A hierarchical bounding box-based video image fast cropping and dimension reduction method and system

Abstract

The application provides a video image fast clipping and dimension reduction method and system based on a hierarchical bounding box, applied to the technical field of image processing, wherein the method collects multi-view dynamic scene point cloud data, and then divides the space by using a hierarchical bounding box to establish a bounding box structure with density characteristics; subsequently, optical diffraction array pre-filtering is used, and point clouds are separated based on wavelength selectivity and density distribution to generate a first subset and a background subset; then, adaptive pulse code modulation is used on nodes in the first subset to obtain high-precision data, and nodes in the second subset are compressed into low-code-rate data by using run-length encoding; subsequently, the two types of data are input into an improved k-d tree based on the spatial relationship of the bounding box to generate a two-dimensional point set representing key features, and a spatial index is constructed to realize real-time scene retrieval. The application can realize image dimension reduction while retaining key geometric features of the image, thereby improving efficient real-time spatial retrieval in a dynamic scene.

Description

TECHNICAL FIELD

[0001] The application relates to the field of image dimension reduction, in particular to a video image rapid clipping and dimension reduction method and system based on hierarchical bounding boxes. BACKGROUND

[0002] In dynamic three-dimensional scenes such as automatic driving real-time perception and immersive augmented reality interaction, the problem of image dimension reduction needs to be solved. To solve this problem, gigabyte-level three-dimensional point clouds need to be efficiently compressed into megabyte-level two-dimensional feature images that can be processed in real time while preserving key information such as object edge sharpness and surface texture. At the same time, the whole link response from data acquisition to spatial retrieval needs to be completed within a short time to support high-precision real-time decision-making on mobile platforms with limited resources.

[0003] The current mainstream solution uses a hierarchical feature distillation architecture. This architecture extracts spatiotemporal features of point clouds through a three-dimensional convolutional neural network, filters high-value three-dimensional points using a channel attention mechanism, projects the filtered point clouds onto a two-dimensional feature plane through graph pooling operations, and finally constructs a spatial index tree to realize fast retrieval of dimension-reduced images.

[0004] However, the above architecture is difficult to adapt to dynamic scene deformation due to fixed convolution kernels, resulting in the breaking and distortion of moving object edges during dimension reduction. In addition, the cumulative error caused by the phased feature distillation and two-dimensional projection makes the final two-dimensional image lose curvature details. Furthermore, the graph pooling operation destroys the original spatial topology, causing structural misalignment between the dimension-reduced image and the three-dimensional scene. SUMMARY

[0005] The application provides a video image rapid clipping and dimension reduction method and system based on hierarchical bounding boxes to solve the problem of poor image dimension reduction effect in the prior art.

[0006] In a first aspect, the application provides a video image rapid clipping and dimension reduction method based on hierarchical bounding boxes, comprising:

[0007] Collecting point cloud data of the image to be processed in a multi-view dynamic scene, dividing the point cloud data in space using hierarchical bounding boxes to obtain a bounding box hierarchy, and the bounding box hierarchy is composed of multiple bounding box nodes;

[0008] Pre-filtering the multi-view dynamic scene through a diffraction element array to divide the bounding box nodes in the bounding box hierarchy into a first subset and a second subset based on the wavelength selectivity and the distribution characteristics of the point cloud density;

[0009] The bounding box nodes in the first subset are processed by using adaptive pulse code modulation to dynamically allocate the number of coding bits according to the point cloud curvature change rate, to obtain coding data, and the bounding box nodes in the second subset are compressed by using run-length encoding, to obtain compressed data;

[0010] Based on the spatial topological relationship of the bounding box hierarchical structure, the coding data and the compressed data are input into an improved k-d tree for non-uniform clustering processing, to obtain a two-dimensional feature point set;

[0011] According to the spatial distribution of the two-dimensional feature point set, a spatial index structure is constructed to realize real-time spatial retrieval of a multi-view dynamic scene.

[0012] Optionally, the processing of the bounding box nodes in the first subset by using adaptive pulse code modulation to dynamically allocate the number of coding bits according to the point cloud curvature change rate, to obtain coding data, and the processing of the bounding box nodes in the second subset by using run-length encoding to obtain compressed data, comprises:

[0013] Each point in the bounding box nodes in the first subset is traversed, the curvature change rate of the points in the local neighborhood is calculated, and the points in the local neighborhood are divided into one of a plurality of predefined change rate levels according to the curvature change rate;

[0014] The three-dimensional coordinate values of the bounding box nodes in the first subset are obtained, the three-dimensional coordinate values are scaled and converted into binary codes of a specified number of bits according to the allocated number of coding bits, to form coding data;

[0015] The three-dimensional coordinate values of the bounding box nodes in the second subset are obtained, and a preset spatial scanning order is used for scanning, to identify points with the same three-dimensional coordinate values that appear continuously as a continuous sequence, and to represent the combination of the continuous sequence as compressed data.

[0016] In a second aspect, the present application provides a video image fast cropping and dimensionality reduction system based on hierarchical bounding boxes, comprising:

[0017] A collection module is configured to collect point cloud data of a to-be-processed image in a multi-view dynamic scene, and divide the point cloud data in space by using a hierarchical bounding box, to obtain a bounding box hierarchical structure, wherein the bounding box hierarchical structure is composed of a plurality of bounding box nodes;

[0018] A separation module is configured to pre-filter the multi-view dynamic scene by using a diffraction element array, to divide the bounding box nodes in the bounding box hierarchical structure into a first subset and a second subset based on the wavelength selectivity and the distribution characteristics of the point cloud density;

[0019] The first processing module is configured to process the bounding box nodes in the first subset by using adaptive pulse code modulation to dynamically allocate coding bits according to a point cloud curvature change rate, to obtain coding data, and to compress the bounding box nodes in the second subset by using run-length encoding, to obtain compressed data.

[0020] The second processing module is configured to input the coding data and the compressed data into an improved k-d tree for non-uniform clustering processing based on a spatial topological relationship of the bounding box hierarchy, to obtain a two-dimensional feature point set.

[0021] The construction module is configured to construct a spatial index structure according to a spatial distribution of the two-dimensional feature point set, to realize real-time spatial retrieval of a multi-view dynamic scene.

[0022] In a third aspect, the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component, to realize the method for fast video image cropping and dimension reduction based on hierarchical bounding boxes according to the first aspect.

[0023] In a fourth aspect, the present application provides a computer storage medium, which stores a computer program; when the computer program is executed by a computer, the method for fast video image cropping and dimension reduction based on hierarchical bounding boxes according to the first aspect is realized.

