House point cloud registration method and device, electronic device and readable storage medium

Abstract

The present application discloses a house point cloud registration method and device, an electronic device and a readable storage medium, and belongs to the field of computer technology. Among them, the house point cloud registration method includes: based on the source house point cloud, determining a first optimization target. Based on the source house point cloud, determining a degree of freedom registration method. Performing a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative pose of the source house point cloud and the target house point cloud. Based on the source house point cloud and the target house point cloud, constructing a source point cloud pyramid and a target point cloud pyramid. Based on the first relative pose, the optimization target and the degree of freedom registration method, performing multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid to obtain a second relative pose of the source house point cloud and the target house point cloud. Based on the second relative pose, splitting the point cloud points into multiple point cloud fragments, performing a second registration on the multiple point cloud fragments, obtaining a global pose, and completing the house point cloud registration.

Description

Technical Field

[0001] The present application belongs to the field of computer technology, and specifically relates to a method and device for house point cloud registration, an electronic device, and a readable storage medium. Background Art

[0002] In the related art, in applications such as VR (Virtual Reality) house viewing, RGB-D (depth image) cameras are often used to shoot houses to obtain three-dimensional point cloud models. The houses being photographed usually have multiple rooms, and need to be photographed at multiple points, and the point cloud registration algorithm is used to estimate the poses between multiple points. In the related art, when solving the point cloud registration problem, there is no targeted optimization for the indoor house scene of VR house viewing, and it is difficult to achieve accurate registration of indoor house point clouds. The wall and ground parts of the point cloud are prone to stratification and other problems. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a house point cloud registration method and device, an electronic device and a readable storage medium, which can solve the problems that the relevant technology has not been optimized specifically for the indoor scene of VR house viewing, it is difficult to achieve accurate registration of indoor house point clouds, and the wall and ground parts of the point cloud are prone to stratification.

[0004] In a first aspect, an embodiment of the present application provides a house point cloud registration method, including: for registering a source house point cloud with a target house point cloud, the house point cloud registration method includes: based on the source house point cloud, determining a first optimization target. Based on the source house point cloud, determining a degree of freedom registration method. Performing a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative pose of the source house point cloud and the target house point cloud. Based on the source house point cloud and the target house point cloud, constructing a source point cloud pyramid and a target point cloud pyramid. Based on the first relative pose, the optimization target and the degree of freedom registration method, performing a multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid to obtain a second relative pose of the source house point cloud and the target house point cloud. Based on the second relative pose, splitting the point cloud points into multiple point cloud fragments, performing a second registration on the multiple point cloud fragments, obtaining a global pose, and completing the house point cloud registration.

[0005] In the second aspect, the embodiment of the present application provides a house point cloud registration device, including a first determination module, a second determination module, a first acquisition module, a first construction module, a second acquisition module and a third acquisition module. The first determination module is used to determine a first optimization target based on a source house point cloud; the second determination module is used to determine a degree of freedom registration method based on the source house point cloud; the first acquisition module is used to perform a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative pose of the source house point cloud and the target house point cloud; the first construction module is used to construct a source point cloud pyramid and a target point cloud pyramid based on the source house point cloud and the target house point cloud; the second acquisition module is used to perform multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid based on the first relative pose, the optimization target and the degree of freedom registration method to obtain a second relative pose of the source house point cloud and the target house point cloud; the third acquisition module, based on the second relative pose, splits the point cloud point into multiple point cloud segments, performs a second registration on the multiple point cloud segments, obtains a global pose, and completes the house point cloud registration.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the house point cloud registration method of the first aspect are implemented.

[0007] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the house point cloud registration method of the first aspect are implemented.

[0008] The house point cloud registration method of the present application can improve the applicability of the house point cloud registration method and the accuracy of the registration results by selecting the first optimization goal and the degree of freedom registration method; reduce the coarse registration pose accuracy requirements by constructing a pyramid; and achieve accurate registration of the point cloud by splitting the point cloud points into multiple point cloud fragments and performing a second registration on the multiple point cloud fragments. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 One of the flow diagrams of the house point cloud registration method provided in the embodiment of the present application is shown;

[0010] Figure 2 The second flowchart of the house point cloud registration method provided in the embodiment of the present application is shown;

[0011] Figure 3 The third flowchart of the house point cloud registration method provided in the embodiment of the present application is shown;

[0012] Figure 4 A fourth flow chart of the house point cloud registration method provided in an embodiment of the present application is shown;

[0013] Figure 5 The fifth flow chart of the house point cloud registration method provided in the embodiment of the present application is shown;

[0014] Figure 6 The sixth flow chart of the house point cloud registration method provided in the embodiment of the present application is shown;

[0015] Figure 7 One of the schematic diagrams of the housing point cloud registration method provided in the embodiment of the present application is shown;

[0016] Figure 8 The second schematic diagram of the housing point cloud registration method provided in the embodiment of the present application is shown;

[0017] Fig. 9 The structure block diagram of the house point cloud registration device provided in the embodiment of the present application is shown;

[0018] Fig.10 A structural block diagram of an electronic device provided in an embodiment of the present application is shown;

[0019] Fig.11 A schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application is shown.

