A 3D model compression and rendering optimization system for virtual interaction scenarios

Abstract

The application discloses a 3D model compression and rendering optimization system for a virtual interaction scene, and relates to the technical field of 3D model optimization.The application locks the same type of feature points by comparing feature vectors, records core position features by using a single standard point, and only stores angle data of the standard point instead of complete position information for the remaining same type of points.Meanwhile, the removed position data is backed up in the cloud, which not only realizes lightweight of the local model, but also provides guarantee for subsequent data recovery, balances the 'compression rate' and 'data security', selects the micro-section sequence with the smallest data fluctuation by calculating the 'data difference variance' of different division processes, and accurately matches the optimization thread number of the rendering center, so that the dual goals of 'thread load balancing-maximization of rendering efficiency' are realized, the rendering waiting time of the 3D model in the virtual scene is effectively reduced, and the user interaction experience is improved.

Description

A 3D model compression and rendering optimization system for virtual interactive scenarios Technical Field

[0001] This invention relates to the field of 3D model optimization technology, specifically a 3D model compression and rendering optimization system for virtual interactive scenarios. Background Technology

[0002] With the rapid popularization of virtual interactive scenarios such as virtual reality (VR), augmented reality (AR), virtual conferencing, and industrial digital twins, the contradiction between the data scale and rendering efficiency of 3D models, as the core carrier of scene construction, has become increasingly prominent, becoming a key bottleneck restricting the upgrading of virtual interactive experience.

[0003] The specific background can be discussed from three aspects: technical requirements, shortcomings of existing solutions, and pain points in the scenario. From the perspective of technical requirements, virtual interactive scenarios place dual demands on the "high precision" and "high smoothness" of 3D models.

[0004] To recreate realistic interaction logic, 3D models need to contain massive amounts of point data. Complex three-dimensional structures are built through dense points, and the number of points in some high-precision models can even reach millions, resulting in a sharp increase in the size of the model file. On the other hand, virtual interactive scenes need to support real-time response, requiring 3D models to load and render quickly. This requires finding a balance between "high model precision" and "lightweight data", and traditional technology systems can hardly meet both of these requirements at the same time.

[0005] Existing solutions often fail to address similar features within 3D models, requiring repeated storage of complete location data for each set of points, resulting in significant data redundancy. For example, for a set of corner points with identical structures, traditional solutions require recording the spatial coordinates of each point separately, without realizing that storage can be simplified using a "reference point + relative angle" approach, making it difficult to overcome the bottleneck in compression rate.

[0006] From the perspective of the shortcomings of existing rendering optimization solutions, current rendering technology is difficult to match the dynamic needs of virtual interactive scenarios. On the one hand, "batch rendering of disordered data" leads to low efficiency. Existing rendering systems mostly process the decompressed 3D model point data in batches without mining the data patterns. When faced with disordered feature point data, the algorithm needs to frequently adjust the calculation logic, resulting in slow rendering speed. In scenarios such as multi-user synchronous interaction and complex scene switching, stuttering and latency are likely to occur. On the other hand, "rigid thread allocation" causes resource waste.

[0007] Most systems use a fixed thread partitioning strategy (such as distributing data evenly among threads), without considering the differences in processing complexity of different data segments. Some threads are overloaded due to large fluctuations and poor regularity in the data they process, while other threads are idle because the data is simple. This results in low utilization of hardware resources and further exacerbates the contradiction between rendering efficiency and scene requirements.

[0008] In summary, existing 3D model compression and rendering technologies are no longer adequate to meet the development needs of virtual interactive scenarios, and there is an urgent need for an integrated optimization solution that combines "high-precision shape preservation, efficient compression, intelligent rendering, and security and reliability". Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a 3D model compression and rendering optimization system for virtual interactive scenarios. It solves the problem that existing solutions often fail to specifically process points with similar features within the 3D model, still requiring repeated storage of complete location data for each group of points, resulting in a large amount of data redundancy.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a 3D model compression and rendering optimization system for virtual interactive scenarios, comprising:

[0011] The compression processing center performs compression processing on the 3D models of virtual interactive scenes, specifically including:

[0012] The feature confirmation terminal confirms the positional features associated with different model points within the 3D model, then verifies these positional features. Based on the verification results, it confirms the feature points within the 3D model and simultaneously records the positional features of these feature points. The specific method is as follows:

[0013] Each model point within the 3D model is confirmed and recorded as a reference point. Then, the adjacent model points associated with the reference point are confirmed and recorded as subordinate points of the reference point. The reference point and subordinate points are then connected sequentially, and the subordinate points are connected in pairs simultaneously. It is confirmed whether a micro-model body can be generated. If so, it means that the subordinate point and the reference point are not in the same plane. In this case, the reference point is directly recorded as a feature point, and the spatial position of the feature point and the spatial position of the subordinate point are recorded simultaneously to generate the positional feature of the feature point. If not, it means that the subordinate point and the reference point are in the same plane. In this case, no marking is made, and the feature of the reference point is not obvious.

[0014] For similar point markers, based on the different location features associated with different feature points, the feature vector associated with each feature point is determined. Then, the confirmed feature vectors are compared and verified to lock similar feature points from a number of feature points. The specific method is as follows:

[0015] The location features associated with different feature points are determined, and the number of associated auxiliary points of the corresponding feature points is identified from the determined location features. Multiple groups of feature points with the same number of auxiliary points are integrated and recorded as the feature point set to be processed.

[0016] From a set of feature points to be processed, identify similar feature points: identify multiple associated points of a single feature point, connect these multiple associated points, and confirm whether all the connected associated points lie in the same plane.

[0017] If so, directly identify the perpendicular line between the feature point and the plane, and generate a feature vector belonging to the feature point with the feature point as the initial point and the perpendicular point as the endpoint.

[0018] If not, confirm the plane associated with multiple subordinate points, simultaneously confirm the midpoint of each plane, and then use the same method to confirm the planes generated between the midpoints of the planes, and continue to confirm until several midpoints of the planes are all located on the same plane. Then confirm the perpendicular line between the feature point distance and the plane, and generate the feature vector belonging to the feature point with the feature point as the initial point and the perpendicular point as the end point.

[0019] The feature vectors associated with different feature points are compared and verified. Feature points associated with multiple sets of feature vectors of the same length are classified into the same type of feature points and marked in the corresponding set of feature points to be processed.

[0020] On the data optimization recording end, data optimization processing is performed on the confirmed similar feature points. A single feature point is selected as a standard point from among these similar feature points. Then, the angle data between other feature points and the standard point is confirmed. Simultaneously, the positional features of other feature points are also optimized, completing the data compression and optimization process of the 3D model. The specific method is as follows:

[0021] Randomly select one feature point from the confirmed similar feature points as a standard point, and record the positional features of this standard point. Then, overlap other feature points with the standard point and synchronously control the other feature points to rotate in a circle. Stop when the positional features of other feature points completely overlap with the standard point. Record the associated angle data at the corresponding time. Bind the angle data with other feature points and record it as the auxiliary data of other feature points. Simultaneously, remove the positional features associated with other feature points.

[0022] The above data compression and optimization process is performed on each group of similar feature points to complete the data compression and optimization process of the 3D model.

[0023] The rendering optimization center performs rendering optimization on the decompressed 3D models, specifically including:

[0024] The point feature sorting end decompresses the 3D model after data compression to obtain a standard 3D model. Then, based on the differences in point features within the 3D model, it generates a set of feature vector sorting sequences. The specific method is as follows:

[0025] Identify a standard point from similar feature points, then identify other feature points associated with this standard point. Based on the positional features of the standard point and the angle data associated with other feature points, rotate the standard point in the opposite direction to obtain the positional features associated with other feature points, and then bind these positional features with the corresponding other feature points.

[0026] Based on the feature vectors associated with similar feature points, the length of the feature vector is denoted as CL. i Where i represents different groups of similar feature points, and several groups of CL are associated within each set of feature points to be processed. i Confirm them one by one, and then sort the confirmed CL groups in ascending order of their values. i The associated feature vectors are sorted to confirm the feature vector sorting sequence.