[0024] The method for fast video image cropping and dimension reduction based on hierarchical bounding boxes provided in the embodiments of the present application has the following beneficial effects:

[0025] By collecting multi-view dynamic scene point cloud data and obtaining a hierarchical structure through hierarchical bounding box division, the present application can realize efficient spatial organization of point cloud data, and further accurately retain the geometric features and point cloud density distribution of different regions, to further provide a structured basis for subsequent data separation and processing; meanwhile, the first subset and the second subset are separated by using a diffraction element array pre-filter, which can accurately distinguish key geometric information and redundant second based on wavelength selectivity and point cloud density characteristics, to further reduce the interference of invalid data on subsequent processing, and further improve the data processing pertinence;

[0026] Then, by processing the first subset by using adaptive pulse code modulation and processing the second subset by using run-length encoding, coding resources can be dynamically allocated according to point cloud curvature, and efficient compression of the second subset can be realized while ensuring high precision of geometric details, to further balance data accuracy and storage / transmission efficiency.

[0027] Then, the encoded data is input into the improved k-d tree based on the bounding box hierarchy topology relationship to obtain a two-dimensional feature point set through non-uniform clustering, which can accurately retain key geometric features in the dimension reduction process, and then realize effective conversion from three dimensions to two dimensions, and further provide simplified and key feature data for subsequent retrieval; finally, a spatial index structure is constructed according to the two-dimensional feature point set, which can realize fast spatial retrieval of multi-view dynamic scenes, and then greatly improve the retrieval efficiency and meet the response requirements of real-time interactive scenes.

[0028] Further, the application can ensure high-precision encoding of the first subset by dynamically allocating the number of encoding bits according to the curvature change rate, and then accurately retain key features such as object edges; then the second subset is compressed by using run-length encoding, which can effectively reduce the code rate of the second subset, thereby realizing efficient compression of data, greatly optimizing the data amount while ensuring the accuracy of key geometric features, and improving the processing and transmission efficiency.

[0029] These aspects or other aspects of the application will be more apparent in the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0031] Figure 1 A flow chart of a video image fast clipping and dimension reduction method based on hierarchical bounding box provided by the application is shown;

[0032] Figure 2 A flow chart of generating the bounding box nodes in the first subset and the bounding box nodes in the second subset provided by the application is shown;

[0033] Figure 3 A structural schematic diagram of a video image fast clipping and dimension reduction system based on hierarchical bounding box provided by the application is shown;

[0034] Figure 4 A structural schematic diagram of a computing device provided by the application is shown. DETAILED DESCRIPTION

[0035] In order to make the person skilled in the art better understand the application scheme, the technical solutions in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application.

[0036] In some of the flowcharts described in the specification and claims of the present application and in the above description of the drawings, a plurality of operations are included which occur in a particular order, but it should be clearly understood that the operations can be performed in an order other than that in which they appear herein or in parallel, and the serial numbers of the operations, such as 101, 102, etc., are merely used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, the flowcharts can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions herein, such as "first", "second", etc., are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit the "first" and "second" to be different types.

[0037] Research shows that in dynamic three-dimensional scenes such as real-time perception of autonomous driving and immersive augmented reality interaction, the image dimensionality reduction problem of multi-view point cloud data faces severe challenges. Although the current mainstream hierarchical feature distillation architecture can achieve dimensionality reduction, it has obvious shortcomings, such as: fixed convolution kernels are difficult to adapt to the deformation of objects in dynamic scenes, resulting in the edges of moving objects being distorted and broken after dimensionality reduction; the cumulative error produced by the feature distillation and two-dimensional projection processes in stages causes the final two-dimensional image to lose the curvature details of the objects; and the graph pooling operation destroys the original spatial topological relationship, causing structural misalignment between the dimensionality reduced image and the three-dimensional scene, which seriously affects the accuracy and real-time performance of subsequent decision-making.

[0038] To solve the above problems, the present application proposes a video image fast cropping and dimensionality reduction method based on hierarchical bounding box. The core idea of this method is: first, collect point cloud data of a multi-view dynamic scene, and divide the space using hierarchical bounding boxes to obtain a hierarchical structure; then separate a first subset and a second subset using an array of diffraction elements for pre-filtering; next, use adaptive pulse code modulation for the first subset to retain high-precision information, and use run-length encoding for the second subset to achieve efficient compression; subsequently, based on the topological relationship of the bounding boxes, obtain a two-dimensional feature point set and construct a spatial index using an improved k-d tree clustering. The present application also solves the problem of fixed convolution kernels not adapting to deformation through dynamic adaptive hierarchical division, reduces cumulative error to retain curvature details through targeted subset separation and encoding processing, and avoids structural misalignment through clustering processing relying on the original topological relationship, effectively meeting the high-precision and real-time requirements of image dimensionality reduction in dynamic scenes.

[0039] The technical solutions in the embodiments of the present application will be described clearly and completely in the embodiments of the present application in conjunction with the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0040] Figure 1 A flowchart of a video image fast clipping and dimension reduction method based on hierarchical bounding box is provided for the embodiments of the present application, as shown in Figure 1 The method comprises the following steps:

[0041] 101. Collecting point cloud data of an image to be processed in a multi-view dynamic scene, and dividing the space of the point cloud data by using a hierarchical bounding box to obtain a bounding box hierarchy, wherein the bounding box hierarchy is composed of a plurality of bounding box nodes.

[0042] In one embodiment, the point cloud data refers to a set of three-dimensional space points of the image to be processed in a dynamic scene collected by a multi-view sensor, wherein each point in the set contains corresponding coordinate information; the bounding box hierarchy refers to a tree-shaped spatial index composed of multiple levels of cubic containers, and each container node or bounding box node or bounding box is used to record the distribution characteristics of internal points.

[0043] In one embodiment, step 101 comprises:

[0044] 1011. The point cloud data is taken as an initial bounding box node of the first level, and a minimum point threshold and a minimum space volume threshold are set for the initial bounding box node. The space volume of the initial bounding box node is calculated, and the total number of points contained in the initial bounding box node is counted.

[0045] In one embodiment, the initial bounding box node or root node bounding box node or root node bounding box is a minimum point threshold, which refers to a minimum point limit for stopping the division of a single bounding box node; and a minimum space volume threshold refers to a minimum physical size for stopping the division of a node.

[0046] In step 1011, the point cloud data of the image to be processed collected in the dynamic scene is taken as a root node bounding box, and the space division termination condition is set, such as a minimum point threshold of 120 points and a minimum space volume threshold of 0.8 m3. Then, the space volume of the initial bounding box node is calculated, such as 30 m x 25 m x 20 m = 15000 m3, and the total number of points in the interior is accurately counted, such as 128000 points.

