A method and apparatus for managing an ODN element waveform

Abstract

The application discloses an ODN element waveform management method and device, and relates to the technical field of optical distribution network. The ODN element waveform management method comprises the following steps: collecting the coordinates of discrete points of the curve of reflected light waveforms of different wavelengths of an ODN, and dividing out the coordinate set of points of an element waveform with a single type; constructing the relationship between the coordinate set of points of all types of element waveforms and corresponding element waveform label names, and extracting the sample data features of element waveform curves with the same element waveform label name; constructing element waveform label vector clusters based on element waveforms with the same element waveform label name, and constructing the non-full connection directed graph network relationship between a plurality of element waveform label vector clusters; and identifying element waveforms based on the non-full connection directed graph network relationship. The application can effectively improve the accuracy of element waveform identification, and save the resources and costs consumed by neural network or intelligent model training.

Description

Technical Field

[0001] This application relates to the field of Optical Distribution Network (ODN) technology, specifically to a method and apparatus for managing ODN primitive waveforms. Background Technology

[0002] Given the large number of optical splitters already deployed in the existing ODN network, and the prospect of deploying even more in the future, there is a need to provide system-level management methods for real-time automatic management of optical splitter resources and their status. This is necessary to meet the urgent need for efficient processing of the waveform big data of reflected light from massive optical splitters, thereby achieving comprehensive, dynamic, real-time, and efficient management of the ODN network and providing management technology support for the "autopilot" operation of the ODN network.

[0003] A prerequisite for automated optical distribution resource and state management is the classification and identification of a subset of the reflected light waveform curves of the ODN network—the meta-waveform—to facilitate further processing of related reasoning logic. The meta-waveform refers to a local waveform curve with three continuous inflection points obtained by traversing forward and backward from the corresponding position of the reflection event to obtain the complete waveform of the reflected light returned by a single probe of the ODN line. Its typical characteristics are visually represented as: trough (first inflection point) - peak (second inflection point) - trough (third inflection point).

[0004] Furthermore, due to limitations in the accuracy of sampling data from optical sensing monitoring equipment and the attenuation effect of ODN lines on reflected light signals, waveform curves with different label types exhibit high morphological similarity. Consequently, the feature values ​​extracted from these samples via neural networks have low discriminative power, leading to reduced accuracy in identifying novel waveforms in the test input. For example, Figure 1 Meta-waveform sample data with different label types and similar shapes. Figure 1 The first curve at the top is a magnified view of the second curve at the bottom. From the curves above, we can see that the 6th and 13th element waveforms are similar in shape, but their label types differ. The label for the 6th element waveform is "cascaded reflection at the fracture surface of a two-way port direct-connected fiber," and the label for the 13th element waveform is "non-cascaded reflection at the fracture surface of a two-way port direct-connected fiber," and the label for the 13th element waveform is "non-cascaded reflection at the fracture surface of a two-way port direct-connected fiber."

[0005] One solution to this problem is to use massive amounts of sample data to build neural networks or intelligent models with more complex structures and extract more accurate sample feature values. However, this cannot avoid problems such as excessively long sample collection cycles, excessive labeling workload, more time-consuming training, and more complex network models.

[0006] Furthermore, the features corresponding to a certain type of label currently trained are low-dimensional and discrete, failing to reflect the high-dimensional, non-fully connected directed graph interconnections between the data in the meta-waveform, thus hindering accurate classification after feature matching of the meta-waveform. For example, in Figure 2 In the middle, meta-waveform 4 is forward dependent on meta-waveform 1; meta-waveform 13 is forward dependent on 10; meta-waveforms 11 to 18 are clustered meta-waveforms, which are aggregated and complementary.

[0007] In terms of representation patterns, the labels corresponding to the meta-waveforms trained by existing methods are also independent of each other and have no correlation, failing to map to a relationship in a high-dimensional non-fully connected directed graph. Summary of the Invention

[0008] This application provides a method and apparatus for managing ODN meta-waveforms, which can effectively improve the accuracy of meta-waveform recognition and save the resources and costs consumed in training neural networks or intelligent models.

[0009] In a first aspect, embodiments of this application provide a method for managing ODN meta-waveforms, the method comprising:

[0010] The coordinates of discrete points of the reflected light waveforms of different wavelengths in the optical distribution network (ODN) are collected, and the coordinate set of points with a single type of meta-waveform is segmented out.

[0011] Construct the coordinate set of points for all types of meta-waveforms and the relationship between the corresponding meta-waveform label names, and extract the sample data features of meta-waveform curves with the same meta-waveform label name;

[0012] Meta-waveform label vector clusters are constructed based on meta-waveforms with the same meta-waveform label name, and non-fully connected directed graph network relationships are constructed between multiple meta-waveform label vector clusters. The non-fully connected directed graph network relationships include forward and backward dependencies and sibling aggregation complementary dependencies between meta-waveform label vector clusters.

[0013] Based on the non-fully connected directed graph network relationship, the meta-waveform is identified.

[0014] In conjunction with the first aspect, in one implementation, the identification of the meta-waveform based on the non-fully connected directed graph network relationship includes:

[0015] Based on the non-fully connected directed graph network relationship, the meta-waveform label vector cluster number of the current input waveform is searched to obtain the sequence of uplink meta-waveform label vector cluster numbers that have a forward-backward dependency relationship with the meta-waveform label vector cluster number.

[0016] Based on each element of the sequence of uplink meta-waveform tag vector cluster numbers, a search is performed again based on the non-fully connected directed graph network relationship to obtain the number of meta-waveform tag vector clusters that have a homogeneous and complementary dependency relationship with the uplink meta-waveform tag vector clusters.

[0017] Based on the sequence of uplink meta-waveform label vector cluster numbers and the numbers of meta-waveform label vector clusters with complementary dependencies at the same level, the meta-waveform label vector cluster numbers of the current input waveform are matched.

[0018] When a match is completed, determine the meta-waveform label type of the meta-waveform label vector cluster of the current input waveform.

[0019] In conjunction with the first aspect, in one implementation method,

[0020] The forward-backward dependency relationship indicates that the distribution shape of the subsequent waveform depends on the distribution position and shape of the preceding waveform.

[0021] The complementary dependency relationship of the same level aggregation represents the aggregation of meta-waveforms at the same reflection position in different scenarios.

[0022] In conjunction with the first aspect, in one implementation, the type of the meta-waveform tag includes:

[0023] One-way reflection, one-way port reflector reflection, one-way port direct-connected ONU reflection, one-way port reflector direct-connected ONU reflection, one-way port direct-connected fiber fracture surface reflection, two-way reflection, two-way port direct-connected ONU reflection, two-way port reflector direct-connected ONU reflection, two-way port direct-connected fiber fracture surface cascade reflection, two-way port reflector non-cascade reflection, two-way port direct-connected fiber fracture surface non-cascade reflection, and two-way port reflector cascade reflection.

[0024] In conjunction with the first aspect, in one embodiment, after segmenting the set of coordinates of points having a single type of meta-waveform, the method further includes:

[0025] The step of normalizing the coordinate set of points of the original waveform includes loss compensation and scaling.

[0026] Secondly, embodiments of this application provide an ODN meta-waveform management device, the ODN meta-waveform management device comprising:

[0027] The acquisition module is used to acquire the coordinates of discrete points of the reflected light waveform curves of different wavelengths in the optical distribution network (ODN) and segment out the coordinate set of points with a single type of meta-waveform.

[0028] The module is used to construct the coordinate set of points of all types of meta-waveforms and the relationship between the corresponding meta-waveform label names, and to extract the sample data features of meta-waveform curves with the same meta-waveform label name.