[0027] On the point optimization processing end, based on the confirmed feature vector sorting sequence and the number of optimization threads within this rendering optimization center, the feature vector sorting sequence is divided into multiple different optimization sequences, and each optimization sequence is optimized separately. Specifically:

[0028] Let G be the number of optimization threads in this rendering optimization center. Divide the sorted sequence of feature vectors associated with a single set of feature points to be processed into G micro-segments. Execute several partitioning processes and determine the process characteristics associated with each partitioning process. Confirm the length value CL associated with each micro-segment sequence. i Minimum value CL i min and maximum value CL i max, using: (CL) i max-CL i min) = Jc confirms the data difference Jc associated with the corresponding micro-segment sequence, and then performs variance processing on the multiple sets of data differences Jc associated with multiple micro-segment sequences to confirm the process characteristics belonging to this partitioning process;

[0029] From the process characteristics associated with different partitioning processes, the minimum value is selected, the partitioning process associated with the minimum value is denoted as the optimal process, the micro-segment sequence of the optimal process is denoted as the optimized sequence, and different optimization threads are used to optimize different optimized sequences.

[0030] This invention provides a 3D model compression and rendering optimization system for virtual interactive scenarios. Compared with existing technologies, it has the following advantages:

[0031] This invention employs a three-layer identification logic of "baseline point - auxiliary point - micro-model volume verification," identifying only non-coplanar points as feature points (such as corner points of a cube), effectively eliminating redundant points without significant features within the plane. Compared to the traditional "full-point recording" mode, this significantly reduces the amount of point data that needs to be stored, while accurately preserving the core features of the model's three-dimensional structure, avoiding model contour distortion during compression, and ensuring the visual integrity of the 3D model in virtual interactive scenes.

[0032] By comparing feature vectors, similar feature points (such as points with the same number of subordinate points and the same feature vector length) are identified. The core location features are recorded by a single standard point, while other similar points only store the angle data relative to the standard point, rather than the complete location information. At the same time, the location data that is removed is backed up in the cloud, which not only makes the local model lightweight, but also provides a guarantee for subsequent data recovery, balancing "compression ratio" and "data security".

[0033] By calculating the "data difference variance" of different partitioning processes, the micro-segment sequence with the smallest data fluctuation (optimized sequence) is selected and precisely matched with the number of optimized threads (G) in the rendering center. The feature vector sequence is divided into 3 micro-segments with the smallest data difference variance. Each thread independently processes a set of regular data, which avoids single-thread overload and prevents multi-thread resource waste, achieving the dual goal of "thread load balancing - maximizing rendering efficiency". This effectively reduces the rendering waiting time of 3D models in the virtual scene and improves the user interaction experience (such as no delay when switching virtual scenes and no lag when operating models). Attached Figure Description

[0034] Figure 1 is a schematic diagram of the principle framework of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Please refer to Figure 1. This application provides a 3D model compression and rendering optimization system for virtual interactive scenarios, including a compression processing center and a rendering optimization center, wherein the compression processing center and the rendering optimization center are electrically connected from the output node to the input node.

[0037] The compression processing center includes a point feature confirmation end, a similar point mark end, and a data optimization recording end. The rendering optimization center includes a point feature sorting end and a point optimization processing end. The compression processing center compresses the 3D model of the virtual interactive scene, while the rendering optimization center optimizes the rendering of the decompressed 3D model.

[0038] The point feature confirmation end confirms the positional features associated with different model points within the 3D model, then verifies the positional features, and confirms the feature points within the 3D model based on the verification results, while simultaneously recording the positional features of the feature points. Specifically, a model body consists of several faces, and a face consists of several points. There are corresponding intersections between each face. The intersections generated by three or more faces belong to the corresponding feature points, which can be understood as the corner points of a cube. Such corner points are feature points.

[0039] The specific method for confirming feature points within a 3D model is as follows:

[0040] Each model point within the 3D model is confirmed and recorded as a reference point. Then, the adjacent model points associated with the reference point are confirmed and recorded as subordinate points of the reference point. The reference point and subordinate points are then connected sequentially, and the subordinate points are connected in pairs simultaneously. It is confirmed whether a micro-model body can be generated (it must be a solid, not a two-dimensional plane). If it can, it means that the subordinate point and the reference point are not in the same plane. In this case, the reference point is directly recorded as a feature point, and the spatial position of the feature point and the spatial position of the subordinate point are recorded simultaneously to generate the positional feature of the feature point. If it cannot, it means that the subordinate point and the reference point are in the same plane. In this case, no marking is made, and the feature of the reference point is not obvious.