[0047] 1012. When the total number of points is greater than the minimum point threshold and the space volume is greater than the minimum space volume threshold, cutting is performed along three coordinate axis directions, two symmetrical space division positions are generated in each coordinate axis direction, the center region volume ratio between the two symmetrical space division positions in the same coordinate axis direction is calculated, and the coordinate axis direction with the smallest center region volume ratio is selected as the division direction.

[0048] The spatial boundary surface refers to a virtual plane cutting the space along the coordinate axis direction, and the central region volume ratio refers to the volume proportion of the middle unallocated region after symmetrical cutting.

[0049] In step 1012, when the total number of points is greater than the minimum point threshold 120 points and the space volume is greater than the minimum space volume threshold 0.8 , the fixed cutting gap is calculated based on the bounding box level depth by the following formula , such as: , wherein, 0.01 m, and 0.05% represents the proportion coefficient of the cutting gap to the maximum axis length, Lmax represents the maximum length of the bounding box in the X, Y, and Z coordinate axis directions, so Lmax=max(Lx, Ly, Lz), Lx represents the length in the X axis direction, Ly represents the length in the Y axis direction, and Lz represents the length in the Z axis direction.

[0050] And, , wherein Xmax represents the maximum coordinate value of the bounding box in the X axis direction, and Xmin represents the minimum coordinate value of the bounding box in the X axis direction; similarly , wherein Ymax represents the maximum coordinate value of the Y axis, and Ymin represents the minimum coordinate value of the Y axis; , wherein Zmax represents the maximum coordinate value of the Z axis, and Zmin represents the minimum coordinate value of the Z axis.

[0051] For example, when Xmin, Ymin, and Zmin are 0, Lx=30.0−0.0=30.0 m, Ly=25.0−0.0=25.0 m, and Lz=20.0−0.0=20.0 m of the above bounding box, so Lmax=max(30.0, 25.0, 20.0)=30.0 m, and further m.

[0052] Further, the double cutting surfaces with a cutting gap δ generated at the geometric center position of each axis are calculated, that is, two symmetrical spatial boundary surface positions are generated in the X, Y, and Z coordinate axis directions, which are respectively:

[0053] The double cutting surface of the Z axis is: and , that is, the double cutting surface of the Z axis is 10.0 m and 10.01 m;

[0054] The double cutting surface of the Y axis is: and , that is, the double cutting surface of the Y axis is 12.5 m and 12.51 m;

[0055] The double cutting surface of the X axis is: and , that is, the double cutting surface of the X axis is 15 m and 15.01 m;

[0056] Then, the volume ratio of the central region between the two symmetrical space interfaces is calculated Rn= (the product of the length of the two axes) ÷ the total volume of the bounding box, where n is the subscript of R, and includes x, y, z three subscripts. That is:

[0057] Rx= (0.01 x 25 x 20) ÷ 15000 = 5 ÷ 15000 ≈ 0.00067 ≈ 0.03%;

[0058] Ry= (0.01 x 30 x 20) ÷ 15000 = 6 ÷ 15000 = 0.0004 = 0.04%;

[0059] Rz= (0.01 x 30 x 25) ÷ 15000 = 7.5 ÷ 15000 = 0.0005 = 0.05%;

[0060] By comparing the ratio values, Rmin = min (0.03%, 0.04%, 0.05%) = 0.03% = Rx, the X axis with the smallest volume ratio of the central unallocated region is finally selected as the cutting direction.

[0061] 1013、Based on the division direction, the region of the initial bounding box node is divided into two sub-regions, a lower level bounding box node is created for the sub-region, a hierarchical relationship link is established for the lower level bounding box node and the initial bounding box node, and is added to the processing queue.

[0062] In step 1013, based on the selected X axis cutting direction, the X axis range of the left child node is [Xmin, 15.0), that is, [0.0, 15.0) m, and the length is 15.0 m; the X axis range of the right child node is (15.01, Xmax], that is, (15.01, 30.0], and the length is 14.99 m, so the left range child node volume = 15.0 x 25.0 x 20.0 = 7500 , the right range child node volume = 14.99 x 25.0 x 20.0 = 7495 .

[0063] Further, the points contained in the left child node need to satisfy , the points contained in the right child node need to satisfy ; then the parent node 128000 point cloud data is executed to obtain the left range point number = 63200, the right range point number = 64800, and the sum of the two is 128000, which is consistent with the point number of the parent node; then the bidirectional pointer of the parent node pointing to the left and right child nodes is established, and the left child node volume 7500 > 0.8 ​and the point number 63200 > 120, the right child node volume 7495 > 0.8 and the point number 64800 > 120, both satisfy the recursive segmentation condition, and finally two child nodes are added to the processing queue.

[0064] 1014、When there is an unprocessed bounding box node in the processing queue, repeat the spatial division process until there is no unprocessed bounding box node in the processing queue, and output the bounding box hierarchy structure composed of all levels of bounding box nodes.

[0065] In step 1014, based on the spatial volume and the total number of points of the unprocessed bounding box node, for example, the left child node of the first item in the queue (63200 points, 7500 , because the point number and the volume both exceed the corresponding threshold, it is necessary to perform the cycle processing of steps 1012-1013; the process is similar to the steps of step 1013, and will not be repeated here.

[0066] For example, the recursive depth of 7 layers is performed, from the root node with a depth of 0 to the first segmentation with a depth of 1, the second segmentation with a depth of 2, and so on, until the seventh segmentation is terminated, and finally the bounding box tree structure is formed.

[0067] The present application can completely eliminate the problem of empty nodes generated by the traditional equal division method in a non-uniform scene by dynamically selecting the cutting direction based on the minimum center volume ratio strategy; combined with the double threshold termination mechanism of spatial volume and point cloud number, it can accurately match the feature distribution density of the actual scene; the generated bounding box hierarchy structure improves the spatial retrieval efficiency and establishes a high-precision spatial index framework for subsequent density feature analysis.

[0068] 102、Pre-filtering the multi-view dynamic scene through the diffraction element array to divide the bounding box nodes in the bounding box hierarchy structure into a first subset and a second subset based on the wavelength selectivity and the distribution characteristics of the point cloud density.

[0069] The diffraction element array refers to a physical device composed of a plurality of independent optical processing units arranged according to a preset angle difference, which separates incident light through wavelength selectivity, and each unit corresponds to a specific spectral analysis range of a certain viewing angle, wherein the first subset is a set of bounding box nodes carrying geometric contour details, and the second subset is a set of bounding box nodes carrying continuous background features.