[0029] The construction module also constructs a meta-waveform tag vector cluster based on meta-waveforms with the same meta-waveform tag name, and constructs a non-fully connected directed graph network relationship between multiple meta-waveform tag vector clusters. The non-fully connected directed graph network relationship includes forward and backward dependencies and sibling aggregation complementary dependencies between meta-waveform tag vector clusters.

[0030] The identification module identifies the original waveform based on the non-fully connected directed graph network relationship.

[0031] In conjunction with the second aspect, in one implementation, the identification module identifies the meta-waveform based on the non-fully connected directed graph network relationships, including:

[0032] Based on the non-fully connected directed graph network relationship, the meta-waveform label vector cluster number of the current input waveform is searched to obtain the sequence of uplink meta-waveform label vector cluster numbers that have a forward-backward dependency relationship with the meta-waveform label vector cluster number.

[0033] Based on each element of the sequence of uplink meta-waveform tag vector cluster numbers, a search is performed again based on the non-fully connected directed graph network relationship to obtain the number of meta-waveform tag vector clusters that have a homogeneous and complementary dependency relationship with the uplink meta-waveform tag vector clusters.

[0034] Based on the sequence of uplink meta-waveform label vector cluster numbers and the numbers of meta-waveform label vector clusters with complementary dependencies at the same level, the meta-waveform label vector cluster numbers of the current input waveform are matched.

[0035] When a match is completed, determine the meta-waveform label type of the meta-waveform label vector cluster of the current input waveform.

[0036] In conjunction with the second aspect, in one implementation, the forward-backward dependency relationship indicates that the distribution pattern of the subsequent waveform depends on the distribution position and pattern of the preceding waveform.

[0037] The complementary dependency relationship of the same level aggregation represents the aggregation of meta-waveforms at the same reflection position in different scenarios.

[0038] In conjunction with the second aspect, in one implementation, the type of the meta-waveform tag includes:

[0039] One-way reflection, one-way port reflector reflection, one-way port direct-connected ONU reflection, one-way port reflector direct-connected ONU reflection, one-way port direct-connected fiber fracture surface reflection, two-way reflection, two-way port direct-connected ONU reflection, two-way port reflector direct-connected ONU reflection, two-way port direct-connected fiber fracture surface cascade reflection, two-way port reflector non-cascade reflection, two-way port direct-connected fiber fracture surface non-cascade reflection, and two-way port reflector cascade reflection.

[0040] In conjunction with the second aspect, in one embodiment, after segmenting the coordinate set of points having a single type of meta-waveform, the acquisition module is further configured to:

[0041] The step of normalizing the coordinate set of points of the original waveform includes loss compensation and scaling.

[0042] The beneficial effects of the technical solutions provided in this application include at least the following:

[0043] The ODN meta-waveform management method in this application, at a low-dimensional level, performs discrete labeling of meta-waveforms with relatively independent semantics and completes feature training for the same type of meta-waveforms. In application, for the meta-waveforms corresponding to a subset of unstructured data of discrete coordinate points of the light reflection waveform curves of the ODN network, identification can map them to structured data that the algorithm can process; by constructing a convolutional neural network, meta-waveform features can be flexibly extracted; and the limitations of judgment based on fixed threshold features are avoided.

[0044] At the high-dimensional level, multiple labels from the low-dimensional level are used to establish multi-label vector clusters; non-fully connected directed graph relationships are established between multi-label vector clusters; thus, a more accurate determination of the label type corresponding to the sample in the non-fully connected directed graph network is provided, which depends on the forward and backward interconnection relationships. This can effectively improve the accuracy of meta-waveform recognition and save the resources and costs consumed in training neural networks or intelligent models.

[0045] Furthermore, the provided layer-specific non-fully connected directed graph network type labels and their associated characteristics will provide effective support for the construction of optical split network resource instance objects, the construction of topology connection relationships, and the monitoring of the status of optical split network resources and topology relationships. Attached Figure Description

[0046] Figure 1 A schematic diagram of meta-waveform sample data of different label types with similar shapes;

[0047] Figure 2 Example diagram showing the relationship between meta-waveforms of forward and backward dependencies and complementary dependencies of sibling aggregations;

[0048] Figure 3This is a flowchart of an embodiment of the ODN meta-waveform management method of this application;

[0049] Figure 4 This application presents a flowchart illustrating the construction process of a non-fully connected directed graph with meta-waveform labels based on three layers.

[0050] Figure 5 This is a schematic diagram illustrating the feature extraction of the meta-waveform based on the convolutional neural network in this application;

[0051] Figure 6 This is a schematic diagram illustrating an example of creating a non-fully connected directed graph with meta-waveform labels based on three layers in this application.

[0052] Figure 7 This is a non-fully connected directed graph formed by the primitive waveform label vector clusters of this application;

[0053] Figure 8 This is the main process for the precise matching of the waveform labels of the original application through a non-connected directed graph. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0055] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0056] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0057] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0058] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0060] Firstly, embodiments of this application provide a method for managing ODN meta-waveforms.

[0061] Reference Figure 3 , Figure 3 This is a flowchart illustrating an embodiment of the ODN waveform management method of this application. Figure 1 As shown, the management methods for ODN meta-waveforms include:

[0062] S1. Collect the coordinates of discrete points of the reflected light waveform curves of different wavelengths in the optical distribution network (ODN), and segment out the coordinate set of points with a single type of meta-waveform.

[0063] S2. Construct the coordinate set of points for all types of meta-waveforms and the relationship between the corresponding meta-waveform label names, and extract the sample data features of meta-waveform curves with the same meta-waveform label name.

[0064] S3. Construct a meta-waveform label vector cluster based on meta-waveforms with the same meta-waveform label name, and construct a non-fully connected directed graph network relationship between multiple meta-waveform label vector clusters. The non-fully connected directed graph network relationship includes forward and backward dependencies and sibling aggregation complementary dependencies between meta-waveform label vector clusters.

[0065] S4. Based on the non-fully connected directed graph network relationship, identify the original waveform.

[0066] It is worth noting that, in order to solve the problems existing in the prior art, this application provides a method to address the above problems from two dimensions.

[0067] At the low-dimensional level, discrete labels are used to annotate discrete meta-waveforms with relatively independent semantics, and feature training is performed for meta-waveforms of the same class. In practical applications, for the meta-waveforms corresponding to a subset of the discrete coordinate points of the light reflection waveform curves of the ODN network, through identification, they can be mapped to structured data that the algorithm can process.

[0068] At the high-dimensional level, multiple labels from the low-dimensional level are used to establish multi-label vector clusters. Subsequently, a non-fully connected directed graph relationship is established between the multi-label vector clusters, and the system automatically stores the above relationship model in the background. At the same time, to address the issue of low overall distinguishability of meta-waveforms with different label types caused by factors such as the impact of limited light footprint monitoring equipment sampling accuracy and optical line attenuation, a more accurate determination of the label type of the sample in the non-fully connected directed graph network is provided, which depends on the forward and backward interconnection relationship. This will not only rely on massive sample data, more complex structural layers, and more time-consuming training of neural networks or intelligent models, but also, by combining the current sample training and the non-fully connected directed graph network relationship between label vector clusters, the accuracy of meta-waveform recognition can be effectively improved, and the resources and costs consumed in training neural networks or intelligent models can be saved.

[0069] The provided layer-by-layer non-fully connected directed graph network type labels and their associated characteristics will provide effective support for the construction of optical split network resource instance objects, the construction of topology connection relationships, and the monitoring of the status of optical split network resources and topology relationships.

[0070] The following section provides a further introduction and explanation of the ODN meta-waveform management method in this application:

[0071] 1. The ODN meta-waveform management method in this application mainly includes: raw data acquisition, label data processing, establishing an intelligent model or neural network to train the raw sample data and output feature sets, and then supporting the type classification of the meta-waveform of newly acquired test data.