[0020] in, Figures 9 to 11 The corresponding relationship between the reference numerals and the component names is as follows:

[0021] 100: House point cloud registration device; 110: First determination module; 120: Second determination module; 130: First acquisition module; 140: First construction module; 150: Second acquisition module; 160: Third acquisition module; 1000: Electronic device; 1002: Processor; 1004: Memory; 1100: Electronic device; 1101: Radio frequency unit; 1102: Network module; 1103: Audio output unit; 1104: Input unit; 11041: Graphics processor; 11042: Microphone; 1105: Sensor; 1106: Display unit; 11061: Display panel; 1107: User input unit; 11071: Touch panel; 11072: Other input devices; 1108: Interface unit; 1109: Memory; 1110: Processor. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0023] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0024] The following is combined with Figures 1 to 11 , the house point cloud registration method and device, electronic device and readable storage medium provided in the embodiments of the present application are described in detail through specific embodiments and their application scenarios.

[0025] In an embodiment of the present application, a method for aligning a house point cloud is provided. Figure 1 FIG. 1 shows one of the flow charts of the house point cloud registration method provided in the embodiment of the present application, such as Figure 1 As shown, the house point cloud registration method includes:

[0026] Step 102: Determine a first optimization target based on the source house point cloud.

[0027] Step 104: Determine the degree of freedom registration method based on the source house point cloud.

[0028] Step 106: perform a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative position and posture of the source house point cloud and the target house point cloud.

[0029] Step 108 , constructing a source point cloud pyramid and a target point cloud pyramid based on the source house point cloud and the target house point cloud.

[0030] Step 110, based on the first relative pose, optimization target and degree of freedom alignment method, multi-scale point cloud alignment is performed on the source point cloud pyramid and the target point cloud pyramid to obtain a second relative pose of the source house point cloud and the target house point cloud.

[0031] Step 112, based on the second relative pose, split the point cloud points into multiple point cloud segments, perform a second registration on the multiple point cloud segments, obtain the global pose, and complete the house point cloud registration.

[0032] It can be understood that a point cloud refers to a set of point data on the product appearance surface obtained by measuring instruments in reverse engineering. In this embodiment, a house point cloud refers to a point cloud obtained by photographing or measuring a house using an instrument, wherein the house includes multiple indoor rooms.

[0033] In this embodiment, according to the different characteristics of the collected source house point cloud, the first optimization target corresponding to it is selected, so that the optimization target can cope with different usage scenarios, thereby improving the applicability of the house point cloud registration method and the accuracy of the registration results.

[0034] It is understandable that when collecting point clouds, the collection environment and equipment will be different. In this embodiment, the degree of freedom registration method is determined in a targeted manner based on the acquisition method of the source house point cloud and the initial characteristics of the point cloud. This can avoid the subsequent pose optimization from falling into local optimality or transitional optimization, thereby improving the usability of the house point cloud registration method and the accuracy of the registration results.

[0035] In this embodiment, the first point cloud registration is performed on the source house point cloud and the target house point cloud, that is, the source house point cloud and the target house point cloud are roughly matched. The rough matching can use manual assistance or feature point-based methods to obtain the relative pose of the source point cloud and the target point cloud, that is, the first relative pose.

[0036] It is understandable that when registering point clouds, it is usually necessary to perform a rough registration on the two point clouds to be registered to obtain the approximate relative pose between them. In the registration method of the related art, the rough registration method has high requirements. In this embodiment, based on the source house point cloud and the target house point cloud, a point cloud pyramid is constructed, and multi-scale registration is performed to obtain the second relative pose, which can reduce the accuracy requirements for the rough registration pose and improve the accuracy of the final registration effect of the point cloud.

[0037] In this embodiment, after the point cloud of the house is registered between multiple points, the second relative pose between multiple points in the whole house is obtained. Further, the point cloud of each point is refined and split into multiple point cloud fragments, and non-rigid registration is performed within the point. According to the second relative pose, further registration is performed to obtain the global pose, which can further reduce the global point cloud registration error and improve the accuracy of the final registration effect of the point cloud.

[0038] The house point cloud registration method of this embodiment can improve the applicability of the house point cloud registration method and the accuracy of the registration result by selecting the first optimization target and the degree of freedom registration method; reduce the coarse registration pose accuracy requirement by constructing a pyramid; and achieve accurate registration of the point cloud by splitting the point cloud points into multiple point cloud fragments and performing a second registration on the multiple point cloud fragments, thereby avoiding the problem of stratification of the wall and ground parts of the indoor house point cloud.

[0039] In some embodiments of the present application, determining a first optimization target based on the source house point cloud specifically includes:

[0040] Determine whether the texture of the source house point cloud is sparse. If the texture of the source house point cloud is sparse, determine that the first optimization target is the structural information of the source house point cloud; otherwise, determine that the first optimization target is the texture information and structural information of the source house point cloud.

[0041] It can be understood that when the distance between point clouds is greater than a set threshold, the point cloud texture can be considered to be sparse, otherwise, the point cloud texture can be considered to be rich.

[0042] A single optimization target cannot be applicable to the registration of all point clouds, so this embodiment sets different optimization targets for point clouds with different characteristics. First, it is determined whether the point cloud texture is sparse. If the point cloud texture is sparse, the structural information of the point cloud is used as the first optimization target for optimization. If the point cloud texture is rich but the structure is not obvious, the structure and texture of the point cloud are combined for optimization, and the first optimization target is determined to be the texture information and structural information of the source house point cloud, that is, optimization is performed based on the grayscale of the point cloud.

[0043] This embodiment adopts two different optimization targets according to the actual situation. Different optimization targets are suitable for different point clouds. When applied, they can cope with different usage scenarios and improve the applicability of the house point cloud registration method. In addition, by dynamically adjusting the optimization target, the method can be more adaptable to different reconstruction scenarios.