[0070] In a specific embodiment, as shown in Figure 2 , step 102 includes:

[0071] 1021、receive the reflected light signal from the multi-view dynamic scene and input the reflected light signal into a diffraction element array, the diffraction element array is composed of a plurality of diffraction units arranged at a preset angle difference, and each diffraction unit separates the reflected light signal into a first light path and a second light path for independent transmission according to wavelength selectivity.

[0072] Among them, the reflected light signal refers to the light wave energy set reflected by the object surface to the collection device, including intensity, wavelength and other physical attribute data, which is used to represent the original optical characteristics of the scene; the diffraction unit refers to the smallest functional module in the array that performs spectral separation, which separates the mixed wavelength incident light into independent transmission channels according to the preset wavelength band; the first light path refers to the high-frequency reflected light carrying the edge and texture of the object; the second light path refers to the low-frequency diffuse reflected light carrying the smooth area.

[0073] In step 1021, input the multi-view light signal and receive the reflected light signal from the N cameras, wherein each signal contains wavelength λ and incident angle θ, and then input the light signal matrix L M×N Input the diffraction element array, wherein M is the number of spectral channels; then each diffraction unit modulates the incident light according to the preset tilt angle and outputs the separated light path that satisfies the Bragg condition, such as: λ1∈[λmin, λc], wherein λ1 represents the light wavelength of the first light path, λmin represents the minimum wavelength of the first light path, λc is the cutoff wavelength, λ2∈(λc, λmax], wherein λ2 represents the light wavelength of the second light path, λmax represents the maximum wavelength of the second light path, wherein the cutoff wavelength λc can be determined by the period of the diffraction unit, and the size of the cutoff wavelength is not limited in the embodiment of the application.

[0074] 1022, respectively set independent first processing end and second processing end for the first light path and the second light path, so as to receive the light intensity distribution of the first light path through the first processing end, and generate the first image data; receive the phase distribution of the second light path through the second processing end, and generate the second image data.

[0075] Among them, the light intensity distribution refers to the light energy intensity data per unit area, which reflects the mutation characteristics of the geometric profile of the object surface; the first image data carries the geometric profile details, and the second image data carries the continuous background features; the phase distribution refers to the spatial displacement data of the light wave waveform, which reflects the uniformity and continuity of the surface material.

[0076] In step 1022, the light intensity distribution I(x, y) of the first light path T1 is measured using the photoelectric sensor array at the first processing end, and the first image data of 1024x1024 pixels is output, which carries the geometric profile details; then, the phase distribution f(x, y) of the second light path T2 is analyzed by the Michelson interferometer at the second processing end, and the second image data of 1024x1024 is generated based on the phase distribution, which carries the continuous background features.

[0077] 1023, extract the distribution features of the point cloud density inside each bounding box node, the distribution features of the point cloud density include the density variation amplitude, mark the bounding box node whose density variation amplitude exceeds the preset density threshold as a first associated node; mark the bounding box node whose density variation amplitude does not exceed the preset density threshold as a second associated node.

[0078] Wherein, the distribution features of the point cloud density refer to the number distribution statistics of the spatial sampling units in the bounding box node.

[0079] In step 1023, the parent bounding box node with the size of Lx x Ly x Lz is divided into sub-nodes, and the calculation formula is: Wherein, r is the subdivision granularity, which represents the size reference of the sub-node in each direction; then, the density variation amplitude of each sub-node is calculated, that is, the standard deviation of the point cloud density in each sub-node: Wherein, represents the number of points in the kth sub-node, represents the local density of the jth point in the kth sub-node, represents the average local density of all points in the sub-node, and if the standard deviation of the point cloud density >τ d is marked as a first associated node, ≤τ d is marked as a second associated node, wherein τ d is a preset density variation threshold.

[0080] 1024, map the pixel cluster in the first image data whose luminance gradient exceeds the preset gradient threshold to the corresponding first associated node, map the pixel region in the second image data whose phase distribution is uniform to the corresponding second associated node, and determine the set to which each bounding box node belongs as the first subset or the second subset according to the mapping relationship.

[0081] Wherein, the luminance gradient refers to the luminance difference rate of adjacent pixels, which is used to identify edge mutations; the pixel cluster mapping refers to the operation of establishing coordinate association between the optical feature data and the spatial bounding box node; and the size of the preset gradient threshold is not specifically limited in the embodiments of the present application.

[0082] In step 1024, a gradient map corresponding to the first image data is determined based on the convolution kernel of the Sobel operator in the x and y directions, and then the gradient map ∇G is binarized based on a preset gradient threshold; then the binary image coordinates are mapped to a three-dimensional space based on the intrinsic matrix K of the camera, and node binding is performed to obtain the bounding box nodes in the first subset and the bounding box nodes in the second subset.

[0083] The application can accurately identify key geometric features and background areas through the dual verification mechanism of optical hardware level spectrum separation and spatial density statistics; then dynamically associates physical optical characteristics with point cloud spatial structure, which can eliminate high-frequency distortion and misjudgment risk caused by traditional digital filtering, and further establishes a high-reliability classification basis for subsequent differentiated processing.

[0084] 103, the bounding box nodes in the first subset are processed by adaptive pulse code modulation to dynamically allocate the number of coding bits according to the point cloud curvature change rate, to obtain coding data, and the bounding box nodes in the second subset are compressed by run-length encoding to obtain compressed data.

[0085] The curvature change rate refers to the change rate of the bending degree of the point cloud surface in the local neighborhood, which can be calculated by the coordinates of adjacent points, and is used to represent the complexity of geometric features.

[0086] Moreover, the specific implementation process of adaptive pulse code modulation and run-length encoding is not specifically limited in the embodiments of the application, and in a specific implementation manner, step 103 includes:

[0087] 1031, each point in the bounding box nodes in the first subset is traversed, the curvature change rate of the points in the local neighborhood is calculated, and the points in the local neighborhood are divided into one of a plurality of change rate levels according to the curvature change rate.

[0088] The plurality of change rate levels correspond to different coding bit numbers from high to low according to the curvature change rate, and the higher the change rate level, the more coding bits are allocated.

[0089] The change rate level refers to a predefined curvature change interval division, and different levels correspond to different coding accuracies; the coding bit number is the binary data length representing the three-dimensional coordinate value, and the higher the bit number, the higher the coordinate accuracy.