[0072] 2. Raw data acquisition: When the test time slot for optical reflection of the ODN network arrives, the network management system will issue a command to acquire two parts of the reflected light waveform of the ODN network connected to a PON port: event data and an unstructured data set containing the coordinates of discrete points of the reflected light waveform curve of the ODN optical splitter network.

[0073] Based on event data, a subset of point coordinate data of meta-waveforms with a single definite semantic is segmented from the unstructured data set of coordinates of discrete points of the reflected light waveform curve; a single definite semantic refers to the independent type classification of meta-waveforms, where one meta-waveform is different from another.

[0074] Subsequently, the subset is normalized within the same planar reference window: damage compensation, scaling, etc.

[0075] 3. Label data processing and feature training:

[0076] First, before labeling the samples, on the graphical user interface, the types of meta-waveforms contained in the subset of the unstructured data set containing the coordinates of discrete points of the curves of reflected light waveforms of a series of different wavelengths are selected and labeled.

[0077] Secondly, the annotation should include the type name of the graphical labeling tool used; and the tool should be used to select, undo, and save the waveform.

[0078] See Figure 5 As shown, the structure of the training network used consists of two channels of convolution plus three fully connected layers. Each channel uses three convolutional layers with no pooling layers, and the convolutional kernels are 3*3. The dimensions of the fully connected layers are 576, 90, 45, and 10. Finally, the softmax function gives ten probability values, representing 10 different types of meta-waveforms.

[0079] The vector cluster of meta-waveform labels of the meta-waveform, the meta-waveform type classification has the following characteristics:

[0080] 1) It describes the characteristics of the elementary waveform of light reflection at a certain topological level in the optical distribution network topology, and this elementary waveform is a part of the complete waveform of light reflection;

[0081] 2) The meta-waveform also describes some abnormal situations, such as the meta-waveform of light reflection caused by the break in the optical fiber of the direct optical splitter.

[0082] 3) Under the main requirements, after the optical splitter at a certain node in the topology is fully equipped with a reflector, the tags of all corresponding wavelengths of the elementary waveforms are formed into a vector cluster in wavelength order; depending on the different levels of the topology nodes, the number of vectors of the elementary waveform tags of the elementary waveforms is different; when the elementary waveform is located at the first split and has a fully equipped reflector, the number of vectors of the elementary waveform tags of the elementary waveform located at that location is equal to the splitting ratio; when the elementary waveform is located at the second split and has a fully equipped reflector, and is a non-cascaded reflector splitter, the number of vectors of the elementary waveform tags of the elementary waveform located at that location is equal to the splitting ratio minus 1;

[0083] 4) The meta-waveform tag is for sample data of the light reflection curve of the same wavelength, which can come from different time points;

[0084] 4. The construction process of the correlated meta-waveform label vector cluster is explained as follows:

[0085] 1) Definition of a cluster of relevance meta-waveform label vectors: A set of meta-waveform labels representing relationships within a non-fully connected directed graph network, mapping the inherent dependencies between samples within the non-fully connected directed graph network. This relationship can be used to effectively construct the topological relationships between nodes in a real ODN network.

[0086] 2) In the construction of the relevance meta-waveform label vector cluster, a construction mode combining guided and self-learning was adopted:

[0087] 2.1) In the annotation area of ​​the system's graphical front-end page, test curves for multiple wavelengths of the same ODN network topology are displayed. The upper left corner is the origin of the coordinates of this area. The horizontal coordinates of the starting point of each curve are aligned, and the horizontal loading and rendering are performed from left to right. Any two curves do not intersect, and all curves are arranged sequentially from top to bottom.

[0088] 2.2) For the waveforms loaded in the labeled areas, perform correlation processing on the meta-waveform label vector clusters in different layers; see [link to relevant documentation]. Figure 4 As shown, the processing involves handling three layers:

[0089] In the first layer, the annotation process of the original waveform corresponding to the original waveform label is performed, and the automatic annotation of waveform data samples of the same wavelength at different time points under the same topology is completed by the system's subsequent inference self-learning.

[0090] In the second layer, the aggregation relationship of multiple wavelength element waveforms is established when the one-way and two-way splitters are fully configured. At the same time, the clustering processing of the element waveform label vector of the wavelength element waveform is also corresponding. The number of aggregations can be multiple or one. When the current stage port reflector has no reflection, the subsequent stage facility entity corresponds to the aggregation of multiple element waveforms, such as the element waveform reflected by the one-way reflector, the element waveform of the two-way reflector without cascade reflection, and the element waveform of the one-way direct-connected ONU without reflection by the previous stage reflector. When the current stage port reflector has reflection, the subsequent stage facility entity corresponds to the aggregation of one element waveform, such as the two-way reflector cascade reflection.

[0091] A non-fully connected directed graph network is constructed in the third layer to support accurate identification of the input meta-waveforms in testing or applications. This network includes two types of relationships: connectivity and complementarity; and any two relationships are mutually exclusive.

[0092] The aforementioned connection relationship means that the reflection from the front-end reflector affects the reflection from the subsequent reflection points (optical splitter, optical splitter port reflector, ONU, fiber break surface, etc.).

[0093] 2.2.1) Marking and processing of the meta-waveform labels of the meta-waveforms in the first layer;

[0094] 2.2.1.0) In the annotation area of ​​the system front-end page, for the test curves of multiple wavelengths of the same ODN network topology, according to the reflection type in the ODN optical splitting topology, select the label tool with the corresponding type name and select the set of curve points of the original waveform in the above curves.

[0095] The waveform types corresponding to the annotations include, but are not limited to:

[0096] One-way reflection, one-way port reflector reflection, one-way port direct-connected ONU reflection, one-way port reflector direct-connected ONU reflection, one-way port direct-connected fiber optic fracture surface reflection, two-way reflection, two-way port direct-connected ONU reflection, two-way port reflector direct-connected ONU reflection, two-way port direct-connected fiber optic fracture surface cascade reflection, two-way port reflector non-cascade reflection, two-way port direct-connected fiber optic fracture surface non-cascade reflection, two-way port reflector cascade reflection.

[0097] 2.2.1.1) In the waveform of the curve, based on the topological hierarchy and distance of the optical splitter planned on the topology, determine whether the element waveform composed of the local waveform of the curve is a single-splitter type element waveform? No, go to 2.2.1.2); Yes, in the toolbox, select the element waveform label rectangle of type single-splitter, select multiple element waveforms in the same column of multiple waveforms from top to bottom, and fill in the label name, such as single-splitter reflection; extract the coordinate set of the points of the element waveform curve, and establish the storage of the data set of the coordinate set of the points of the element waveform curve and the corresponding label name pair;

[0098] 2.2.1.2) In the waveform of the curve, based on the topological hierarchy and distance of the optical splitter reflectors planned on the topology, is the element waveform composed of the local waveforms of the curve a non-cascaded type element waveform of a binary reflector? No, go to 2.2.1.3); Yes, in the toolbox, you can select the elliptical box of the element waveform label of the binary reflector, select multiple element waveforms in the same column of multiple waveforms from top to bottom, fill in the label name, such as non-cascaded reflection of binary reflector; and complete the coordinate set of the points of the element waveform curve, and establish the storage of the data set of the coordinate set of the points of the element waveform curve and the corresponding label name pair;

[0099] 2.2.1.3) In the waveform of the curve, based on the topological hierarchy and distance of the optical splitter reflectors planned on the topology, is the element waveform composed of the local waveform of the curve a single-direct-connected ONU element waveform of a non-cascaded type of two-split reflector? No, go to 2.2.1.4); Yes, in the toolbox, the trapezoidal box of the single-direct-connected ONU element waveform label can be extracted. For multiple element waveforms in the same column of multiple waveforms, select from top to bottom and fill in the label name, such as single-direct-connected ONU reflection; and extract the coordinate set of the points of the element waveform curve, and establish the storage of the set data of the coordinate set of the points of the element waveform curve and the corresponding label name pair;

[0100] 2.2.1.4) In the first layer, construct and store the relationship between the coordinate set of points and corresponding label name pairs of other types of meta-waveform curves included in the waveform of the light reflection curve;

[0101] 2.2.1.5) In the first layer, has the relationship between the coordinate set of points and corresponding signature pairs of all types of meta-waveform curves in the waveform of the light reflection curve been constructed? Yes, go to 2.2.1.6); No, go to 2.2.1.0.