[0044] In this embodiment, the optimization target is selected according to the characteristics of the point cloud. For the optimization with geometric structure (structural information) as the optimization target, the loss function is defined as:

[0045]

[0046] Among them, E represents the energy sum, n represents the number of point clouds, i represents the point cloud index, and q i represents the coordinates of the i-th 3D point of the target point cloud, R represents the rotation matrix from the current point cloud to the target point cloud, and p i represents the coordinates of the i-th 3D point in the current point cloud, t represents the translation matrix from the current point cloud to the target point cloud, represents the normal vector of the i-th 3D point of the target point cloud, and ||·||2 represents the L2 norm operation.

[0047] For the optimization goal of combining geometric structure (structural information) and texture information, the loss function is:

[0048]

[0049] Among them, ω represents the weight coefficient, C(q i ) represents the color value of the i-th 3D point of the target point cloud, C(p i ) shows the color value of the i-th 3D point in the current point cloud.

[0050] In some embodiments of the present application, Figure 2 FIG. 2 shows a second flow chart of the house point cloud registration method provided in an embodiment of the present application. Figure 2 As shown, based on the source house point cloud, the degree of freedom registration method is determined, including:

[0051] Step 202, determining a method for acquiring the source house point cloud.

[0052] Step 204: When the acquisition method of the source house point cloud remains horizontal, determine to adopt a four-degree-of-freedom registration method.

[0053] Step 206: When the acquisition method of the source house point cloud remains horizontal and has no vertical displacement, determine to adopt a three-degree-of-freedom registration method.

[0054] Step 208: When the source house point cloud is obtained in other ways, determine to use a six-degree-of-freedom registration method.

[0055] In the point cloud registration link of this embodiment, the optimization degrees of freedom are limited according to specific circumstances. For example, when collecting house point clouds, if each point cloud is guaranteed to be basically non-inclined through means such as a pan / tilt or a tripod, only four degrees of freedom need to be optimized during the optimization process. On the basis of the above, if the height of each point cloud can be guaranteed to be basically consistent, only three degrees of freedom need to be optimized during the optimization process. Specifically, the registration degrees of freedom are evaluated according to different point cloud acquisition methods. For ordinary point clouds, a six-degree-of-freedom registration method is selected; for point clouds that ensure horizontality, a four-degree-of-freedom registration method is selected; for point clouds that ensure horizontality and consistent horizontal height, a three-degree-of-freedom registration method is selected.

[0056] It is understandable that in applications such as VR house viewing, the initial pose of the house point cloud obtained will be different due to the different ways of collecting point clouds. This embodiment uses optimization methods with different degrees of freedom for different collection characteristics. Six degrees of freedom optimization is used for ordinary point clouds, four degrees of freedom optimization is used for horizontal point clouds, and three degrees of freedom optimization is used for horizontal point clouds with no vertical displacement. Through targeted optimization methods, the robustness of the house point cloud registration method is improved.

[0057] In some embodiments of the present application, in the six-degree-of-freedom alignment method, the six degrees of freedom include the freedom of movement along the three rectangular coordinate axes of x, y, and z and the freedom of rotation along the three rectangular coordinate axes of x, y, and z.

[0058] It can be understood that six degrees of freedom means that an object has six degrees of freedom in space, namely, the freedom of movement along the three rectangular coordinate axes of x, y, and z and the freedom of rotation around these three coordinate axes. In this embodiment, for ordinary point clouds, a six-degree-of-freedom registration method is used to ensure the accuracy of the house point cloud registration method.

[0059] In some embodiments of the present application, in the four-degree-of-freedom alignment method, the four degrees of freedom include the freedom of movement along the three rectangular coordinate axes of x, y, and z in the x, y, and z rectangular coordinates and the degree of rotation freedom along the z rectangular coordinate axis, and the degrees of rotation freedom along the x and y rectangular coordinate axes are both zero.

[0060] It is understandable that when collecting house point clouds, if the collection is performed by means of a pan / tilt or a tripod, the point clouds will not be tilted. In view of the above situation, this embodiment only optimizes four degrees of freedom during the optimization process, that is, four degrees of freedom optimization is used for the horizontal point clouds. For different point cloud collection methods, targeted optimization methods are selected to improve the robustness of the house point cloud registration method, thereby improving the accuracy of the house point cloud registration method.

[0061] In some embodiments of the present application, in the three-degree-of-freedom alignment method, the three degrees of freedom include the degree of freedom of movement along the x and y rectangular coordinate axes and the degree of freedom of rotation along the z rectangular coordinate axis in the x, y, and z rectangular coordinates, and the degree of freedom of movement along the z rectangular coordinate axis and the degree of freedom of rotation along the x and y rectangular coordinate axes are both zero.

[0062] It is understandable that when collecting house point clouds, if the collection is performed by means of a pan / tilt or tripod, the point clouds will not be tilted, and the height of each point cloud is basically consistent. In view of the above situation, this embodiment only optimizes three degrees of freedom during the optimization process, that is, three degrees of freedom optimization is used for horizontal point clouds with consistent horizontal heights. For different point cloud collection methods, targeted optimization methods are selected to improve the robustness of the house point cloud registration method, thereby improving the accuracy of the house point cloud registration method.

[0063] In some embodiments of the present application, Figure 3 FIG. 3 shows a flow chart of the house point cloud registration method provided in the embodiment of the present application. Figure 3 As shown, based on the source house point cloud and the target house point cloud, a source point cloud pyramid and a target point cloud pyramid are constructed, specifically including:

[0064] Step 302 : setting the number of pyramid levels, and setting the corresponding voxel threshold based on the number of pyramid levels.

