be noted that after the required spatial topological relationship is obtained in real time through the above steps, the current spatial topological relationship can be used as the training basis. At the same time, the existing neural network is called out, and the current neural network will be trained with a specific model through the current spatial topological relationship, so that the required drift compensation model can be generated synchronously. Among them, it should be pointed out that the drift compensation model is mainly used to detect whether there are data defects in the original image data, and after determining that there are data defects, data compensation can be performed in time, so that complete image data corresponding to the object can be formed to facilitate subsequent processing.

[0035] In step S40 , the target image data are input into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time.

[0036] Among them, it should be noted that after the target image data corresponding to the above-mentioned target object is finally obtained through the above steps, the current target image data will be immediately input into the existing three-dimensional programs such as UG or SolidWorks in real time, and the target three-dimensional model corresponding to the above-mentioned target object can be finally created in real time through the current three-dimensional program, thereby effectively avoiding the phenomenon of model missing, and at the same time being able to accurately create the required three-dimensional model, which correspondingly improves the model creation efficiency.

[0037] Second embodiment Furthermore, the step of performing real-time alignment processing on the plurality of original image data by using a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data includes: When a plurality of the original image data are acquired in real time, the acquisition time corresponding to each of the original image data is detected in sequence; The corresponding time base is determined in real time according to the acquisition time through a preset PTP protocol, and a plurality of the original image data are aligned in real time according to the time base to generate a plurality of the standard image data accordingly.

[0038] It should be noted that in order to objectively and accurately complete the alignment processing of the current raw image data, the present invention will call out the existing dynamic timestamp synchronization algorithm in real time. Specifically, the present invention will first detect the acquisition time corresponding to each current raw image data. At the same time, it can determine the time base corresponding to each current acquisition time in real time through the existing PTP protocol, and can complete the real-time alignment processing of the current raw image data in real time according to the current time base. It should be pointed out that in most of the acquisition times of the current raw image data, some images have the same acquisition time, while other images have different acquisition times. Based on this, it will be judged in real time which of the two parts has a larger number, that is, it will be judged in real time whether the number of the same acquisition time is greater than the number of the different acquisition time. Specifically, if so, the time node corresponding to the same current acquisition time is used as the above-mentioned time base. Correspondingly, if not, the time node corresponding to the different current acquisition time is used as the above-mentioned time base, so as to facilitate subsequent processing.

[0039] Furthermore, the step of performing real-time alignment processing on the plurality of original image data according to the time reference to correspondingly generate the plurality of standard image data includes: When the time reference is acquired in real time, a target time difference between the acquisition time of each of the original image data and the time reference is detected in real time; A time compensation factor corresponding to each of the original image data is determined in real time according to the time reference and the target time difference through a preset algorithm, and the acquisition time of each of the original image data is corrected to be synchronized with the time reference according to the time compensation factor.

[0040] Among them, it should be noted that after the required time base is obtained in real time through the above steps, the target time difference between the acquisition time of each current original image data and the current time base will be immediately determined. Specifically, if the acquisition time is the same as the time base, the corresponding target time difference is 0. Correspondingly, if the acquisition time is different from the time base, the target time difference between the two will be calculated in real time. Based on this, the present invention will also immediately call out the preset algorithm, and can immediately determine the time compensation factor corresponding to each current original image data in real time according to the current time base and the target time difference. Based on this, the acquisition time of each current original image data will eventually be adaptively corrected to the above time base according to the size of the current time compensation factor, so that the acquisition time of each current original image data can be quickly and effectively synchronized for subsequent processing.

[0041] Furthermore, the expression of the preset algorithm is:

[0042] Among them, Δ t i represents the time reference, Δ t targe represents the target time difference, k represents the adjustment parameter, α i Represents the time compensation factor of the i-th original image data.

[0043] Furthermore, the step of creating a corresponding drift compensation model in real time according to the spatial topological relationship includes: When the spatial topological relationship is acquired in real time, a preset neural network is called out in real time from a preset database; Performing a full scan of the spatial topological relationship to detect in real time a number of topological nodes corresponding to the spatial topological relationship; Detecting in real time the target topology value contained in each of the topological nodes, and creating in real time a corresponding topology matrix according to each of the target topological values; The drift compensation model is created in real time according to the topology matrix and the preset neural network, and the topology matrix is ​​unique.