[0102] 2.2.1.6) Based on the coordinates of the selected points on the light reflection curve, the background system continuously performs three inflection point determinations: trough (first inflection point) - peak (second inflection point) - trough (third inflection point), and extracts the set of coordinates of the points on the curve of the original waveform that passes from the first inflection point trough to the third inflection point trough.

[0103] 2.2.1.6) Extract the features of the sample data of the original waveform curves with the same label name. The calculation process is as follows:

[0104] 2.2.1.6.1) Load the set of point coordinates and the set of label pairs of the original waveform curve; this set is recorded as set C;

[0105] The system automatically filters out a sequence S of coordinates of points on the same curve at different times, based on the same topology and wavelength, by loading sampled data of unlabeled light reflection curves.

[0106] Based on the two trough coordinates of each meta-waveform in the existing sample C, after matching the corresponding coordinate positions of the curve in sample S, the meta-waveform is extracted, and the set of meta-waveform curve coordinate sequences for that element is updated; the specific execution process is as follows:

[0107] 1) Is the traversal of set C complete? No, go to 2); Yes, go to 10.

[0108] 2) Obtain an element C0; the maximum ordinate of the point in the original waveform of C0 is Y0;

[0109] 3) Is the traversal of sequence S complete? No, go to 4); Yes, go to 1).

[0110] 4) Obtain an element S0;

[0111] 5) Using the left trough coordinate C01 of C0, match a new trough coordinate S11 in S0, and the deviation between C01 and S11 does not exceed ±5 meters? Yes, go to 6).

[0112] 6) Using the right trough coordinate C02 of C0, match a new trough coordinate S12 in S0, and the deviation between C02 and S12 does not exceed ±5 meters? Yes, go to 7); No, go to 3).

[0113] 7) Extract the original waveform between S11 and S12, and the ordinate value of the highest point of the original waveform is Y1;

[0114] 8) Is the absolute value of the difference between Y0 and Y1 less than 0.3 dB? Yes, go to 9); No, go to 3.

[0115] 9) Extract the element waveforms in S using subscripts S11 and S12, record them as S', and load S' into the set C of element waveform curve coordinate sequences;

[0116] 10) Exit.

[0117] The identification of the original waveform is accomplished using a convolutional neural network algorithm. The main process of the algorithm is as follows:

[0118] A. Initialize network parameters using random numbers

[0119] 1.1 Define the initial values ​​of the weight parameters of the fully connected network from the input layer to the first hidden layer; the input layer has 576 neurons and the first hidden layer has 90 neurons; the initial values ​​are random numbers divided by 1000.

[0120] 1.2 Define the initial values ​​of the weight parameters for the fully connected network from the first hidden layer to the second hidden layer; wherein the first hidden layer has 90 neurons and the second hidden layer has 45 neurons; the initial values ​​are random numbers divided by 100.

[0121] 1.3 Define the initial values ​​of the weight parameters of the fully connected network from the second hidden layer to the third output hidden layer; wherein the second hidden layer has 45 neurons and the third output hidden layer has 10 neurons; the initial values ​​are random numbers divided by 10;

[0122] 1.4. Save the maximum cross-entropy max2 for each training session; max2 = 2; load the number of training iterations m; load the number of labels LN;

[0123] 1.5. Starting from 0, how many training iterations have been completed? If no, proceed to 1.6; if yes, proceed to 1.12.3.

[0124] 1.6 The learning rate is defined as learningrate, where learningrate = (max² / 10.0) 1.7 ;

[0125] 1.7. If learningrate > 0.01? Yes, then learningrate = 0.01; use cross-entropy loss max2 to obtain a smaller learning rate learningrate;

[0126] 1.8 If the current counter is an integer multiple of the total number of labels LN, calculate the training progress percentage, output the cross-entropy loss max2, and the learning rate learningrate;

[0127] 1.9. If the learning rate is less than 0.0000000001, record the network parameters and output them to a file;

[0128] 1.10. Shuffle the order of the samples participating in the learning process;

[0129] 1.11. Is the traversal of all training samples complete? No, proceed to 1.12; Yes, proceed to 1.5;

[0130] 1.12. Define max2 = 0;

[0131] Obtain the sample label value Y;

[0132] 1.12.1 Calculate the forward propagation function by calling the C forward propagation function;

[0133] 1.12.2 Calculate the backpropagation function and call the backpropagation function F.

[0134] 1.12.3 Calculate the maximum cross-entropy in this round to guide the learning rate. Input the sample value Y, call H to calculate the maximum cross-entropy loss in this round, which is used to guide the adjustment of the learning rate;

[0135] B. Activation function;

[0136] It will be called in the C forward propagation function;

[0137] 1.13.1 The number of neurons, n, that carry the hidden layer;

[0138] An array a[] that contains the values ​​of the neurons in the hidden layer;

[0139] Define an array b[] containing the activation function values ​​of the hidden layer;

[0140] 1.13.2. Define a counter j = 0, j < 90? Yes, go to 1.13.3; no, go to 1.13.4;

[0141] 1.13.3、(b[j])=max(a[j],a[j]*0.05); Go to 1.13.2;

[0142] 1.13.4. Exit.

[0143] C. Forward propagation function

[0144] 1) First-layer fully connected output operation

[0145] Input: Number of neurons in the input layer, number of neurons in the current layer, array of input values ​​in the input layer, array of weights for neurons in the first layer;

[0146] Output: The output array of the fully connected neurons in the first layer;

[0147] Call: E performs forward propagation to perform matrix multiplication and calculate the output of the current layer;

[0148] 2) First-level activation output operation

[0149] Inputs: Number of neurons in the first layer, outputs of fully connected neurons in the first layer;

[0150] Output: Array of outputs from the first layer activation function;

[0151] Call: B activation function;

[0152] 3) Second-layer fully connected output operation

[0153] Input: Number of neurons in the first layer, number of neurons in the second layer, output array of the activation function in the first layer, and weight array of neurons in the second layer;

[0154] Output: The output array of the fully connected second-layer neurons;

[0155] Call: E performs forward propagation to perform matrix multiplication and calculate the output of the current layer;

[0156] 4) Second layer activation output operation

[0157] Input: Number of neurons in the second layer, and the output array of the fully connected second-layer neurons;

[0158] Output: The output array of the second-layer activation function;

[0159] Call: B activation function;

[0160] 5) Third-layer fully connected output operation

[0161] Input: The number of neurons in the second layer, the number of neurons in the third layer, the array of output values of the second activation function, the array of weights of the neurons in the third layer;

[0162] Output: The array of fully connected outputs of the third layer;

[0163] 6) Classification output operation of the third layer

[0164] Input: The array of fully connected outputs of the third layer

[0165] Output: The classification result of the third layer;

[0166] [[ID=J15]]Call: D softmax classifier

[0167] D, softmax classifier

[0168] Define a temporary variable p = 0; Load the array of fully connected outputs of the output layer op[]; Classification result array [];

[0169] 1) Define a counter i = 0; i < number of labels? Yes, go to 2); No, go to 3);

[0170] 2) Sum: p += exp(op[i]); Go to 1);

[0171] 3) i = 0; i < number of labels? Yes, go to 4); No, go to 5);

[0172] 4) Classification result array [i] = exp(op[i]) / p; Go to 3);

[0173] 5) Exit.