[0065] Step 304 , based on the voxel threshold, a voxel downsampling method is used to downsample the source house point cloud and the target house point cloud respectively, to obtain the point cloud in each layer of the pyramid, and then to obtain the source point cloud pyramid and the target point cloud pyramid.

[0066] In this embodiment, the number of pyramid layers is set according to the application scenario, and the corresponding threshold is set according to the number of pyramid layers. For example, the number of pyramid layers can be set to four layers, and the voxel threshold corresponding to each layer can be the first voxel value, the second voxel value, the third voxel value and the fourth voxel value.

[0067] In this embodiment, a voxel downsampling method is used to downsample the source house point cloud and the target house point cloud respectively to obtain the point cloud in each layer of the pyramid. The pyramid can be reasonably constructed through downsampling, and then multi-scale registration of the point cloud can be achieved, which can reduce the accuracy requirements of the registration method for the coarse registration pose, and can also improve the point cloud registration accuracy.

[0068] In some embodiments of the present application, the multiple voxel thresholds decrease sequentially with the number of pyramid levels.

[0069] In this embodiment, the number of voxel thresholds is the same as the number of pyramid levels, and the values ​​of the multiple voxel thresholds decrease in sequence as the number of pyramid levels increases. For example, the values ​​of the multiple voxel thresholds decrease exponentially as the number of pyramid levels increases.

[0070] For example, when constructing a point cloud pyramid, the number of pyramid layers is set to four, and the corresponding first voxel threshold is 0.16, the second voxel threshold is 0.08, the third voxel threshold is 0.04, and the fourth voxel threshold is 0.01. The units of the above voxel thresholds are all meters. For the point cloud P, downsampling is performed in sequence according to the voxel size of 0.16, 0.08, 0.04, and 0.01, and the point cloud pyramid P = {P1, P2, P3, P4} is obtained, where P1 represents the first layer, P2 represents the second layer, P3 represents the third layer, and P4 represents the fourth layer. The first layer of point cloud is the sparsest, and the fourth layer of point cloud is the densest. The number of point clouds in each layer is about 4 times that of the previous layer.

[0071] This embodiment can reasonably set the size of multiple voxel thresholds to reasonably set the number of point clouds in each layer of the pyramid, thereby achieving multi-scale registration of point clouds through the pyramid, and ultimately improving the accuracy of point cloud registration.

[0072] In some embodiments of the present application, Figure 4 FIG. 4 shows a flow chart of a method for aligning house point clouds provided in an embodiment of the present application. Figure 4 As shown, the voxel downsampling method specifically includes:

[0073] Step 402: Obtain a first value of a voxel.

[0074] Step 404, obtaining the number of point clouds and point cloud data in a first space, where the first space is a space whose length, width and height are all first values.

[0075] Step 406: based on the number of point clouds and the point cloud data, obtain the average value of the point cloud data.

[0076] Step 408 , setting the first space to include a point cloud, and the data of the point cloud is the average value of the point cloud data.

[0077] For example, a specific method of voxel sampling may be, for a downsampling of a voxel size a, counting all points p1, p2, p3, ..., p in each a×a×a three-dimensional space. n (p n represents the nth point), only one point is retained in the voxel According to the above method, the source point cloud pyramid P is constructed respectively. stc With the target point cloud pyramid P tgt .

[0078] In this embodiment, a specific implementation method of voxel downsampling is given. Through downsampling, source point cloud pyramids and target point cloud pyramids can be constructed, providing a prerequisite for subsequent multi-scale registration of point clouds through pyramids, thereby improving the accuracy of point cloud registration.

[0079] In some embodiments of the present application, based on the first relative pose, optimization target and degree of freedom registration method, multi-scale point cloud registration is performed on the source point cloud pyramid and the target point cloud pyramid to obtain a second relative pose of the source house point cloud and the target house point cloud, specifically including:

[0080] Set the first relative pose as the initial registration value, and use the registration result of the previous layer as the initial registration value of the next layer. According to the order of the pyramid, based on the optimization target and the degree of freedom registration method, align each layer of the source point cloud pyramid and the target point cloud pyramid. The final registration result is the second relative pose of the source house point cloud and the target house point cloud.

[0081] In this embodiment, starting from a low-resolution point cloud, a rough registration of the point cloud is first performed, and then the result of the rough registration is used as an initial value to further register on a higher-resolution point cloud, layer by layer. For example, the registration can be performed from the bottom layer to the top layer in the order of a pyramid, and the registration result of the previous layer is used as the initial value of the registration of the next layer.

[0082] In this embodiment, in each layer of the pyramid, the source point cloud and the target point cloud are registered according to the optimization target and degree of freedom registration method in the above embodiment, and the final registration result is the second relative pose.

[0083] The pyramid of this embodiment performs multi-scale registration, which can effectively reduce the accuracy requirements for coarse registration poses and effectively improve the point cloud registration accuracy.

[0084] In some embodiments of the present application, Figure 5 FIG. 5 shows a flow chart of a method for aligning house point clouds provided in an embodiment of the present application. Figure 5 As shown, based on the optimization goal and degree of freedom registration method, each layer of the source point cloud pyramid and the target point cloud pyramid is registered, including:

[0085] Step 502: Determine the gradient matrix according to the degree of freedom registration method.

[0086] Step 504, based on the initial registration value, gradient matrix and optimization target of each layer, solve the normal equations, and iterate to obtain the optimized posture, that is, the registration result of each layer.