[0044] Among them, it should be noted that, in order to accurately and effectively train the required drift compensation model, after the present invention obtains the required spatial topological relationship in real time through the above steps, it is necessary to call out the existing neural network in the preset database accordingly. Among them, it should be pointed out that the present invention will immediately perform a full scan of the current spatial topological relationship, and can detect in real time several topological nodes corresponding to the current spatial topological relationship. Among them, it should be noted that, in order to facilitate control, the existing nodes will all have corresponding control values ​​set inside them. Similarly, the present invention can also detect in real time the target topological values ​​respectively contained in the interior of each current topological node, and can immediately perform matrix processing on each current target topological value to generate the required topological matrix in real time. The topological matrix can be recognized by the above-mentioned neural network, so that the topological matrix can be input into the interior of the above-mentioned neural network again for subsequent processing.

[0045] Furthermore, the step of creating the drift compensation model in real time according to the topology matrix and the preset neural network includes: When the topology matrix is ​​acquired in real time, the encoding layer, the training layer, and the output layer contained in the preset neural network are detected in real time; Performing real-time encoding processing on the topology matrix through a preset encoder in the encoding layer to generate corresponding target training codes in real time, and detecting a number of corresponding training nodes included in the training layer in real time; The drift compensation model is created in real time according to the target training code and a plurality of the training nodes, and the target training code is unique.

[0046] It should be noted that after the required topological matrix is ​​obtained in real time through the above steps, in order to facilitate subsequent training, it is necessary to obtain the network structure contained in the current neural network in real time. Specifically, the neural network provided by the present invention includes a coding layer, a training layer and an output layer. Specifically, the present invention will first perform real-time encoding processing on the current topological matrix through a preset encoder in the current coding layer, and can encode the corresponding target training code in real time, and then input the target training code into the above-mentioned training layer. It should be pointed out that the training layer contains several training nodes, and the training of the current target training code will be completed in sequence through the current several training nodes to facilitate subsequent processing.

[0047] Furthermore, the step of creating the drift compensation model in real time according to the target training code and the plurality of training nodes includes: When the target training code is acquired in real time, the target training code is split into training sub-codes respectively adapted to each of the training nodes; Loading each of the training subcodes into each of the training nodes to perform iterative training on each of the training nodes, and determining in real time whether a highest training value is generated during the iterative training; If it is determined in real time that the highest training value is generated in the iterative training, several target nodes are generated accordingly, and the target nodes are integrated to generate the drift compensation model in real time, and the drift compensation model is output accordingly through the output module.

[0048] It should be noted that after the required target training code is obtained in real time through the above steps, since the internal training corresponding to each current training node is not the same, based on this, the present invention will detect the training attributes corresponding to each current training node in real time, and will directly perform corresponding attribute matching within the current target training code according to the current training attributes, so that the training sub-code corresponding to the training attribute can be matched within the current target training code. Based on this, each current training sub-code is split and processed. At the same time, each current training sub-code is loaded into each current training node, and finally iterative training is performed within each current training node. It is also determined in real time whether each current training node generates the highest training value. Specifically, if so, it indicates that the training is completed and the required target node is trained. Based on this, each current target node is immediately integrated to form the required drift compensation model. Correspondingly, if not, the training is not completed, and iterative training needs to continue until the highest training value is trained, so that the required drift compensation model can be trained objectively and accurately for subsequent processing.

[0049] See also Figure 2 , the third embodiment of the present invention provides: A three-dimensional object reconstruction system based on edge computing power, wherein the system includes: An acquisition module is configured to acquire a plurality of raw image data corresponding to a target object from multiple perspectives in real time through a preset distributed edge device, and to perform real-time alignment processing on the plurality of raw image data using a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data; A detection module is used to detect in real time a number of adjacent edge nodes corresponding to the distributed edge device, and obtain in real time the spatial topological relationship generated by the number of adjacent edge nodes; a correction module, configured to create a corresponding drift compensation model in real time according to the spatial topological relationship, and to correct the offset errors of the plurality of standard image data in real time using the drift compensation model to correspondingly generate a plurality of target image data; The execution model is used to input the target image data into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time.

