A remote sensing process recommendation method based on logical structure and attribute characteristics

Abstract

The present application relates to the technical field of data mining, and particularly relates to a remote sensing process recommendation method based on logical structure and attribute characteristics. Sim(r) ; part of the attributes in the attribute set is taken, and similarity calculation is performed with the feature set of other processes in the process library to generate M Sim(f) ; M Sim(r) is multiplied with M Sim(f) to generate top n recommended processes with the highest similarity to the process to be recommended, an algorithm module with a recommended position of length (alpha) + 1 in the recommended process is taken as a recommended result, a starting step is returned, and the process is repeated until a complete remote sensing process is constructed. The present application can quickly and accurately construct a remote sensing process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data mining, and particularly relates to a remote sensing process recommendation method based on logical structure and attribute characteristics. BACKGROUND

[0002] Data mining refers to a process of searching for information hidden in a large amount of data through an algorithm. As an application direction of data mining, a recommendation system has been widely applied in various industries in recent years. In the field of remote sensing, a large amount of process structure information and attribute information is converted into remote sensing process characteristics, and data relationships between processes in a process library are mined, so that the effect of process recommendation is achieved.

[0003] A remote sensing workflow is usually used to describe the execution order between algorithms and the logical relationship between algorithms. The entire process data is stored in a text file, including algorithm, parameter and other data. A complete remote sensing workflow refers to a process that has a start and end marker, has not less than one algorithm unit when constructing the process, has complete arrow direction and parameter information, and can finally generate correct results.

[0004] A Task-based Process Structure Tree (TPST) is a tree structure with logical nodes generated according to a process model, wherein each leaf node corresponds to a task node in the process model, and a non-leaf node corresponds to a logical node. The structure includes Sequence, And and Xor. SUMMARY

[0005] In order to solve the above technical problems, the purpose of the present application is to provide a remote sensing process recommendation method based on logical structure and attribute characteristics, and the technical solution adopted is as follows:

[0006] A: First, select the required process components and algorithm modules, fill in the parameters, link the algorithm modules with arrows, generate a complete process model, and generate a process structure tree based on logical units from the complete process model. Then generate an incomplete process model as a process to be recommended;

[0007] B: Use the process structure tree to generate a path table containing logical structure, collect process parameter information and characteristic attributes, and construct an attribute set F=(T R ,L I ,L P ,E) of all features of the process object.

[0008] Wherein, F is an attribute set; T R is a path table set generated according to the process structure tree; L IL is a set of process attributes contained in the process construction process; L P E is a control table; E is a set of algorithm module parameters in the process construction process; E

[0009] C: convert all processes in the process library into a path table; take the path table of the process to be recommended and the path table in the process library for similarity matching to generate a similarity matrix M based on the logical structure Sim(r) ;

[0010] D: obtain L I and L P in step B, and perform similarity calculation with the feature set of other processes in the process library to generate a similarity matrix M based on the feature attributes Sim(f) ;

[0011] E: multiply the similarity matrix M based on the logical structure M Sim(r) and the similarity matrix M based on the feature attributes M Sim(f) to obtain a new similarity matrix, generate Top n recommended processes with the highest similarity to the process to be recommended, Top n is greater than 1; obtain the algorithm module at the recommended position length(α)+1 in the recommended process as the recommended result, return to step A, repeat the process until a complete remote sensing process is constructed.

[0012] Preferably, the step A comprises the following specific steps:

[0013] A1: build process components and algorithm modules on the process construction platform through component dragging, and select logical structure modules such as sequence, parallel, and condition to connect the processes in series, and finally generate a complete process model guided by arrows;

[0014] A2: convert the process model into a process structure tree using the Petri net structure model.

[0015] Preferably, the step B comprises the following specific steps:

[0016] B1: obtain the process structure tree set, perform pre-order traversal on the structure tree, and first observe whether the root node is a logical node;

[0017] B2: when the root node is a logical node, observe the child nodes of the root node; if there is a logical node in the node at this layer, store the logical node identifier in the path table, and divide the process segment connected by the logical node using the start marker and the end marker, and store the start marker and the end marker of the logical node in the path table; then traverse the next layer of nodes, store the start marker and the end marker of the logical node in the path table, and continue until the deepest layer of the structure tree is reached;

[0018] B3: after reaching the deepest layer of the structure tree, traversing from the right child node of the parent node, when the right child node of the parent node exists, the right child node is in the same layer as the node; when the right child node of the parent node does not exist, the task nodes connected by the parent node have been connected completely, and the logic segment is exited; in this way, if the layer number of the node is 2 less than the layer number of the previous node, the two logic segments are exited, and the storage position of the task unit is after the two end marks;

[0019] B4: the path table stores the logical relationship in the process, the process attributes and algorithm attributes in the process construction process are stored in the form of task log text, and the data are represented by symbols and codes; the correspondence table stores the actual meaning of the corresponding symbols and codes;

[0020] Preferably, step C comprises the following specific steps:

[0021] C1: repeating the processes of steps B1 to B4, converting all process models in the process library and the process to be recommended into path tables, and obtaining a path table set with different lengths;

[0022] C2: performing similarity calculation by using a longest common subsequence method based on a logical structure, performing sequence matching based on LSL between the process to be recommended and all processes in the process library, and obtaining the longest common subsequence;

[0023] C3: multiplying the length value of the obtained longest common subsequence by 2 as a numerator, and multiplying the sum of the lengths of the two sequences participating in matching by 2 as a denominator, to obtain the similarity Sim(r):

[0024]

[0025] Wherein, Sim(r) represents the similarity, LSL(r1, r2) represents the longest common subsequence of the logical structure contained in the array r1 and the array r2, |r1| represents the number of elements in the array r1, and |r2| represents the number of elements in the array r2;

[0026] Then, similarity calculation is performed with the processes in the process library respectively, to obtain a similarity matrix M based on a logical structure Sim(r) .