[0090] In step 1031, in the process of traversing each point cloud data in the first subset, a spherical detection area is first constructed with a fixed radius of 0.1 meters centered on the current point; then the three-dimensional coordinates of all adjacent points in the spherical detection area are counted, and the standard deviation of these three-dimensional coordinate values and the center point coordinate is calculated as the curvature change rate : , representing the coordinates of the neighboring points, is the average coordinate of the center point, and T is the number of neighboring points;

[0091] Subsequently, a first threshold of 0.02 meters is set based on the maximum deformation tolerance allowed by the steel welding quality standard, a second threshold of 0.01 meters is set based on the connection accuracy specification of the assembly, and the curvature change rate is compared with the preset three change rate level threshold intervals, wherein the high change rate level σ>0.02 meters corresponds to a weld and other complex surfaces, the medium change rate level 0.01 meters≤σ≤0.02 meters corresponds to a bolt and other moderate rough surfaces, and the low change rate level σ<0.01 meters corresponds to a flat steel plate and other smooth surfaces; then the number of coding bits is allocated according to the change rate level to obtain 8-bit coding for high change rate points, which can represent 256 states, 4-bit coding for medium change rate, which can represent 16 states, and 2-bit coding for low change rate, which can represent 4 states, so as to realize the dynamic compression strategy that the greater the curvature change, the higher the coding accuracy.

[0092] 1032. Obtain the three-dimensional coordinate values of the bounding box nodes in the first subset, scale and convert the three-dimensional coordinate values into binary codes of a specified number of bits according to the allocated number of coding bits to form coding data.

[0093] The step 1032 can specifically include the following processes:

[0094] Step a1: Obtain the three-dimensional coordinate values of the bounding box nodes in the first subset, wherein the three-dimensional coordinate values include three component values, calculate the maximum and minimum values of the three component values, and determine a scaling coefficient according to the correspondence between the maximum value, the minimum value and a preset integer range.

[0095] In step a1, all points in the current bounding box node are traversed to dynamically calculate the numerical boundaries of the three-dimensional coordinate axes, and the minimum coordinate value Xmin and the maximum coordinate value Xmax of the X axis are obtained. Similarly, the minimum coordinate value Ymin and the maximum coordinate value Ymax of the Y axis and the minimum coordinate value Zmin and the maximum coordinate value Zmax of the Z axis are calculated, and the coordinate span of each axis is calculated. and ΔZ are similarly calculated and will not be described here.

[0096] Step a2: Based on the scaling coefficient and the allocated number of coding bits, determine the specified number of bits corresponding to the binary code, convert the three component values into integer values within the preset integer range respectively, and convert the integer values into binary codes of the specified number of bits, aggregate the three binary codes to form a binary code.

[0097] The minimum value is zero, and the maximum value is the result of the power operation with two as the base number and the allocated number of coding bits as the exponent minus one.

[0098] In step a2, the coding bit number is assigned to the curvature rate level mark of each point, wherein 16 bits can be assigned to high rate points σ>0.02m, the integer range is 0-65, 12 bits can be assigned to medium rate points 0.01m≤σ≤0.02m, the integer range is 0-4, and 4 bits can be assigned to low rate points σ<0.01m, the integer range is 0-15; then the conversion between coordinates and integers is performed, and a linear mapping from floating-point coordinates to integer values is performed on each point, and the scaling formula is: wherein a∈{x, y, z}, represents the a-component coordinate value of the current point, is the minimum coordinate value representing the a-component, is the maximum coordinate value representing the a-component, and b represents the coding bit number corresponding to the curvature level of the current point, represents the converted a-component integer value, and the symbol represents the floor function.

[0099] Step a3: arranging the binary codes of all the bounding box nodes in the first subset in order to form the coding data.

[0100] In step a3, when generating the binary code, the integer value is converted into a binary sequence of a specified number of bits, and is padded with zeros to a fixed length in the high bits, and then the three-axis binary codes are concatenated in the order X→Y→Z to generate the complete point code, and finally all the point codes are concatenated in the order of the spatial index of the bounding box nodes to generate the coding data.

[0101] 1033. Obtain the three-dimensional coordinate values of the bounding box nodes in the second subset, and scan according to a preset spatial scanning order to identify points with the same three-dimensional coordinate values that appear continuously as a continuous sequence, and represent the combination of the continuous sequence as compression data.

[0102] wherein the spatial scanning order refers to a rule for traversing the point cloud according to a fixed path; the continuous sequence refers to a set of continuous points with the same coordinate values on the scanning path, and the combination representation of the continuous sequence includes: three-dimensional coordinate points and corresponding occurrence times.

[0103] In step 1033, all the points in the node are arranged in the order of the preset spatial scanning order, i.e., the order of ascending Z-axis, ascending Y-axis, and ascending X-axis; then the coordinate difference between adjacent points is compared in real time by the formula wherein represents the three-dimensional coordinates of the i-th point, represents the three-dimensional coordinates of the i-1-th point, represents the spatial distance between the two adjacent points, and when When the distance is less than the preset distance threshold, the same point is determined; then the continuous same points are merged into a single compression unit, and the storage structure includes the reference coordinate and the repetition count; finally, all the compression units are connected in the scanning order to output the compressed data.

[0104] The application can achieve precision enhancement in the key geometric feature area by dynamically allocating the number of encoding bits according to the curvature change rate; then, the spatial continuity detection is used to compress the second, which can significantly reduce the redundant storage, and the dual-mode processing mechanism can ensure the integrity of the core features while achieving efficient reduction of the overall data volume, thereby establishing the basis for real-time transmission and retrieval.

[0105] 104、Based on the spatial topological relationship of the bounding box hierarchy structure, the encoded data and the compressed data are input into the improved k-d tree for non-uniform clustering processing to obtain a two-dimensional feature point set.

[0106] In the application, the "improved k-d tree" is a self-adaptive spatial division structure specially designed for non-uniform clustering and dimension reduction of point clouds, and the improvement is reflected in the division strategy and termination condition: first, it inherits the basic form of recursive division of space of the traditional k-d tree, but improves the division basis, that is, for each region to be divided, the coordinate variances of the internal point cloud in the X, Y and Z coordinate axis directions are calculated, and the direction with the largest variance is selected as the division axis this time, to ensure that the division is along the most significant feature direction;

[0107] Secondly, a plurality of equally spaced candidate division positions are generated on the selected division axis, and the point number difference after each candidate position divides the current region into two sub-regions is calculated, and finally the position with the smallest point number difference is selected as the division plane, so as to realize the balance of the number of points in the sub-regions and adapt to the non-uniform distribution of point clouds;

[0108] Finally, its division termination condition is not a single point number threshold as in the traditional one, but a dual judgment combining the spatial volume and the number of point clouds, that is, when the spatial volume of the sub-region is less than the preset minimum volume threshold or the number of points contained therein is less than the preset minimum point number threshold, the division is stopped and it is marked as the final spatial sub-unit; then the geometric centers of all the spatial sub-units are projected to form a two-dimensional feature point set.