[0050] Furthermore, the acquisition module is specifically used to: When a plurality of the original image data are acquired in real time, the acquisition time corresponding to each of the original image data is detected in sequence; The corresponding time base is determined in real time according to the acquisition time through a preset PTP protocol, and a plurality of the original image data are aligned in real time according to the time base to generate a plurality of the standard image data accordingly.

[0051] Furthermore, the acquisition module is specifically used to: When the time reference is acquired in real time, a target time difference between the acquisition time of each of the original image data and the time reference is detected in real time; A time compensation factor corresponding to each of the original image data is determined in real time according to the time reference and the target time difference through a preset algorithm, and the acquisition time of each of the original image data is corrected to be synchronized with the time reference according to the time compensation factor.

[0052] Furthermore, the expression of the preset algorithm is:

[0053] Among them, Δ t i represents the time reference, Δ t targe represents the target time difference, k represents the adjustment parameter, α i Represents the time compensation factor of the i-th original image data.

[0054] Furthermore, the correction module is specifically used to: When the spatial topological relationship is acquired in real time, a preset neural network is called out in real time from a preset database; Performing a full scan of the spatial topological relationship to detect in real time a number of topological nodes corresponding to the spatial topological relationship; Detecting in real time the target topology value contained in each of the topological nodes, and creating in real time a corresponding topology matrix according to each of the target topological values; The drift compensation model is created in real time according to the topology matrix and the preset neural network, and the topology matrix is ​​unique.

[0055] Furthermore, the correction module is specifically used to: When the topology matrix is ​​acquired in real time, the encoding layer, the training layer, and the output layer contained in the preset neural network are detected in real time; Performing real-time encoding processing on the topology matrix through a preset encoder in the encoding layer to generate corresponding target training codes in real time, and detecting a number of corresponding training nodes included in the training layer in real time; The drift compensation model is created in real time according to the target training code and a plurality of the training nodes, and the target training code is unique.

[0056] Furthermore, the correction module is specifically used to: When the target training code is acquired in real time, the target training code is split into training sub-codes respectively adapted to each of the training nodes; Loading each of the training subcodes into each of the training nodes to perform iterative training on each of the training nodes, and determining in real time whether a highest training value is generated during the iterative training; If it is determined in real time that the highest training value is generated in the iterative training, several target nodes are generated accordingly, and the target nodes are integrated to generate the drift compensation model in real time, and the drift compensation model is output accordingly through the output module.

[0057] A fourth embodiment of the present invention provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the three-dimensional object reconstruction method based on edge computing power as described above is implemented.

[0058] A fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the three-dimensional object reconstruction method based on edge computing power as described above is implemented.

[0059] To sum up, the three-dimensional object reconstruction method and system based on edge computing provided by the above embodiments of the present invention can effectively avoid the phenomenon of model dislocation, and at the same time can quickly and effectively create the required three-dimensional model, thereby improving the efficiency of model creation.

[0060] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0061] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0062] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0063] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0064] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0065] The above-described embodiments merely illustrate several embodiments of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of the present invention. Therefore, the scope of the present invention shall be determined by the appended claims.

Claims

1. A three-dimensional object reconstruction method based on edge computing, characterized in that: The method comprises: Collecting a plurality of raw image data corresponding to a target object under multiple perspectives in real time through a preset distributed edge device, and performing real-time alignment processing on the plurality of raw image data through a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data; Detecting in real time a number of adjacent edge nodes corresponding to the distributed edge device, and obtaining in real time a spatial topological relationship generated by the number of adjacent edge nodes; Creating a corresponding drift compensation model in real time according to the spatial topological relationship, and correcting the offset errors of the plurality of standard image data in real time by using the drift compensation model to correspondingly generate a plurality of target image data; The target image data are input into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time.