[0087] In this embodiment, the gradient matrix is ​​determined according to the degree of freedom registration method, specifically:

[0088] For the six-degree-of-freedom registration method, the corresponding gradient matrix is:

[0089]

[0090] in, Represents the gradient matrix corresponding to the six-degree-of-freedom registration method, Indicates p i The normal vector, p i represents the i-th point, p' i Indicates p i The corresponding points are obtained after rotation and translation, T represents matrix transpose, and × represents vector cross product. The size of the gradient matrix corresponding to the six-degree-of-freedom registration method is 1×6, where the first dimension represents the variable gradient rotated around the x-axis, the second dimension represents the variable gradient rotated around the y-axis, the third dimension represents the variable gradient rotated around the z-axis, the fourth dimension represents the variable gradient translated along the x-axis, the fifth dimension represents the variable gradient translated along the y-axis, and the sixth dimension represents the variable gradient translated along the z-axis.

[0091] For the four-degree-of-freedom registration method, the corresponding gradient matrix is: the first dimension and the third dimension in the corresponding matrix of the six-degree-of-freedom registration method are set to zero.

[0092] For the three-degree-of-freedom registration method, the corresponding gradient matrix is: the first dimension, the third dimension and the fifth dimension in the corresponding matrix of the six-degree-of-freedom registration method are set to zero.

[0093] After determining the gradient matrix, solve the normal equation according to the gradient matrix:

[0094]

[0095] Among them, J i represents the gradient matrix, δ represents the update amount of Lie algebra, ξ represents the Lie algebra representation of the rotation and translation, e i represents the error between the source point and the target point, and T represents the matrix transpose.

[0096] Solve the normal equations iteratively to obtain the optimized pose.

[0097] The optimization target is the loss function given in the above embodiment.

[0098] In this embodiment, a method for obtaining a gradient matrix is ​​given. The registration result of each layer of the pyramid is obtained through the gradient matrix and the loss function, thereby providing an initial registration value for the next layer of the pyramid, and ultimately improving the accuracy of the point cloud registration method.

[0099] In some embodiments of the present application, Figure 6 FIG. 6 shows a flow chart of the house point cloud registration method provided in the embodiment of the present application. Figure 6 As shown, a second registration is performed on multiple point cloud fragments to obtain a global pose, which specifically includes:

[0100] Step 602 , divide the multiple point cloud fragments into multiple point cloud fragment sub-graphs, perform pairwise registration on the point cloud fragments in the point cloud fragment sub-graphs, and obtain the relative pose of every two point cloud fragments.

[0101] Step 604, construct a global pose graph, the nodes of the global pose graph are the pose update increments of the point cloud fragments, initialized to the unit matrix, and set the relative pose of every two point cloud fragments to the constraint edge of the corresponding two point cloud fragment nodes.

[0102] Step 606, based on the second optimization goal, the global pose graph is optimized to obtain the optimized point cloud pose of each point cloud segment node.

[0103] Step 608, obtaining a global pose based on the optimized point cloud pose.

[0104] After the point cloud of the house is registered between multiple points, the relative poses between multiple points in the whole house are obtained. Furthermore, this embodiment refines and splits the point cloud of each point into multiple point cloud fragments. According to the second relative pose, the point cloud fragment submap is divided, and the fragments in the submap are further registered, and then the global pose graph optimization problem is constructed according to the fine registration result, so as to further reduce the global point cloud registration error.

[0105] For example, these include:

[0106] (1) After obtaining the second relative pose of the point cloud between multiple points through the above point cloud registration method, the point cloud of the point is split into multiple point cloud segments to perform refined registration correction of the point cloud within the point.

[0107] (2) Calculate the point cloud centroid of each point cloud fragment and use the point cloud centroid to construct a KDtree (k-dimensional tree). For each point cloud, use the KNN algorithm (K-NearestNeighbor) on the kdtree to search for point cloud fragments within the specified distance threshold, establish matching relationships between sub-map fragments, and prepare for further fine alignment.

[0108] (3) Traverse the matching relationships of each sub-image constructed in (2) above, and perform pairwise registration on the fragment point clouds within the sub-image.

[0109] (4) Construct a global pose graph, where the nodes are the pose update increments T of each fragment point cloud. i node , initialized to the identity matrix. Add the relative pose calculated in (3) as the constraint edge between the nodes of the corresponding two fragment numbers:

[0110]

[0111] Among them, T i node represents the pose update of the i-th point cloud fragment, R represents the relative rotation matrix between the two fragments, and t represents the relative translation matrix between the two fragments.

[0112] (5) Optimize the global pose graph, average the constraints errors, remove the wrong constraints, and update the pose increments corresponding to each node. The optimization goal is:

[0113]

[0114] Where E(T,L) represents the optimization objective, T represents the relative rotation and translation matrix between all two nodes, L represents the relative constraint confidence between all two nodes, i represents the index of the i-th node, j represents the index of the j-th node, and T i represents the rotation and translation matrix of the ith node in the global coordinate system, T j represents the rotation and translation matrix of the jth node in the global coordinate system, T ij represents the relative rotation and translation matrix between nodes i and j, μ represents the weight coefficient, l ij represents the constraint confidence, which can be obtained by calculating the covariance matrix of the registered matching point pairs, l ij ∈(0,1).

[0115]

[0116] Among them, γ(l ij ) represents the confidence mapping function.

[0117] (6) Use the optimized values ​​of each node to update the point cloud fragment pose as the final global pose:

[0118] T i frag' =T i node ·T i frag

[0119] Among them, T i frag' represents the updated point cloud fragment pose, T i node represents the point cloud fragment pose update amount, T i frag Represents the point cloud fragment pose before updating.