[0174] E. Perform matrix multiplication in forward propagation to obtain the output of the current layer;

[0175] 1) Load the number of neurons in the previous layer n;

[0176] Load the number of neurons in the current layer p:

[0177] Load the output array a[] of the neurons in the previous layer;

[0178] Load the weight values b[] of the neurons in the current layer;

[0179] 2) Define the output c[] of the neurons in the current layer;

[0180] 3) Define a counter j = 0; j < p? Yes, go to 4); No, go to 7);

[0181] 4) c[i * n + j] = 0;

[0182] 5) Define a counter k = 0; k < n? Yes, go to 6); No, go to 3);

[0183] 6) c[i*n+j]+=a[i*n+k]*b[k*p+j];Go to 5);

[0184] 7) End.

[0185] F. The flow framework of the backpropagation function;

[0186] Loading parameters: sample value y; fully connected neural network output value d; fully connected neural network weight parameter value p; calculate the maximum cross-entropy loss g in this round, which is used to guide the adjustment of the learning rate;

[0187] F.1) Backpropagation from the output layer to the second layer:

[0188] 1.1) A function for calculating the bias terms of the weights of the output layer network parameters;

[0189] Input: Number of neurons in the output layer m; Number of neurons in the second layer n; Activation array a[] in the output layer; Bias array b[] in the second layer;

[0190] Output: The output weight bias term c[] of the second layer;

[0191] Call: F.4, the function that calculates the bias terms of the weights of the output layer network parameters;

[0192] 1.2) Derive the function from the gradient of the output layer to the gradient of the second layer;

[0193] Input: Number of neurons in the second layer m; Number of neurons in the output layer n; Output layer activation array a; Output layer bias term array b

[0194] Output: An array c[] containing the output results of the gradient of the current layer;

[0195] Call: F.5, the function that derives the gradient from the gradient of the next layer to the gradient of the previous layer.

[0196] 1.3) Calculate the activation function of the second layer backpropagation, and the output term is the value of the second layer bias term;

[0197] Input: m, the number of neurons in the second layer; a[], the array containing the neuron values ​​in the second layer; b[], the array containing the neuron values ​​in the second layer;

[0198] An array c[] that holds the activation function values ​​for the second layer.

[0199] Output: An array containing the values ​​of the second-layer activation function;

[0200] Call: G. Backpropagation activation function

[0201] F.2) Backpropagation from the second layer to the first layer:

[0202] 2.1) Function for calculating the bias term of the weights of the second - layer network parameters;

[0203] 2.2) Function for deriving the gradient of the first layer from the gradient of the second layer;

[0204] 2.3) Calculate the activation function for backpropagation of the first layer, and the output term is the value of the bias term of the first layer;

[0205] F.3) Backpropagation from the first layer to the input layer:

[0206] 3.1) Function for calculating the bias term of the weights of the first - layer network parameters;

[0207] 3.2) Function for deriving the gradient of the input layer from the gradient of the first layer;

[0208] 3.3) Calculate the activation function for backpropagation of the input layer, and the output term is the value of the bias term of the input layer;

[0209] F.4) Function for the bias term of the network parameter weights;

[0210] 4.1) Load the number of neurons m in the previous layer; load the number of neurons n in the current layer; load the activation array a[] of the previous layer; load the bias array b[] of the current layer;

[0211] 4.2) Define the output weight bias term c[] of the current layer;

[0212] Define counter i; define counter j;

[0213] 4.3) Traverse counter i starting from 0; i < m? Yes, go to 4.4); No, go to 4.6);

[0214] 4.4) Traverse counter j starting from 0; j < n? Yes, go to 4.5); No, go to 4.3);

[0215] 4.5) c[i * n + j] = a[i] * b[j]; go to 4.4);

[0216] 4.6) Exit.

[0217] F.5) Function for deriving the gradient of the previous layer from the gradient of the next layer

[0218] 5.1) Load the number of neurons m in the current layer; load the number of neurons n in the next layer;

[0219] Load the activation array a of the next layer; load the bias term array b of the next layer;

[0220] 5.2) Define the output result array c of the gradient of the current layer;

[0221] Define counters i, j; <8000464>

[0222] 5.3) Start counting from 0, is i less than m? If yes, go to 5.4); if no, go to 5.7);

[0223] 5.4) c[i] = 0;

[0224] 5.5) Start counting from 0, is j less than n? If yes, go to 5.6); if no, go to 5.3);

[0225] 5.6) c[i] += a[i * n + j] * b[j];

[0226] 5.7) Exit;

[0227] G. The activation function for backpropagation in the current layer.

[0228] 1) Load the number of neurons m in the current layer;

[0229] Load the array a[] of neuron values in the next layer;

[0230] Load the array b[] of neuron values in the next layer;

[0231] Load the array c[] of activation function values in the current layer;

[0232] 2) Define the counter i;

[0233] From 0, start incrementing the counter i. Is i < m? If yes, go to 4); if no, go to 8);

[0234] 4) If a[i] > 0, then go to 5); otherwise, go to 6);

[0235] 5) c[i] = b[i] * 1;

[0236] 6) If a[i] <= 0, then go to 7);

[0237] 7) c[i] = b[i] * 0.05; Go to 3);

[0238] 8) Exit;

[0239] H. Calculate the maximum cross-entropy loss in this round to guide the adjustment of the learning rate.

[0240] 1) Load the classification values a[];

[0241] 2) Load the label values m of the validation set;

[0242] 3) Define the cross-difference quotient loss u;

[0243] 4) The cross-quotient loss u = (-log10(a[m]));

[0244] 5) Obtain the larger quotient loss value for this round and assign it to max2.

[0245] 2.2.2) Labeling processing of the meta-waveform label vector cluster corresponding to the meta-waveform in the second layer;

[0246] Initialize the number of the meta-waveform label vector cluster in the second layer to 0;

[0247] 2.2.2.0) Select multiple samples of the same type of meta-waveform in the first layer of the current view from top to bottom to complete the construction of the meta-waveform label vector cluster for the second layer? No, go to 2.2.2.1); Yes, go to 2.2.2.4.

[0248] 2.2.2.1) Create a single-reflector element wave label vector cluster? No, go to 2.2.2.2); Yes, in the toolbox, select the rectangle that can select the element wave label vector cluster of the single-reflector element waveform, select multiple single-reflector element waveform rectangles from top to bottom, and fill in the name of the element wave label vector cluster type, such as element wave label vector cluster of single-reflector reflection element waveform; increment the number of the element wave label vector cluster in the second layer by 1;

[0249] Save the processed data.

[0250] Go to 2.2.2.2);

[0251] 2.2.2.2) Create a binary reflector element wave label vector cluster? No, go to 2.2.2.3); Yes, in the toolbox, select the rectangle that can select the binary reflector element wave label vector cluster, select multiple binary reflector element wavelets from top to bottom using elliptical boxes, and fill in the name of the element wave label vector cluster type, such as "Element Wavelet Label Vector Cluster of Divider Reflection Element Wavelet"; increment the number of the element wave label vector cluster in the second layer by 1;

[0252] Save the processed data.

[0253] Go to 2.2.2.3);

[0254] 2.2.2.3) Do you want to create a one-way direct-connect ONU waveform label vector cluster? No, go to 2.2.2.4); Yes, in the toolbox, select the rectangle that can select the waveform of the one-way direct-connect ONU reflector, select multiple one-way direct-connect ONU waveforms from top to bottom in the trapezoidal box, and fill in the label name, such as the waveform label vector cluster of the one-way direct-connect ONU reflection waveform; increment the number of the waveform label vector cluster in the second layer by 1;

[0255] Save the processed data.