[0041] Specifically, within the corresponding 3D model, there are a large number of model points, but there are definitely corresponding points generated by the intersection of multiple planes, and there are quite a lot of such points. In the original compression process, the data features of such points need to be recorded, which will result in a large capacity during the compression process. If the data features between these points are comprehensively optimized and the total amount of data features between these points is reduced, the corresponding 3D model can be effectively optimized.

[0042] Among them, for the same type of point marker, based on the different location features associated with different feature points, the feature vector associated with each feature point is determined, and then the confirmed feature vectors are compared and verified to lock down the same type of feature points from several feature points:

[0043] The location features associated with different feature points are determined, and the number of associated auxiliary points of the corresponding feature points is identified from the determined location features. Multiple groups of feature points with the same number of auxiliary points are integrated and recorded as the feature point set to be processed.

[0044] From a set of feature points to be processed, identify similar feature points: identify multiple associated points of a single feature point, connect these multiple associated points, and confirm whether all the connected associated points lie in the same plane.

[0045] If so, directly identify the perpendicular line between the feature point and the plane, and generate a feature vector belonging to the feature point with the feature point as the initial point and the perpendicular point as the endpoint.

[0046] If not, then confirm the planes associated with multiple subordinate points (i.e., the multiple planes to which they belong), simultaneously confirm the midpoint of each plane, and then use the same method to confirm the planes generated between the midpoints of the planes, and continue to confirm until several midpoints of the planes are all located on the same plane. Then confirm the perpendicular line between the feature point and the plane, and generate the feature vector belonging to the feature point with the feature point as the initial point and the perpendicular point as the end point. Specifically, if there are three subordinate points around a feature point, then there is only one plane, and the feature vector confirmation process can be directly carried out.

[0047] There are four subordinate points, each located at an arbitrary spatial position, and there is an associated plane between each subordinate point. That is, there can be at most three associated planes for the four points. So here we need to determine the center point of each associated plane. Based on the three center points, we can directly determine a set of planes.

[0048] If there are five subordinate points, there are at most four associated planes. Each associated plane has a center point. Then, four plane midpoints are confirmed here. The corresponding plane confirmation process is then executed for the four plane midpoints, thereby confirming three associated planes. Then, the midpoints of the three associated planes are confirmed, thereby confirming the plane where the three midpoints are located, thus directly locking the associated feature vector.

[0049] The feature vectors associated with different feature points are compared and verified. Feature points associated with multiple sets of feature vectors of the same length are classified into the same type of feature points and marked in the corresponding set of feature points to be processed. Specifically, the features of the auxiliary points associated with different sets of feature points to be processed are not the same. If a feature point has three auxiliary points around it, then the number of associated feature points and auxiliary points in the associated feature point set is the same, that is, they are classified in the same point set, which facilitates subsequent compression and verification.

[0050] The data optimization recording end performs data optimization processing on the confirmed similar feature points. A single feature point is selected as a standard point from among these similar feature points. Then, the angle data between other feature points and the standard point is confirmed. Simultaneously, the positional features of other feature points are also optimized, completing the data compression and optimization process of the 3D model. The specific data optimization process includes:

[0051] Randomly select a feature point from the confirmed similar feature points as a standard point, and record the positional features of this standard point. Then, overlap other feature points with the standard point and synchronously control the other feature points to rotate in a circle. Stop when the positional features of other feature points completely overlap with the standard point. Record the associated angle data at the corresponding moment, and bind the angle data with other feature points as the auxiliary data of other feature points. Simultaneously, remove the positional features associated with other feature points. The removed data will be stored in the cloud and removed from this 3D model.

[0052] The above data compression and optimization process is performed on each group of similar feature points to complete the data compression and optimization process of the 3D model.

[0053] Specifically, within a set of similar feature points, there are five sets of feature points. Feature point A is the standard point, and the other four feature points are the other feature points. Since the feature vector lengths are consistent, the trend characteristics associated with the surrounding auxiliary points and the feature points are also consistent. Therefore, based on the rotation characteristics of the other feature points and feature point A during the rotation process, the confirmation process of the other feature points can be confirmed, which facilitates subsequent data decompression processing.

[0054] Furthermore, the actual positions of the corresponding feature points within the 3D model have not changed. Here, the feature data between the points has been compressed and optimized, rather than the points being moved or altered.