[0109] In a specific embodiment, step 104 includes:

[0110] 1041、According to the spatial position and spatial topological relationship of the bounding box node recorded in the bounding box hierarchy structure, each encoded data is decoded into an original three-dimensional coordinate value, and each compressed data is restored to a continuous sequence, the original three-dimensional coordinate value and the continuous sequence are mapped to the corresponding three-dimensional spatial coordinate position, and spatial topological splicing and reorganization are performed.

[0111] The spatial topology splicing refers to recombining scattered data into a complete three-dimensional scene by using the position relationship of bounding box nodes.

[0112] In step 1041, the encoding data can be restored to original three-dimensional coordinate values by using a preset coordinate restoration formula, and the compressed data is decompressed to generate a repeated point set; then, according to the spatial position relationship recorded by the bounding box level, all points are spliced into a complete three-dimensional scene point cloud, providing basic data for subsequent processing.

[0113] 1042, after recombination, the coordinate axis direction with the largest coordinate axis direction of the point cloud coordinate variance in the division region is selected, and the division region is divided into a plurality of candidate positions along the selected coordinate axis direction, the point number difference value of the space on both sides of each candidate position is calculated, the candidate position with the smallest point number difference value is selected as the final division position, and when the spatial volume of the final division position is less than a preset minimum volume threshold or the point number in the space is less than a preset minimum point number threshold, a spatial subunit is generated and the division is terminated.

[0114] The division region is a clustering unit for adjusting the division volume in real time according to the point density.

[0115] In step 1042, the density d is calculated as follows: Wherein, S is the point number, and V is the division volume. If the density d exceeds the preset gradient threshold, the division volume is reduced; if the threshold is not exceeded, the division volume is increased. This dynamic adjustment ensures that the high detail area is finely processed, and the background area is efficiently covered.

[0116] In step 1042, the density d is calculated as follows:

[0117] Step b1: calculate the arithmetic mean of the coordinate values of all three-dimensional coordinate points in the division region in X, Y and Z directions, and obtain the dispersion degree value based on the arithmetic mean and the total number of coordinate values, and select the direction with the largest dispersion degree value as the selected coordinate axis direction.

[0118] In step b1, the arithmetic mean in the direction a is calculated as follows: Wherein, Indicates the arithmetic mean in the direction a, S indicates the total number of three-dimensional coordinate points in the current division region, Indicates the coordinate value of the i-th point in the direction a, and the square sum of the difference between each coordinate value and the arithmetic mean is calculated and divided by the total number of coordinate values to obtain the dispersion degree value. The calculation formula of the dispersion degree value is as follows: And the direction with the largest dispersion degree value is selected as the division direction.

[0119] ​Step b2: traverse the coordinate values of all three-dimensional coordinate points in the current partition region in the direction of the selected coordinate axis, take the minimum coordinate value as the minimum boundary value and the maximum coordinate value as the maximum boundary value; based on the minimum boundary value and the maximum boundary value, generate a plurality of candidate positions by equidistant division with a preset fixed interval value.

[0120] Step b3: for each candidate position in the candidate partition set, divide the current region into a first subunit and a second subunit with the candidate position as the boundary, respectively count the number of three-dimensional coordinate points of the first subunit and the second subunit, calculate the absolute difference value of the number of three-dimensional coordinate points of the two, and select the candidate partition plane position with the minimum absolute difference value as the final partition position.

[0121] In step b3, for each candidate position in the candidate partition set , divide the current region into a first subunit and a second subunit with the candidate position as the boundary , count the number of three-dimensional coordinate points of the two subunits and respectively, calculate the absolute difference value of the two , and select the candidate partition plane position with the minimum absolute difference value as the final partition position.

[0122] Step b4: preset a minimum volume threshold of a space unit and a preset minimum number threshold, and define them as a preset minimum volume threshold and a preset minimum number threshold respectively. When the volume of the final partition position is less than the preset minimum volume threshold or the total number of three-dimensional coordinate points in the unit is less than the preset minimum number threshold, generate a space subunit mark and stop partitioning, otherwise, divide the current region into two new region units with the final partition position, and recursively execute the above partitioning process for each new region unit until all generated new region units meet the stop partitioning condition.

[0123] In step b4, preset a minimum volume threshold of a space unit and a preset minimum number threshold . When the volume of the final partition position after division or the number of three-dimensional coordinate points in the unit , generate a space subunit mark and stop partitioning; otherwise, divide the current region into two new region units with the final partition position, and recursively execute the above partitioning process for each new region unit until all new region units meet the stop partitioning condition.

[0124] ​​​1043. The arithmetic mean position of all points in the spatial sub-unit is used as the geometric center, projected onto a preset two-dimensional plane to generate two-dimensional coordinate points, and the two-dimensional coordinate points of all spatial sub-units are combined to form a two-dimensional feature point set.

[0125] In step 1043, the geometric center coordinates of all points in the sub-unit where the division terminates are calculated, using the following formula: Where C represents the geometric center of the spatial sub-unit. This indicates the number of point clouds within a spatial sub-unit. , , Let C represent the three-dimensional coordinate components of the i-th point within the sub-unit. Then, project the center point C onto the horizontal plane to generate a two-dimensional coordinate point. ,in, and The x and y component coordinates of the geometric center C are given respectively, and the two-dimensional coordinates of all spatial sub-units are output. Then, they are arranged in order according to the spatial topological relationship to form a two-dimensional feature point set.

[0126] This application preserves the complete details of key geometric features through high-precision encoding, and significantly reduces the data volume by using "coordinates × number" compression storage for the background region. Subsequently, the spatial partitioning granularity is dynamically adjusted based on the point cloud density, which can achieve a balance between fine processing of high-detail areas and efficient coverage of smooth areas. Then, the three-dimensional point cloud is condensed into a two-dimensional key feature point set, which can accelerate the feature extraction process by combining non-uniform clustering mechanism while ensuring the integrity of the core structure.

[0127] 105. Construct a spatial index structure based on the spatial distribution of the two-dimensional feature point set to realize real-time spatial retrieval of dynamic scenes from multiple perspectives.

[0128] In one specific implementation, step 105 includes:

[0129] Step 1051: Using the planar region covered by the two-dimensional feature point set as the root node, divide the planar region into multiple sub-regions of equal area.

[0130] Step 1052: Count the number of two-dimensional feature points in each sub-region. If the number of two-dimensional feature points is less than or equal to a preset minimum point threshold, assign a region identifier to the corresponding sub-region, and associate the region identifier with the spatial location and spatial topology relationship of the bounding box node to construct a spatial index structure.

[0131] The root node refers to a two-dimensional boundary range of the entire scene, the preset minimum point threshold refers to a preset sub-region termination division condition, the spatial identifier refers to a code uniquely identifying a sub-region, and the bidirectional mapping relationship refers to mutual correlation of a sub-region code and a spatial position and a spatial topological relationship of a three-dimensional bounding box.