[0120] In some embodiments of the present application, Figure 7 One of the schematic diagrams of the housing point cloud registration method provided in the embodiment of the present application is shown. Figure 8 The second schematic diagram of the housing point cloud registration method provided in the embodiment of the present application is shown. Figure 7 and Figure 8 As shown, the house point cloud registration method includes:

[0121] Part I:

[0122] (1) Based on the source house point cloud 702 and the target house point cloud 704, the point cloud acquisition method, texture and structural characteristics are evaluated, and the point cloud optimization degrees of freedom and optimization target 706 are evaluated (i.e., the first optimization target and the degree of freedom registration method are determined).

[0123] (2) Using a coarse registration method, the source house point cloud and the target house point cloud are preliminarily registered (i.e., point cloud coarse registration 708) to obtain their approximate relative pose, i.e., the first relative pose.

[0124] (3) Perform voxel downsampling on the source point cloud and the target point cloud, construct a source point cloud pyramid 710, and construct a target point cloud pyramid 712.

[0125] (4) Using the coarse registration pose as the initial value, multi-scale pose estimation (i.e., multi-scale point cloud registration 714) is performed. Starting from the first layer of the pyramid, the pose is iteratively updated from coarse to fine, and the estimation result of each layer is used as the initial value of the estimation of the next layer of the pyramid. Each layer is iteratively optimized according to the optimization objective and optimization degrees of freedom evaluated in (1).

[0126] (5) Based on the result of the multi-scale estimation, the relative pose of the point cloud between the multiple points is output (output point pose 716), that is, the second relative pose.

[0127] in, Figure 7 In the above, (1) to (5) correspond to the above (1) to (5) respectively.

[0128] Part II:

[0129] (6) Divide the single-point point cloud into multiple point cloud segments 802.

[0130] (7) Divide the point cloud fragment sub-graph 804 according to the position relationship.

[0131] (8) The point cloud segments within the sub-image are aligned pairwise 806.

[0132] The point cloud registration algorithm of the first part (1) to (5) is used to perform pairwise registration on the segment point clouds in the sub-image.

[0133] (9) Add the registration result as a constraint to the global pose graph 808.

[0134] (10) Perform global optimization 810.

[0135] (11) Update the pose 812.

[0136] Through the above process, the updated pose is obtained and the house point cloud registration is completed.

[0137] The house point cloud registration method provided in the embodiment of the present application can be executed by a house point cloud registration device. In the embodiment of the present application, the house point cloud registration method is executed by a house point cloud registration device as an example to illustrate the house point cloud registration device provided in the embodiment of the present application.

[0138] In some embodiments of the present application, a house point cloud registration device is provided. Fig. 9 The structure block diagram of the house point cloud registration device provided in the embodiment of the present application is shown as follows: Fig. 9As shown, the house point cloud registration device 100 includes a first determination module 110, a second determination module 120, a first acquisition module 130, a first construction module 140, a second acquisition module 150 and a third acquisition module 160. The first determination module 110 is used to determine a first optimization target based on the source house point cloud. The second determination module 120 is used to determine a degree of freedom registration method based on the source house point cloud. The first acquisition module 130 is used to perform a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative pose of the source house point cloud and the target house point cloud. The first construction module 140 is used to construct a source point cloud pyramid and a target point cloud pyramid based on the source house point cloud and the target house point cloud. The second acquisition module 150 is used to perform multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid based on the first relative pose, the optimization target and the degree of freedom registration method to obtain a second relative pose of the source house point cloud and the target house point cloud. The third acquisition module 160 is used to split the point cloud points into multiple point cloud segments based on the second relative pose, perform a second registration on the multiple point cloud segments, obtain the global pose, and complete the house point cloud registration.

[0139] The house point cloud registration method of this embodiment can improve the applicability of the house point cloud registration method and the accuracy of the registration result by selecting the first optimization target and the degree of freedom registration method; reduce the coarse registration pose accuracy requirement by constructing a pyramid; and achieve accurate registration of the point cloud by splitting the point cloud points into multiple point cloud fragments and performing a second registration on the multiple point cloud fragments, thereby avoiding the problem of stratification of the wall and ground parts of the indoor house point cloud.

[0140] The house point cloud registration device 100 provided in the embodiment of the present application can implement each process of the above-mentioned point cloud model detection method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0141] The house point cloud registration device in the embodiment of the present application can be an electronic device or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0142] The house point cloud registration device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0143] The house point cloud registration device provided in the embodiment of the present application can implement each process implemented in the above method embodiment, and will not be described again here to avoid repetition.

[0144] Alternatively, if Fig.10 As shown, an embodiment of the present application further provides an electronic device 1000, which includes a processor 1002 and a memory 1004. The memory 1004 stores programs or instructions that can be executed on the processor 1002. When the program or instructions are executed by the processor 1002, the various steps of the above-mentioned method embodiment are implemented and the same technical effect can be achieved. To avoid repetition, they are not described here.

[0145] It should be noted that the electronic devices in the embodiments of the present application include the above-mentioned mobile electronic devices and non-mobile electronic devices.

[0146] Fig.11 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of the present application.

[0147] The electronic device 1100 includes but is not limited to: a radio frequency unit 1101, a network module 1102, an audio output unit 1103, an input unit 1104, a sensor 1105, a display unit 1106, a user input unit 1107, an interface unit 1108, a memory 1109, and a processor 1110 and other components.

[0148] Those skilled in the art will appreciate that the electronic device 1100 may also include a power source (such as a battery) for supplying power to each component, and the power source may be logically connected to the processor 1110 through a power management system, thereby implementing functions such as managing charging, discharging, and power consumption management through the power management system. Fig.11 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be described in detail here.