[0256] Go to version 2.2.2.0);

[0257] 2.2.2.4) At this point, the numerical value of the number of the meta-waveform label vector cluster in the second layer is N; Exit.

[0258] 2.2.2.5) In the second layer, complete the construction of the vector cluster numbering for other types of meta-waveform labels;

[0259] 2.2.3) Construction of the non-fully connected directed graph network relationship of the meta-waveform label vector cluster in the third layer;

[0260] 2.2.3.0) Define a two-dimensional array with indices i and j, respectively, and initialize i = 0 and j = 0;

[0261] Initialize a two-dimensional integer array arc[i][j] representing the edges of a directed graph, with each element of the array initialized to 65535;

[0262] Where i represents the forward node number in the edge of the directed graph, and the number of the meta-waveform label vector cluster of the selected graphic in the second layer is also present.

[0263] j represents the backward node number in the edge of the directed graph, which is the number of the meta-waveform label vector cluster of the selected graphic in the second layer.

[0264] When the direction of the directed graph from j to i holds, and j is not equal to i, the value of arc[i][j] is set to 10, which indicates the backward-to-forward dependency between the vector clusters of the two meta-waveform labels.

[0265] When the direction of the directed graph from j to i holds, and j is not equal to i, the value of arc[i][j] is assigned to 11, which indicates that there is a complementary dependency relationship between the vector clusters of the two meta-waveform labels.

[0266] When the relationship that the direction of the directed graph is from j to i holds, and j equals i, the value of arc[i][j] is assigned to 12, which indicates that there is a type autocorrelation relationship within a single meta-waveform label vector cluster;

[0267] 2.2.3.1) In the third layer, have all directed dependencies between the meta-waveform label vector clusters of all selected graphics in the second layer been constructed? If no, proceed to 2.2.3.2); if yes, proceed to 2.2.3.6).

[0268] 2.2.3.2) In the third layer, select the second layer bounding box graphic of the original waveform label vector cluster, identify it as the forward node, and obtain its number value as i;

[0269] For example, select the rectangular frame of the graphic rectangle of the waveform label vector cluster of the reflector element in the second layer and define it as the forward node;

[0270] Go to 2.2.3.3);

[0271] 2.2.3.3) In the third layer, select the second layer bounding box graphic of another different meta-waveform label vector cluster, identify it as the backward node, and obtain its number value j;

[0272] For example, select a rectangular box of the meta-waveform label vector cluster reflected by a dichotomous reflector in the second layer;

[0273] Go to 2.2.3.4);

[0274] 2.2.3.4) Establish a directed connection from backward node j to forward node i, which serves as an edge in a non-fully connected directed graph; when there is a forward-backward dependency between two vector clusters of meta-waveform labels, it is recorded as arc[i][j] = 10; when there is a sibling complementary dependency between two vector clusters of meta-waveform labels, it is recorded as arc[i][j] = 11; when there is a type autocorrelation within a single vector cluster of meta-waveform labels, it is recorded as arc[i][i] = 12.

[0275] Go to 2.2.3.5);

[0276] 2.2.3.5) Save the connection relationships between nodes in the backward-to-forward non-fully connected directed graph between the above-mentioned meta-waveform label vector clusters, arc[i][j]; go to 2.2.3.1);

[0277] 2.2.3.6) Exit;

[0278] The final non-fully connected directed graph of the meta-waveform label vector cluster can be found in [reference needed]. Figure 6 As shown.

[0279] 3) In the construction of the meta-waveform label vector cluster with high cohesion and correlation, the attribute model of the meta-waveform with the following attribute items is adopted:

[0280] ① Meta waveform tag type;

[0281] ② The coordinate set of the elementary waveform points;

[0282] ③ The feature set after training the meta-waveform;

[0283] ④ The name of the type of the meta-waveform label vector cluster;

[0284] ⑤ Meta-waveform label vector cluster number; initial value 0; globally unique number, and cannot be repeated;

[0285] 4) In the construction of the above-mentioned meta-waveform label vector clusters with high cohesion correlation, a sequence list of the attribute model of the meta-waveform is included, such as Figure 7 The left rectangular box shown represents the attribute models of the meta-waveforms of λ1, λ2, λ3, and λ4 arranged in order at the same position on the same topology;

[0286] 5) Construct a hash table mapping. The primary key of this hash table is the meta-waveform label type name, and the value is the number of the meta-waveform label vector cluster; expressed as map_ml_cl<meta-waveform label type name, meta-waveform label vector cluster number>;

[0287] The meta-waveform matches the meta-waveform label type, and outputs a hash table mmap_wl_st with multiple primary keys. Its primary key is the wavelength, and the value is a map. The primary key of this map is the distance, and the value is a vector. The elements of the vector are a structure, and the attribute items of this structure include: meta-waveform label type, meta-waveform label vector cluster number, percentage; expressed as mmap_wl_st<wavelength, map<distance, vector<struct: meta-waveform label type, meta-waveform label vector cluster number, percentage>>>;

[0288] Among them, the map distance refers to a classification of the relationship between meta-label vector clusters: when there is a forward and backward dependence relationship between two meta-label vector clusters, it is recorded as 10; when there is a same-level complementary dependence relationship between two meta-label vector clusters, it is recorded as 11; when there is a type self-correlation relationship between two meta-label vector clusters, it is recorded as 12; when there is no mutual relationship between two meta-label vector clusters, it is recorded as 65535.

[0289] The basic process of constructing this data model is as follows: Obtain the wavelength, distance, and curve coordinate set of a meta-waveform, and match the sample feature values corresponding to this wavelength trained through this curve coordinate set to obtain at least two matching percentages; and record them in mmap_wl_st<wavelength, map<distance, vector<struct: meta-waveform label type, meta-waveform label vector cluster number, percentage>>> in sequence;

[0290] 6) Application of the label vector cluster;

[0291] 6.1.1) The forward-dependent node path of the current vertex;

[0292] Data structure: Hash table, the primary key is the current vertex, and the value is the sequence of the nodes that the current vertex depends on forward; these nodes are arranged in order from near to far; expressed as map_v_vec;

[0293] The vertex refers to the number of the meta-label vector cluster, and its semantics is the naming of the meta-label vector cluster;

[0294] 6.1.2) The sibling aggregation dependency relationship on the node path of the forward dependency of the current vertex;

[0295] Data structure: Hash table, with the primary key being the node with the cascading reflection feature and the value being the node at the same level with the non-cascading reflection feature; expressed as map_v_cv;

[0296] 6.2) The generation method for the node path of the forward dependency of the current vertex and the sibling aggregation dependency relationship on the node path of the forward dependency of the current vertex is as follows;

[0297] 6.2.1) Load the two-dimensional array arc[i][j]; where i == j == 0, and the maximum value of i and j is M; define the vector sequence VEC; define the counter m = 0; z = m;

[0298] 6.2.2) m < M? Yes, go to 6.2.3); No, go to 6.2.14);

[0299] 6.2.3) Obtain the new one-dimensional array arc[][z];

[0300] 6.2.4) Define the counter p = 0; p < i;

[0301] 6.2.5) Has the traversal of arc[p][z] ended? No, go to 6.2.6); Yes, go to 6.2.12);

[0302] 6.2.6) arc[p][z] == 10, or arc[p][z] == 12? Yes, go to 6.2.7); No, go to 6.2.9);

[0303] 6.2.7) Add p to VEC;

[0304] 6.2.8) z = p; go to 6.2.3);

[0305] 6.2.9) arc[p][z] == 11? Yes, go to 6.2.10); No, go to 6.2.11);

[0306] 6.2.10) Add the pair {z, p} to map_v_cv;