[0055] Second Embodiment

[0056] In this embodiment, compared to the above embodiments, the main focus is on the data optimization process of the 3D model.

[0057] The feature sorting end decompresses the 3D model after data compression to obtain a standard 3D model. Then, based on the differences in feature variations within the 3D model, it generates a set of feature vector sorting sequences and transmits them to the feature optimization and care end. The generation method of the feature vector sorting sequences is as follows:

[0058] Identify a standard point from similar feature points, then identify other feature points associated with this standard point. Based on the positional features of the standard point and the angle data associated with other feature points, rotate the standard point in the opposite direction to obtain the positional features associated with other feature points, and then bind these positional features with the corresponding other feature points.

[0059] Based on the feature vectors associated with similar feature points, the length of the feature vector is denoted as CL. i Where i represents different groups of similar feature points, and several groups of CL are associated within each set of feature points to be processed. i Confirm them one by one, and then sort the confirmed CL groups in ascending order of their values. i The associated feature vectors are sorted to confirm the feature vector sorting sequence.

[0060] In the point optimization processing stage, based on the confirmed feature vector sorting sequence and the number of optimization threads within the rendering optimization center, the feature vector sorting sequence is divided into multiple different optimization sequences, and each optimization sequence is optimized separately:

[0061] Let G be the number of optimization threads in this rendering optimization center. Divide the sorted sequence of feature vectors associated with a single set of feature points to be processed into G micro-segments. Execute several partitioning processes and determine the process characteristics associated with each partitioning process. Confirm the length value CL associated with each micro-segment sequence. i Minimum value CL i min and maximum value CL i max, using: (CL) i max-CL i min) = Jc confirms the data difference Jc associated with the corresponding micro-segment sequence, and then performs variance processing on the multiple sets of data differences Jc associated with multiple micro-segment sequences to confirm the process characteristics belonging to this partitioning process;

[0062] From the process characteristics associated with different partitioning processes, the minimum value is selected, the partitioning process associated with the minimum value is recorded as the optimal process, the micro-segment sequence of the optimal process is recorded as the optimized sequence, and different optimization threads are used to optimize different optimized sequences.

[0063] Specifically, the feature vectors associated with each point are not the same. Here, we do not consider the direction, but only the length. According to the optimization logic of the system, the optimization rate of gradually changing features is faster than that of randomly changing features. The gradually changing features are those whose length values ​​gradually increase from small to large. By following this optimization method, each point can be optimized in turn to ensure the specific optimization effect, while not causing system burden during the optimization process, thus improving optimization efficiency and reducing optimization time.

[0064] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0065] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A 3D model compression and rendering optimization system for virtual interactive scenarios, characterized in that, include: The compression processing center compresses the 3D models of the virtual interactive scene. Specifically, this includes: a point feature confirmation end, which confirms the positional features associated with different model points within the 3D model, verifies these features, and confirms the feature points within the 3D model based on the verification results, simultaneously recording the positional features of the feature points; a similar point marking end, which determines the feature vector associated with each feature point based on the different positional features associated with different feature points, then compares and verifies the confirmed feature vectors to lock in similar feature points from a pool of feature points; and a data optimization and recording end, which performs data optimization processing on the confirmed similar feature points, selects a single feature point from the similar feature points as a standard point, and then confirms... The angle data between other feature points and the standard point are used to simultaneously optimize the positional features of other feature points, completing the data compression and optimization process of the 3D model. The rendering optimization center performs rendering optimization on the decompressed 3D model, specifically including: the point feature sorting end, which decompresses the 3D model after the data compression process to obtain a standard 3D model, and then generates a set of feature vector sorting sequences based on the differences in the changes of point features within the 3D model; the point optimization processing end, which divides the feature vector sorting sequence into multiple different optimization sequences based on the confirmed feature vector sorting sequence and the number of optimization threads existing in this rendering optimization center, and performs separate optimization processing on each optimization sequence.