[0149] The processor 1110 is used to determine a first optimization target based on the source house point cloud.

[0150] The processor 1110 is used to determine a degree of freedom registration method based on the source house point cloud.

[0151] The processor 1110 is used to perform a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative position and posture of the source house point cloud and the target house point cloud.

[0152] Processor 1110 is used to construct a source point cloud pyramid and a target point cloud pyramid based on the source house point cloud and the target house point cloud.

[0153] Processor 1110 is used to perform multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid based on the first relative pose, optimization target and degree of freedom registration method to obtain a second relative pose of the source house point cloud and the target house point cloud.

[0154] Processor 1110 is used to split the point cloud points into multiple point cloud segments based on the second relative pose, perform a second registration on the multiple point cloud segments, obtain a global pose, and complete the house point cloud registration.

[0155] The house point cloud registration method of this embodiment can improve the applicability of the house point cloud registration method and the accuracy of the registration result by selecting the first optimization target and the degree of freedom registration method; reduce the coarse registration pose accuracy requirement by constructing a pyramid; and achieve accurate registration of the point cloud by splitting the point cloud points into multiple point cloud fragments and performing a second registration on the multiple point cloud fragments, thereby avoiding the problem of stratification of the wall and ground parts of the indoor house point cloud.

[0156] The processor 1110 provided in the embodiment of the present application can implement each process of the above-mentioned point cloud model detection method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0157] It should be understood that in the embodiment of the present application, the input unit 1104 may include a graphics processor (Graphics Processing Unit, GPU) 11041 and a microphone 11042, and the graphics processor 11041 processes the image data of the static picture or video obtained by the image capture device (such as a camera) in the video capture mode or the image capture mode. The display unit 1106 may include a display panel 11061, and the display panel 11061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 1107 includes a touch panel 11071 and at least one of other input devices 11072. The touch panel 11071 is also called a touch screen. The touch panel 11071 may include two parts: a touch detection device and a touch controller. Other input devices 11072 may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.

[0158] The memory 1109 can be used to store software programs and various data. The memory 1109 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory 1109 may include a volatile memory or a non-volatile memory, or the memory 1109 may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM). The memory 1109 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

[0159] The processor 1110 may include one or more processing units; optionally, the processor 1110 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to an operating system, a user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor 1110.

[0160] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned house point cloud registration method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0161] The processor is the processor in the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0162] An embodiment of the present application further provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned house point cloud registration method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0163] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0164] An embodiment of the present application provides a computer program product, which is stored in a storage medium. The program product is executed by at least one processor to implement the various processes of the above-mentioned house point cloud registration method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0165] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0166] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0167] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A house point cloud registration method, characterized in that: Used to register the source house point cloud with the target house point cloud, the house point cloud registration method includes: Based on the source house point cloud, determining a first optimization target; Determining a degree of freedom registration method based on the source house point cloud; Performing a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative position and posture of the source house point cloud and the target house point cloud; Based on the source house point cloud and the target house point cloud, construct a source point cloud pyramid and a target point cloud pyramid; Based on the first relative pose, the optimization target and the degree of freedom registration method, performing multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid to obtain a second relative pose of the source house point cloud and the target house point cloud; Based on the second relative pose, split the point cloud points into multiple point cloud segments, perform a second registration on the multiple point cloud segments, obtain a global pose, and complete the house point cloud registration; The determining of a first optimization target based on the source house point cloud specifically includes: Determine whether the texture of the source house point cloud is sparse, and if the texture of the source house point cloud is sparse, determine that the first optimization target is the structural information of the source house point cloud; otherwise, determine that the first optimization target is the texture information and structural information of the source house point cloud; The determining of the degree of freedom registration method based on the source house point cloud specifically includes: Determine a method for acquiring the source house point cloud; Under the condition that the acquisition method of the source house point cloud remains horizontal, determining to adopt a four-degree-of-freedom registration method; Under the condition that the acquisition method of the source house point cloud remains horizontal and has no vertical displacement, determining to adopt a three-degree-of-freedom registration method; In the case where the source house point cloud is obtained in other ways, it is determined to adopt a six-degree-of-freedom registration method; The step of performing multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid based on the first relative pose, the optimization target and the degree of freedom registration method to obtain a second relative pose of the source house point cloud and the target house point cloud specifically includes: The first relative pose is set as the initial registration value, the registration result of the previous layer is used as the initial registration value of the next layer, and each layer of the source point cloud pyramid and the target point cloud pyramid is registered according to the order of the pyramid based on the optimization target and the degree of freedom registration method, so as to obtain a final registration result, which is the second relative pose of the source house point cloud and the target house point cloud; The performing a second registration on the plurality of point cloud fragments to obtain a global pose specifically includes: Dividing the plurality of point cloud fragments into a plurality of point cloud fragment sub-graphs, performing pairwise registration on the point cloud fragments in the point cloud fragment sub-graphs, and obtaining a relative pose of every two point cloud fragments; Constructing a global pose graph, wherein the nodes of the global pose graph are the pose update increments of the point cloud fragments, are initialized to the unit matrix, and the relative pose of each two point cloud fragments is set as the constraint edge of the corresponding two point cloud fragment nodes; Based on the second optimization goal, the global pose graph is optimized to obtain the optimized point cloud pose of each point cloud fragment node; Based on the optimized point cloud pose, a global pose is obtained.