[0307] 6.2.11) arc[p][z] == 65535? Yes, go to 6.2.12); No, output an error, go to 6.2.12.2);

[0308] 6.2.12) Result recording;

[0309] 6.2.12.1) Combine m and VEC into a pair and insert it into map_v_vec;

[0310] Clear the vector sequence VEC;

[0311] 6.2.12.2) m = m + 1, and the next vertex will be traversed;

[0312] 6.2.13) z = m; Go to 6.2.2);

[0313] 6.2.14) End;

[0314] 7) See Figure 8 as shown, the process of obtaining the weakly connected paths of a non-fully connected directed graph related to a meta-waveform;

[0315] Load variables;

[0316] Input variables:

[0317] mmap_wl_st<wavelength, map<distance, vector<struct: meta-waveform label type, meta-waveform label vector cluster number, percentage>>>;

[0318] The forward-dependent node path of the current vertex, data structure: hash table, the primary key is the current vertex, and the value is the sequence of the forward-dependent nodes of the current vertex. The nodes in this sequence are arranged in order from near to far, expressed as map_v_vec;

[0319] The same-level aggregation dependence relationship on the forward-dependent node path of the current vertex; data structure: hash table, the primary key is the node with the cascading reflection feature, and the value is the node with the non-cascading reflection feature at the same level; expressed as map_v_cv;

[0320] Through the percentage, meta-waveform label type, and meta-waveform label vector cluster number of the meta-waveform label vector cluster of the current input waveform; using the meta-waveform label vector cluster number as the primary key, search map_v_vec to obtain the sequence v_vec of the upstream meta-waveform label vector cluster numbers related to this number;

[0321] Through each element in v_vec, search map_v_cv to obtain the related meta-waveform label vector cluster numbers at the same level;

[0322] Finally, through the meta-waveform label vector cluster numbers of the current input waveform, match the elements in the sequence of the upstream forward-dependent nodes of v_vec one by one; and, for cascaded elements, query map_v_cv to obtain the related meta-waveform label vector cluster numbers at the same level and match the meta-waveform label vector cluster numbers of other meta-waveforms of the current input waveform;

[0323] When the data at both of the above levels are successfully matched, the type of the meta-waveform label vector cluster of the current input waveform will be determined, thus achieving the goal of accurate positioning.

[0324] For example, for Figure 1 The original waveforms 6 and 13 are identified. The original waveform 6 is initially discretized and characterized as a non-cascaded reflection of the fiber optic fracture surface. The probability of non-cascaded reflection of the two-way port reflector is 49%, and its number is 4. It is also identified as a cascaded reflection of the fiber optic fracture surface. The probability of cascaded reflection of the two-way port reflector is 51%.

[0325] Two primary waveforms cannot be independently determined as a certain primary waveform based on their morphology.

[0326] By looking up the label vector cluster number C of the meta-waveform 6 in map_v_vec, we can obtain the corresponding set of sequences of autocorrelation or uplink meta-waveform label vector clusters.

[0327] By searching each element in v_vec, if two elements are the same and also have the same index C, then the two elements are autocorrelated.

[0328] Then, by querying the nearest upstream meta-waveform N1 through meta-waveform 6, the first probabilistic meta-waveform label corresponding to N1 and the number C1 of the meta-waveform label vector cluster are found through this meta-waveform N1.

[0329] The nearest downstream meta-waveform N2 is found by querying meta-waveform 6; the number C2 of the first probabilistic meta-waveform tag vector cluster corresponding to N2 is found by searching through meta-waveform N2.

[0330] If C0 == C1 and C1 == C2, then the three elemental waveforms mentioned above are autocorrelated; for example Figure 2 In the diagram, the relationship between waveforms 5, 6, and 7 is discussed.

[0331] Traverse the upstream elementary waveforms of N1 and determine whether the vector cluster number of the elementary waveform is autocorrelated.

[0332] Relationship; such as Figure 2 In the diagram, the relationship between the waveforms 5, 3, and 2 is discussed.

[0333] Similarly, whether the vector cluster numbering of the downstream meta-waveform of N2 is autocorrelation;

[0334] In the matching of vector cluster numbers of the upstream and downstream meta-waveforms of meta-waveform 6, a complementary relationship was found between elements in vec; such as Figure 2 In the diagram, the relationship between the primary waveform 4 and the primary waveforms 2, 3, 5, 6, 7, 8, and 9 is discussed.

[0335] In the matching of vector cluster numbers of the upstream and downstream meta-waveforms of meta-waveform 4, a forward dependency relationship corresponding to an element in vec was found; such as Figure 2 In the diagram, the relationship between the original waveform 4 and the original waveform 1 is discussed.

[0336] The key logic of the judgment process in this scenario is as follows:

[0337] The meta-waveform tag vector cluster number of meta-waveforms 6, 7, 8, 9 and meta-waveforms 5, 3, 2 is “2. Non-cascaded reflection of the fracture surface of the two-part fiber, and cascaded reflection of the two-part port reflector”.

[0338] The 4-element waveform label vector cluster is numbered as “3-division fiber fracture surface cascade reflection, and division port reflector cascade reflection”;

[0339] The 1-element waveform label vector cluster is numbered "1. Two-port reflector cascaded reflection";

[0340] Thus, the type and relationship of a set of element waveforms with correlations in multiple non-fully connected directed graphs can be determined step by step with greater accuracy.

[0341] In summary, the ODN meta-waveform management method in this application, at a low-dimensional level, performs discrete labeling of meta-waveforms with relatively independent semantics and completes feature training for the same type of meta-waveforms. In application, for the meta-waveforms corresponding to a subset of the discrete coordinate points of the light reflection waveform curves of the ODN network, identification can map them to structured data that the algorithm can process; by constructing a convolutional neural network, meta-waveform features can be flexibly extracted; and the limitations of judgment based on fixed threshold features are avoided.

[0342] At the high-dimensional level, multiple labels from the low-dimensional level are used to establish multi-label vector clusters; non-fully connected directed graph relationships are established between multi-label vector clusters; thus, a more accurate determination of the label type corresponding to the sample in the non-fully connected directed graph network is provided, which depends on the forward and backward interconnection relationships. This can effectively improve the accuracy of meta-waveform recognition and save the resources and costs consumed in training neural networks or intelligent models.

[0343] Furthermore, the provided layer-specific non-fully connected directed graph network type labels and their associated characteristics will provide effective support for the construction of optical split network resource instance objects, the construction of topology connection relationships, and the monitoring of the status of optical split network resources and topology relationships.

[0344] Secondly, embodiments of this application also provide an ODN primitive waveform management device, which includes an acquisition module, a construction module, and an identification module.

[0345] The acquisition module is used to acquire the coordinates of discrete points of the reflected light waveform curves of different wavelengths in the optical distribution network (ODN) and segment out the coordinate set of points with a single type of meta-waveform.

[0346] The building module is used to construct the coordinate set of points of all types of meta-waveforms and the relationship between the corresponding meta-waveform label names, and to extract the sample data features of meta-waveform curves with the same meta-waveform label name.

[0347] The construction module also constructs a meta-waveform tag vector cluster based on meta-waveforms with the same meta-waveform tag name, and constructs a non-fully connected directed graph network relationship between multiple meta-waveform tag vector clusters. The non-fully connected directed graph network relationship includes forward and backward dependencies and sibling aggregation complementary dependencies between meta-waveform tag vector clusters.

[0348] The identification module identifies the original waveform based on the non-fully connected directed graph network relationship.

[0349] Furthermore, in one embodiment, the identification module identifies the meta-waveform based on the non-fully connected directed graph network relationship, including:

[0350] Based on the non-fully connected directed graph network relationship, the meta-waveform label vector cluster number of the current input waveform is searched to obtain the sequence of uplink meta-waveform label vector cluster numbers that have a forward-backward dependency relationship with the meta-waveform label vector cluster number.