2. The 3D model compression and rendering optimization system for virtual interactive scenarios according to claim 1, characterized in that, The specific method for confirming feature points within the 3D model using the point feature confirmation terminal is as follows: Each model point within the 3D model is confirmed, and the confirmed model point is recorded as a reference point. Then, the adjacent model points associated with the reference point are confirmed and recorded as subordinate points of the reference point. The reference point and subordinate points are then connected sequentially, and the subordinate points are connected pairwise simultaneously. It is then confirmed whether a micro-model body can be generated. If so, it means that the subordinate point and the reference point are not in the same plane, and the reference point is directly recorded as a feature point. Simultaneously, the spatial position of the feature point and the spatial position of the subordinate point are recorded to generate a positional feature belonging to the feature point. If not, it means that the subordinate point and the reference point are in the same plane, and no marking is performed.

3. The 3D model compression and rendering optimization system for virtual interactive scenarios according to claim 1, characterized in that, The specific method for locking similar feature points from several feature points is as follows: determine the location features associated with different feature points, identify the number of associated auxiliary points of the corresponding feature point from the determined location features, and integrate multiple groups of feature points with the same number of auxiliary points into a set of feature points to be processed. The process involves identifying similar feature points from a set of feature points to be processed: Multiple associated points are identified for each feature point, and these associated points are connected. It is then confirmed whether all associated points lie in the same plane. If so, the perpendicular line between the feature point and the plane is directly confirmed, generating a feature vector belonging to this feature point with the feature point as the initial point and the perpendicular as the endpoint. If not, the planes associated with the multiple associated points are confirmed, and the midpoint of each plane is simultaneously confirmed. The same method is used to confirm the planes generated between the midpoints of the associated points, and this process continues until all midpoints of several planes are located in the same plane. The perpendicular line between the feature point and the plane is then confirmed, generating a feature vector belonging to this feature point with the feature point as the initial point and the perpendicular as the endpoint. The feature vectors associated with different feature points are compared and verified. Feature points associated with multiple sets of feature vectors of consistent length are classified as similar feature points and labeled within the corresponding set of feature points to be processed.

4. The 3D model compression and rendering optimization system for virtual interactive scenarios according to claim 1, characterized in that, The specific method for optimizing the positional features of other feature points in the data optimization recording terminal is as follows: randomly select a feature point as a standard point from the confirmed feature points of the same type, record the positional features of this standard point, then overlap other feature points with the standard point, and synchronously control the other feature points to rotate in a circle. When the positional features of other feature points and the standard point are completely overlapped, record the associated angle data at the corresponding time, bind the angle data with other feature points, record it as the auxiliary data of other feature points, and synchronously remove the positional features associated with other feature points. For each group of similar feature points, a data compression and optimization process is performed to complete the data compression and optimization process of the 3D model.

5. The 3D model compression and rendering optimization system for virtual interactive scenarios according to claim 1, characterized in that, The specific method for generating the feature vector sorting sequence at the point feature sorting end is as follows: First, identify a standard point from among similar feature points. Then, identify other feature points associated with this standard point. Based on the positional features of the standard point and the angle data associated with other feature points, rotate the standard point in the opposite direction using the corresponding angle data to obtain the positional features associated with other feature points. Then, bind these positional features with the corresponding other feature points. The length of the feature vector associated with similar feature points is denoted as CL. i Where i represents different groups of similar feature points, and several groups of CL are associated within each set of feature points to be processed. i Confirm them one by one, and then sort the confirmed CL groups in ascending order of their values. i The associated feature vectors are sorted to confirm the feature vector sorting sequence.

6. The 3D model compression and rendering optimization system for virtual interactive scenarios according to claim 1, characterized in that, The point optimization processing terminal divides the feature vector sorting sequence into multiple optimization sequences in the following way: Let G be the number of optimization threads existing in this rendering optimization center; divide the feature vector sorting sequence associated with a single set of feature points to be processed into G micro-segment sequences; execute several division processes; determine the process characteristics associated with each division process; and confirm the length value CL associated with each micro-segment sequence. i Minimum value CL i min and maximum value CL i max, using: (CL) i max-CL i min) = Jc confirms the data difference Jc associated with the corresponding micro-segment sequence, and then performs variance processing on the multiple sets of data differences Jc associated with multiple micro-segment sequences to confirm the process characteristics belonging to this partitioning process; From the process characteristics associated with different partitioning processes, the minimum value is selected, the partitioning process associated with the minimum value is denoted as the optimal process, the micro-segment sequence of the optimal process is denoted as the optimized sequence, and different optimization threads are used to optimize different optimized sequences.

Images