2. The house point cloud registration method according to claim 1, characterized in that: In the six-degree-of-freedom registration method, the six degrees of freedom include the degrees of freedom of movement along the three rectangular coordinate axes of x, y, z and the degrees of freedom of rotation along the three rectangular coordinate axes of x, y, z in the x, y, z rectangular coordinates.

3. The house point cloud registration method according to claim 1, characterized in that: In the four-degree-of-freedom registration method, the four degrees of freedom include the movement freedom along the three rectangular coordinate axes of x, y, and z and the rotation freedom along the z rectangular coordinate axis in the x, y, and z rectangular coordinates, and the rotation freedom along the x and y rectangular coordinate axes are both zero.

4. The house point cloud registration method according to claim 1, characterized in that: In the three-degree-of-freedom alignment method, the three degrees of freedom include the degree of freedom of movement along the x and y rectangular coordinate axes and the degree of freedom of rotation along the z rectangular coordinate axis in the x, y, and z rectangular coordinates. The degree of freedom of movement along the z rectangular coordinate axis and the degree of freedom of rotation along the x and y rectangular coordinate axes are both zero.

5. The house point cloud registration method according to claim 1, characterized in that: The constructing a source point cloud pyramid and a target point cloud pyramid based on the source house point cloud and the target house point cloud specifically includes: Setting the number of pyramid levels, and setting a corresponding voxel threshold based on the number of pyramid levels; Based on the voxel threshold, a voxel downsampling method is adopted to downsample the source house point cloud and the target house point cloud respectively, to obtain the point cloud in each layer of the pyramid, and then to obtain the source point cloud pyramid and the target point cloud pyramid.

6. The house point cloud registration method according to claim 5, characterized in that: The multiple voxel thresholds decrease sequentially with the number of pyramid levels.

7. The house point cloud registration method according to claim 5, characterized in that: The voxel downsampling method specifically comprises: Get the first value of the voxel; Acquire the number of point clouds and point cloud data in a first space, where the first space is a space whose length, width and height are all the first values; Based on the number of point clouds and the point cloud data, obtaining an average value of the point cloud data; It is assumed that the first space includes a point cloud, and the data of the point cloud is an average value of the point cloud data.

8. The house point cloud registration method according to claim 1, characterized in that: The registering each layer of the source point cloud pyramid with the target point cloud pyramid based on the optimization target and the degree of freedom registration method specifically includes: Determining a gradient matrix according to the degree of freedom registration method; Based on the initial registration value of each layer, the gradient matrix and the optimization target, the normal equations are solved and the optimized posture, that is, the registration result of each layer, is obtained by iteration.

9. A house point cloud registration device, characterized in that: Used to align the source house point cloud with the target house point cloud, the house point cloud alignment, the house point cloud alignment device comprises: A first determination module, configured to determine a first optimization target based on the source house point cloud; A second determination module is used to determine a degree of freedom registration method based on the source house point cloud; A first acquisition module is used to perform a first point cloud registration on the source house point cloud and the target house point cloud to obtain a first relative position and posture of the source house point cloud and the target house point cloud; A first construction module is used to construct a source point cloud pyramid and a target point cloud pyramid based on the source house point cloud and the target house point cloud; A second acquisition module is used to perform multi-scale point cloud registration on the source point cloud pyramid and the target point cloud pyramid based on the first relative pose, the optimization target and the degree of freedom registration method to obtain a second relative pose of the source house point cloud and the target house point cloud; A third acquisition module, based on the second relative pose, splits the point cloud points into multiple point cloud segments, performs a second registration on the multiple point cloud segments, obtains a global pose, and completes the house point cloud registration; The first determination module is specifically used to determine whether the texture of the source house point cloud is sparse. If the texture of the source house point cloud is sparse, the first optimization target is determined to be the structural information of the source house point cloud; otherwise, the first optimization target is determined to be the texture information and structural information of the source house point cloud; The second determination module is specifically used to determine the acquisition method of the source house point cloud; when the acquisition method of the source house point cloud remains horizontal, determine to adopt a four-degree-of-freedom registration method; when the acquisition method of the source house point cloud remains horizontal and there is no vertical displacement, determine to adopt a three-degree-of-freedom registration method; when the acquisition method of the source house point cloud is other, determine to adopt a six-degree-of-freedom registration method; The second acquisition module is specifically used to set the first relative pose as the initial registration value, use the registration result of the previous layer as the initial registration value of the next layer, and register each layer of the source point cloud pyramid and the target point cloud pyramid in the order of the pyramid based on the optimization target and the degree of freedom registration method to obtain a final registration result, which is the second relative pose of the source house point cloud and the target house point cloud; The third acquisition module is specifically used to divide the multiple point cloud fragments into multiple point cloud fragment subgraphs, perform pairwise registration on the point cloud fragments in the point cloud fragment subgraphs, and obtain the relative pose of every two point cloud fragments; construct a global pose graph, the nodes of the global pose graph are the pose update increments of the point cloud fragments, initialized as the unit matrix, and set the relative pose of every two point cloud fragments to the constraint edges of the corresponding two point cloud fragment nodes; based on the second optimization goal, optimize the global pose graph to obtain the optimized point cloud pose of each point cloud fragment node; based on the optimized point cloud pose, obtain the global pose.

10. An electronic device, characterized in that: include: A memory on which programs or instructions are stored; A processor, used to implement the steps of the house point cloud registration method as described in any one of claims 1 to 8 when executing the program or instruction.

11. A readable storage medium having a program or instruction stored thereon, characterized in that: When the program or instruction is executed by a processor, the steps of the house point cloud registration method according to any one of claims 1 to 8 are implemented.

Images