2. The method for 3D object reconstruction based on edge computing according to claim 1, wherein: The step of performing real-time alignment processing on the plurality of original image data by using a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data includes: When a plurality of the original image data are acquired in real time, the acquisition time corresponding to each of the original image data is detected in sequence; The corresponding time base is determined in real time according to the acquisition time through a preset PTP protocol, and a plurality of the original image data are aligned in real time according to the time base to generate a plurality of the standard image data accordingly.

3. The edge computing-based 3D object reconstruction method according to claim 2, wherein: The step of performing real-time alignment processing on the plurality of original image data according to the time reference to correspondingly generate the plurality of standard image data includes: When the time reference is acquired in real time, a target time difference between the acquisition time of each of the original image data and the time reference is detected in real time; A time compensation factor corresponding to each of the original image data is determined in real time according to the time reference and the target time difference through a preset algorithm, and the acquisition time of each of the original image data is corrected to be synchronized with the time reference according to the time compensation factor.

4. The method for 3D object reconstruction based on edge computing according to claim 3, wherein: The expression of the preset algorithm is: Among them, Δ t i represents the time reference, Δ t targe represents the target time difference, k represents the adjustment parameter, α i Represents the time compensation factor of the i-th original image data.

5. The method for 3D object reconstruction based on edge computing according to claim 1, wherein: The step of creating a corresponding drift compensation model in real time according to the spatial topological relationship includes: When the spatial topological relationship is acquired in real time, a preset neural network is called out in real time from a preset database; Performing a full scan of the spatial topological relationship to detect in real time a number of topological nodes corresponding to the spatial topological relationship; Detecting in real time the target topology value contained in each of the topological nodes, and creating in real time a corresponding topology matrix according to each of the target topological values; The drift compensation model is created in real time according to the topology matrix and the preset neural network, and the topology matrix is ​​unique.

6. The method for 3D object reconstruction based on edge computing according to claim 5, characterized in that: The step of creating the drift compensation model in real time according to the topology matrix and the preset neural network includes: When the topology matrix is ​​acquired in real time, the encoding layer, the training layer, and the output layer contained in the preset neural network are detected in real time; Performing real-time encoding processing on the topology matrix through a preset encoder in the encoding layer to generate corresponding target training codes in real time, and detecting a number of corresponding training nodes included in the training layer in real time; The drift compensation model is created in real time according to the target training code and a plurality of the training nodes, and the target training code is unique.

7. The method for 3D object reconstruction based on edge computing according to claim 6, characterized in that: The step of creating the drift compensation model in real time according to the target training code and the plurality of training nodes comprises: When the target training code is acquired in real time, the target training code is split into training sub-codes respectively adapted to each of the training nodes; Loading each of the training subcodes into each of the training nodes to perform iterative training on each of the training nodes, and determining in real time whether a highest training value is generated during the iterative training; If it is determined in real time that the highest training value is generated in the iterative training, several target nodes are generated accordingly, and the target nodes are integrated to generate the drift compensation model in real time, and the drift compensation model is output accordingly through the output module.

8. A three-dimensional object reconstruction system based on edge computing, characterized in that: The system comprises: An acquisition module is configured to acquire a plurality of raw image data corresponding to a target object from multiple perspectives in real time through a preset distributed edge device, and to perform real-time alignment processing on the plurality of raw image data using a preset dynamic timestamp synchronization algorithm to generate a plurality of corresponding standard image data; A detection module is used to detect in real time a number of adjacent edge nodes corresponding to the distributed edge device, and obtain in real time the spatial topological relationship generated by the number of adjacent edge nodes; a correction module, configured to create a corresponding drift compensation model in real time according to the spatial topological relationship, and to correct the offset errors of the plurality of standard image data in real time by using the drift compensation model to correspondingly generate a plurality of target image data; The execution model is used to input the target image data into a preset three-dimensional program to create a target three-dimensional model corresponding to the target object in real time.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the three-dimensional object reconstruction method based on edge computing power as described in any one of claims 1 to 7 is implemented.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the three-dimensional object reconstruction method based on edge computing power as described in any one of claims 1 to 7 is implemented.

Images