[0132] The application can realize fast positioning of a target region driven by coding by constructing a hierarchical spatial index structure, then establish a region spatial topological relationship, and optimize resource allocation through a balance mechanism of sparse region coarse indexing and dense region fine indexing, thereby improving the retrieval efficiency and real-time response capability of a multi-view dynamic scene.

[0133] Figure 3 A structure diagram of a video image fast cropping and dimension reduction system based on hierarchical bounding boxes is provided for an embodiment of the application, as shown in the drawing, the system comprises: Figure 3

[0134] The acquisition module 31 is configured to acquire point cloud data of a to-be-processed image in a multi-view dynamic scene, perform spatial division on the point cloud data by using a hierarchical bounding box, and obtain a bounding box hierarchical structure, wherein the bounding box hierarchical structure is composed of a plurality of bounding box nodes.

[0135] The separation module 32 is configured to pre-filter the multi-view dynamic scene by using a diffraction element array, divide the bounding box nodes in the bounding box hierarchical structure into a first subset and a second subset based on wavelength selectivity and distribution characteristics of point cloud density.

[0136] The first processing module 33 is configured to perform processing on the bounding box nodes in the first subset by using adaptive pulse code modulation, dynamically allocate a number of coding bits according to a point cloud curvature change rate, obtain coding data, and perform compression processing on the bounding box nodes in the second subset by using run-length coding, and obtain compressed data.

[0137] The second processing module 34 is configured to input the coding data and the compressed data into an improved k-d tree based on a spatial topological relationship of the bounding box hierarchical structure, and perform non-uniform clustering processing to obtain a two-dimensional feature point set.

[0138] The construction module 35 is configured to construct a spatial index structure according to a spatial distribution of the two-dimensional feature point set, so as to realize real-time spatial retrieval of the multi-view dynamic scene.

[0139] Figure 3 The video image fast cropping and dimension reduction system based on hierarchical bounding boxes can perform the following steps: Figure 1 ​The implementation principle and technical effects of the video image fast clipping and dimension reduction method based on hierarchical bounding boxes according to the illustrated embodiment will not be described again. The specific operation modes of each module and unit in the video image fast clipping and dimension reduction system based on hierarchical bounding boxes according to the above embodiment have been described in detail in the embodiments related to the method, and will not be described in detail here.

[0140] In one possible design, Figure 3 The video image fast clipping and dimension reduction system based on hierarchical bounding boxes according to the illustrated embodiment can be implemented as a computing device, such as Figure 4 According to the illustrated embodiment, the computing device can include a storage component 41 and a processing component 42.

[0141] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42.

[0142] The processing component 42 is configured to perform the above Figure 1 The embodiment of the video image fast clipping and dimension reduction method based on hierarchical bounding boxes.

[0143] The processing component 42 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method.

[0144] The storage component 41 is configured to store various types of data to support the operation of the terminal. The storage component can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk or optical disk.

[0145] Of course, the computing device can also include other components, such as input / output interface, display component, communication component, etc.

[0146] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0147] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0148] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can be a cloud server, and the processing component, the storage component, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0149] The embodiment of the application further provides a computer storage medium, which stores a computer program, and the computer program can realize the above method when being executed by a computer. Figure 1 The embodiment of the application further provides a computer storage medium, which stores a computer program, and the computer program can realize the above method when being executed by a computer.

[0150] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0151] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0152] Through the foregoing description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, so as to make a computer device (which can be a personal computer, a server, etc.) execute the method described in each embodiment or some parts of the embodiment.

[0153] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the application, and not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A method for fast video image cropping and dimensionality reduction based on hierarchical bounding boxes, characterized in that, include: Point cloud data of the image to be processed is acquired in a multi-view dynamic scene. The point cloud data is spatially divided using hierarchical bounding boxes to obtain a bounding box hierarchy structure, which consists of multiple bounding box nodes. By pre-filtering the multi-view dynamic scene using a diffraction element array, the bounding box nodes in the bounding box hierarchy are divided into a first subset and a second subset based on wavelength selectivity and point cloud density distribution characteristics. Adaptive pulse code modulation is used to process the bounding box nodes in the first subset to dynamically allocate the number of bits for encoding according to the curvature change rate of the point cloud, thereby obtaining encoded data. Run-length encoding is used to compress the bounding box nodes in the second subset to obtain compressed data. Based on the spatial topological relationship of the bounding box hierarchy, the encoded data and compressed data are input into the improved kd tree for non-uniform clustering to obtain a two-dimensional feature point set. A spatial index structure is constructed based on the spatial distribution of the two-dimensional feature point set to achieve real-time spatial retrieval of dynamic scenes from multiple perspectives.

2. The method according to claim 1, characterized in that, The pre-filtering of the multi-view dynamic scene using a diffraction element array, based on wavelength selectivity and point cloud density distribution characteristics, divides the bounding box nodes in the bounding box hierarchy into a first subset and a second subset, including: The system receives reflected light signals from the multi-view dynamic scene and inputs the reflected light signals into a diffraction element array. The diffraction element array is composed of multiple diffraction units arranged with a preset angle difference. Each diffraction unit selectively decomposes the reflected light signal into a first optical path and a second optical path for independent transmission according to wavelength. Independent first processing end and second processing end are respectively set for the first optical path and the second optical path, so that the light intensity distribution of the first optical path is received through the first processing end to generate first image data; and the phase distribution of the second optical path is received through the second processing end to generate second image data. Extract the distribution features of point cloud density inside each bounding box node. The distribution features of point cloud density include the density change amplitude. Bounding box nodes whose density change amplitude exceeds a preset density threshold are marked as first associated nodes. Bounding box nodes whose density change amplitude does not exceed the preset density threshold are marked as second associated nodes. In the first image data, pixel clusters with brightness gradients exceeding a preset gradient threshold are mapped to the corresponding first associated nodes. In the second image data, pixel regions with uniform phase distribution are mapped to the corresponding second associated nodes. Based on the mapping relationship, the set to which each bounding box node belongs is determined to be either the first subset or the second subset.

3. The method according to claim 1, characterized in that, The bounding box nodes in the first subset are processed using adaptive pulse code modulation to dynamically allocate the number of encoding bits according to the point cloud curvature change rate, thereby obtaining encoded data. The bounding box nodes in the second subset are then compressed using run-length encoding to obtain compressed data, including: Traverse each point within the bounding box nodes of the first subset, calculate the curvature change rate of points in the local neighborhood, and classify the points in the local neighborhood into one of a number of predefined change rate levels based on the curvature change rate. Obtain the three-dimensional coordinate values ​​of the bounding box nodes in the first subset, scale the three-dimensional coordinate values ​​according to the allocated encoding bit length, and convert them into binary encoding of a specified bit length to form encoded data; Obtain the 3D coordinate values ​​of the bounding box nodes in the second subset, and scan them according to the preset spatial scanning order. Identify points with the same 3D coordinate values ​​that appear consecutively as a continuous sequence, and represent the combination of continuous sequences as compressed data.