[0351] Based on each element of the sequence of uplink meta-waveform tag vector cluster numbers, a search is performed again based on the non-fully connected directed graph network relationship to obtain the number of meta-waveform tag vector clusters that have a homogeneous and complementary dependency relationship with the uplink meta-waveform tag vector clusters.

[0352] Based on the sequence of uplink meta-waveform label vector cluster numbers and the numbers of meta-waveform label vector clusters with complementary dependencies at the same level, the meta-waveform label vector cluster numbers of the current input waveform are matched.

[0353] When a match is completed, determine the meta-waveform label type of the meta-waveform label vector cluster of the current input waveform.

[0354] Furthermore, in one embodiment, the forward-backward dependency relationship indicates that the distribution shape of the subsequent waveform depends on the distribution position and shape of the preceding waveform; the homogeneous aggregation complementary dependency relationship indicates the aggregation of waveforms at the same reflection position in different scenarios.

[0355] Furthermore, in one embodiment, the type of the meta-waveform tag includes:

[0356] One-way reflection, one-way port reflector reflection, one-way port direct-connected ONU reflection, one-way port reflector direct-connected ONU reflection, one-way port direct-connected fiber fracture surface reflection, two-way reflection, two-way port direct-connected ONU reflection, two-way port reflector direct-connected ONU reflection, two-way port direct-connected fiber fracture surface cascade reflection, two-way port reflector non-cascade reflection, two-way port direct-connected fiber fracture surface non-cascade reflection, and two-way port reflector cascade reflection.

[0357] Furthermore, in one embodiment, after segmenting the coordinate set of points having a single type of meta-waveform, the acquisition module is further configured to:

[0358] The step of normalizing the coordinate set of points of the original waveform includes loss compensation and scaling.

[0359] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for managing ODN primitive waveforms, characterized in that, The method for managing the ODN meta-waveform includes: The coordinates of discrete points of the reflected light waveforms of different wavelengths in the optical distribution network (ODN) are collected, and the coordinate set of points with a single type of meta-waveform is segmented out. Construct the coordinate set of points for all types of meta-waveforms and the relationship between the corresponding meta-waveform label names, and extract the sample data features of meta-waveform curves with the same meta-waveform label name; Meta-waveform label vector clusters are constructed based on meta-waveforms with the same meta-waveform label name, and non-fully connected directed graph network relationships are constructed between multiple meta-waveform label vector clusters. The non-fully connected directed graph network relationships include forward and backward dependencies and sibling aggregation complementary dependencies between meta-waveform label vector clusters. Based on the relationships of the non-fully connected directed graph network, the meta-waveform is identified; The identification of the meta-waveform based on the non-fully connected directed graph network relationship includes: Based on the non-fully connected directed graph network relationship, the meta-waveform label vector cluster number of the current input waveform is searched to obtain the sequence of uplink meta-waveform label vector cluster numbers that have a forward-backward dependency relationship with the meta-waveform label vector cluster number. Based on each element of the sequence of uplink meta-waveform tag vector cluster numbers, a search is performed again based on the non-fully connected directed graph network relationship to obtain the number of meta-waveform tag vector clusters that have a homogeneous and complementary dependency relationship with the uplink meta-waveform tag vector clusters. Based on the sequence of uplink meta-waveform label vector cluster numbers and the numbers of meta-waveform label vector clusters with complementary dependencies at the same level, the meta-waveform label vector cluster numbers of the current input waveform are matched. When a match is completed, determine the meta-waveform label type of the meta-waveform label vector cluster of the current input waveform.

2. The ODN meta-waveform management method as described in claim 1, characterized in that: The forward-backward dependency relationship indicates that the distribution shape of the subsequent waveform depends on the distribution position and shape of the preceding waveform. The complementary dependency relationship of the same level aggregation represents the aggregation of meta-waveforms at the same reflection position in different scenarios.

3. The ODN meta-waveform management method as described in claim 1, characterized in that, The types of the meta-waveform labels include: One-way reflection, one-way port reflector reflection, one-way port direct-connected ONU reflection, one-way port reflector direct-connected ONU reflection, one-way port direct-connected fiber fracture surface reflection, two-way reflection, two-way port direct-connected ONU reflection, two-way port reflector direct-connected ONU reflection, two-way port direct-connected fiber fracture surface cascade reflection, two-way port reflector non-cascade reflection, two-way port direct-connected fiber fracture surface non-cascade reflection, and two-way port reflector cascade reflection.

4. The ODN meta-waveform management method as described in claim 1, characterized in that, After segmenting the coordinate set of points having a single type of primitive waveform, the method further includes: The step of normalizing the coordinate set of points of the original waveform includes loss compensation and scaling.

5. A management device for ODN element waveforms, characterized in that, The ODN meta-waveform management device includes: The acquisition module is used to acquire the coordinates of discrete points of the reflected light waveform curves of different wavelengths in the optical distribution network (ODN) and segment out the coordinate set of points with a single type of meta-waveform. The module is used to construct the coordinate set of points of all types of meta-waveforms and the relationship between the corresponding meta-waveform label names, and to extract the sample data features of meta-waveform curves with the same meta-waveform label name. The construction module also constructs a meta-waveform tag vector cluster based on meta-waveforms with the same meta-waveform tag name, and constructs a non-fully connected directed graph network relationship between multiple meta-waveform tag vector clusters. The non-fully connected directed graph network relationship includes forward and backward dependencies and sibling aggregation complementary dependencies between meta-waveform tag vector clusters. The identification module identifies the original waveform based on the relationships of the non-fully connected directed graph network; The identification module identifies the original waveform based on the non-fully connected directed graph network relationship, including: Based on the non-fully connected directed graph network relationship, the meta-waveform label vector cluster number of the current input waveform is searched to obtain the sequence of uplink meta-waveform label vector cluster numbers that have a forward-backward dependency relationship with the meta-waveform label vector cluster number. Based on each element of the sequence of uplink meta-waveform tag vector cluster numbers, a search is performed again based on the non-fully connected directed graph network relationship to obtain the number of meta-waveform tag vector clusters that have a homogeneous and complementary dependency relationship with the uplink meta-waveform tag vector clusters. Based on the sequence of uplink meta-waveform label vector cluster numbers and the numbers of meta-waveform label vector clusters with complementary dependencies at the same level, the meta-waveform label vector cluster numbers of the current input waveform are matched. When a match is completed, determine the meta-waveform label type of the meta-waveform label vector cluster of the current input waveform.

6. The ODN element waveform management device as described in claim 5, characterized in that: The forward-backward dependency relationship indicates that the distribution shape of the subsequent waveform depends on the distribution position and shape of the preceding waveform. The complementary dependency relationship of the same level aggregation represents the aggregation of meta-waveforms at the same reflection position in different scenarios.

7. The ODN element waveform management device as described in claim 5, characterized in that, The types of the meta-waveform labels include: One-way reflection, one-way port reflector reflection, one-way port direct-connected ONU reflection, one-way port reflector direct-connected ONU reflection, one-way port direct-connected fiber fracture surface reflection, two-way reflection, two-way port direct-connected ONU reflection, two-way port reflector direct-connected ONU reflection, two-way port direct-connected fiber fracture surface cascade reflection, two-way port reflector non-cascade reflection, two-way port direct-connected fiber fracture surface non-cascade reflection, and two-way port reflector cascade reflection.

8. The ODN element waveform management device as described in claim 5, characterized in that, After segmenting the coordinate set of points with a single type of primitive waveform, the acquisition module is further configured to: The step of normalizing the coordinate set of points of the original waveform includes loss compensation and scaling.

Images