4. The method according to claim 1, characterized in that, Based on the spatial topological relationship of the bounding box hierarchy, the encoded data and compressed data are input into an improved kd-tree for non-uniform clustering processing to obtain a two-dimensional feature point set, including: Based on the spatial location and spatial topology of the bounding box nodes recorded in the bounding box hierarchy, each encoded data is decoded into the original three-dimensional coordinate value, each compressed data is restored into a continuous sequence, the original three-dimensional coordinate values ​​and the continuous sequence are mapped to the corresponding three-dimensional spatial coordinate positions, and spatial topology splicing and recombination are performed. After reorganization, the coordinate axis direction with the largest variance of point cloud coordinates within the division area is selected, and the area is divided into multiple candidate positions along the selected coordinate axis direction. The difference in the number of points on both sides of each candidate position is calculated, and the candidate position with the smallest difference in the number of points is selected as the final division position. When the spatial volume of the final division position is less than a preset minimum volume threshold or the number of points in the space is less than a preset minimum number of points threshold, a spatial sub-unit is generated and the division is terminated. The arithmetic mean of all points within the spatial sub-unit is used as the geometric center, projected onto a preset two-dimensional plane to generate two-dimensional coordinate points, and the two-dimensional coordinate points of all spatial sub-units are combined to form a two-dimensional feature point set.

5. The method according to claim 1, characterized in that, Constructing a spatial index structure based on the spatial distribution of the two-dimensional feature point set includes: Using the planar region covered by the two-dimensional feature point set as the root node, the planar region is divided into multiple sub-regions of equal area; The number of two-dimensional feature points in each sub-region is counted. If the number of two-dimensional feature points is less than or equal to a preset minimum point threshold, a region identifier is assigned to the corresponding sub-region. The spatial location and spatial topology relationship between the region identifier and the bounding box node are associated to construct a spatial index structure.

6. The method according to claim 1, characterized in that, Point cloud data is spatially partitioned using hierarchical bounding boxes to obtain a bounding box hierarchy. This hierarchy consists of multiple bounding box nodes, including: The point cloud data as a whole is used as the initial bounding box node of the first level. The minimum number of points and the minimum spatial volume threshold within the initial bounding box node are set. The spatial volume of the initial bounding box node is calculated, and the total number of points contained within the initial bounding box node is counted. When the total number of points is greater than the minimum number of points and the spatial volume is greater than the minimum spatial volume threshold, the points are cut along the three coordinate axes respectively. Two symmetrical spatial interface positions are generated in each coordinate axis direction. The volume ratio of the central region between the two symmetrical spatial interfaces under the same coordinate axis is calculated, and the coordinate axis direction with the smallest volume ratio of the central region is selected as the dividing direction. Based on the division direction, the region of the initial bounding box node is divided into two sub-regions. Sub-level bounding box nodes are created for the sub-regions. A hierarchical relationship link is established between the sub-level bounding box nodes and the initial bounding box node, and the nodes are added to the queue to be processed. If there are unprocessed bounding box nodes in the queue, repeat the space partitioning process until there are no unprocessed bounding box nodes in the queue, and output the bounding box hierarchy structure composed of bounding box nodes of all levels.

7. The method according to claim 3, characterized in that, Obtain the 3D coordinate values ​​of the bounding box nodes in the first subset, scale and convert the 3D coordinate values ​​into binary code of a specified number of bits according to the allocated encoding bit width, forming encoded data, including: Obtain the three-dimensional coordinate values ​​of the bounding box nodes in the first subset, wherein the three-dimensional coordinate values ​​include three component values, calculate the maximum and minimum values ​​of the three component values, and determine the scaling factor based on the correspondence between the maximum and minimum values ​​and a preset integer range; Based on the scaling factor and the allocated number of bits, determine the specified number of bits corresponding to the binary code; The three component values ​​are converted into integer values ​​within a preset integer range, and the integer values ​​are converted into binary codes of a specified number of bits. The three binary codes are then combined to form a binary code. Arrange the binary codes of all bounding box nodes in the first subset in order to form encoded data.

8. The method according to claim 4, characterized in that, Select the coordinate axis direction with the largest variance of the point cloud coordinates within the partitioned area, and divide the area into multiple candidate locations along the selected coordinate axis direction, including: Calculate the arithmetic mean of the coordinate values ​​of all three-dimensional coordinate points in the X, Y, and Z directions within the divided region. Based on the arithmetic mean and the total number of coordinate values, obtain the dispersion value, and select the direction with the largest dispersion value as the selected coordinate axis direction. Traverse the coordinate values ​​of all three-dimensional coordinate points in the selected coordinate axis direction within the current partitioned region, and take the coordinate value with the smallest value as the minimum boundary value and the coordinate value with the largest value as the maximum boundary value; Based on the minimum and maximum boundary values, multiple candidate positions are generated by dividing the area equally with a preset fixed spacing value.

9. A fast video image cropping and dimensionality reduction system based on hierarchical bounding boxes, characterized in that, include: The acquisition module is used to acquire point cloud data of the image to be processed in a multi-view dynamic scene, and to spatially divide the point cloud data using hierarchical bounding boxes to obtain a bounding box hierarchy structure, which is composed of multiple bounding box nodes. The separation module is used to pre-filter the multi-view dynamic scene through the diffraction element array, so as to divide the bounding box nodes in the bounding box hierarchy into a first subset and a second subset based on the wavelength selectivity and the distribution characteristics of point cloud density. The first processing module is used to process the bounding box nodes in the first subset using adaptive pulse code modulation to dynamically allocate the number of encoding bits according to the curvature change rate of the point cloud to obtain encoded data, and to compress the bounding box nodes in the second subset using run-length encoding to obtain compressed data. The second processing module is used to input the encoded data and compressed data into the improved kd tree for non-uniform clustering based on the spatial topological relationship of the bounding box hierarchy to obtain a two-dimensional feature point set. The construction module is used to construct a spatial index structure based on the spatial distribution of the two-dimensional feature point set, so as to realize real-time spatial retrieval of dynamic scenes from multiple perspectives.

10. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a method for fast video image cropping and dimensionality reduction based on hierarchical bounding boxes as described in any one of claims 1 to 8.

Images