[0027] Preferably, step D comprises the following specific steps:

[0028] D1: taking the process attribute set L I , encapsulating the attributes in the text description file of the corresponding process, one process corresponding to one text description file; the attribute characteristics in the process construction are stored in the form of nodes in the text file;

[0029] D2: taking the process parameter set L P, one flow corresponds to one attribute set, multiple algorithms, and one algorithm corresponds to multiple parameter sets; the data of each parameter set is stored in the text description file of the corresponding attribute set, and the character type is floating point type. At this time, the feature set L C of each flow contains the data of the attribute set and the parameter set;

[0030] D3: the feature set L C contains attribute features γ and parameter features i, and each of the two features is assigned a weight of one-half; the feature set of the process to be recommended is matched with the attribute features γ of each process in the process library for similarity, and a binary method is adopted. If the character strings are consistent, it indicates a successful match. If the character strings are inconsistent, it indicates a failed match. The quotient of the number of successful matches and the total number of matches is obtained to obtain the attribute feature similarity matrix M Sim(γ) ;

[0031] D4: the feature set of the process to be recommended is matched with the parameter features i of each process in the process library for similarity. First, the algorithm set L A of the process to be recommended is matched with each algorithm in the algorithm set L' A of the other processes. If it is the same type of algorithm, the number of parameters and the meaning of the parameters are consistent. Otherwise, there is no matching significance. The intersection L A ∩L' A is obtained, and the parameters in the algorithms in the intersection are matched. A binary method is adopted for matching. If the floating point values are consistent, it indicates a successful match. The quotient of the number of successful matches and the total number of matches is taken as the parameter feature similarity result. After all calculations are completed, the parameter feature similarity matrix M Sim(i) is obtained.

[0032] D5: according to the weight assigned in step D3, the values of M Sim(γ) and M Sim(i) are added after being multiplied by their respective weights. The formula is as follows:

[0033]

[0034] wherein, U1 represents the number of elements in the user features that match each other, U0 represents the number of elements in the user features that do not match each other, I1 represents the number of elements in the algorithm features that match each other, and I0 represents the number of elements in the algorithm features that do not match each other.

[0035] Then, the similarity matrix M Sim(f) based on the feature attributes is obtained.

[0036] Preferably, step E includes the following specific steps:

[0037] E1: the similarity matrix M Sim(r) based on the logical structure is multiplied by the similarity matrix M Sim(f)Numerical multiplication, the formula is as follows:

[0038]

[0039] Wherein, |r1| represents the number of elements in the array r1; |r2| represents the number of elements in the array r2;

[0040] Then the similarity value of the recommended process and other processes in the process library is obtained;

[0041] E2: Obtain the process construction length length (alpha) of the process to be recommended, and the recommended position length (alpha) +1;

[0042] E3: According to the similarity value, generate the top n Recommended processes with the highest similarity to the process to be recommended, and the algorithm of the recommended position length (alpha) +1 in the process is recommended to the user as the final recommended algorithm.

[0043] The embodiments of the application have at least the following beneficial effects:

[0044] The method provided by the application can mine the potential relationship between each process in the process library, especially the structural relationship and attribute relationship. According to the related data analysis, the maximum expected value of the next activity of the process to be constructed is obtained, and the process structure and algorithm module are recommended, so as to help the user to quickly and accurately construct the remote sensing process. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0046] Figure 1 The flowchart of the remote sensing process recommendation method based on logical structure and attribute characteristics of the application;

[0047] Figure 2 The schematic diagram of the multi-branch structure tree of a certain process;

[0048] Figure 3 The schematic diagram of the structure tree of a certain process;

[0049] Figure 4 The storage format of the process feature set L C . DETAILED DESCRIPTION

[0050] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive objectives, the specific implementation, structure, features and effects of the scheme proposed according to the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0052] Please refer to Figure 1 which shows a flowchart of a remote sensing process recommendation method based on logical structure and attribute characteristics provided by one embodiment of the present application, the method comprising the following steps:

[0053] A: First, select the required process components and algorithm modules, fill in the parameters, link each algorithm module with arrows, generate a complete process model, and generate a process structure tree based on logical units from the complete process model; then generate an incomplete process model as a process to be recommended.

[0054] Step A comprises the following specific steps:

[0055] A1: On the process construction platform, build the process algorithm module by component dragging, and select sequential, parallel, conditional and other logical structure modules to connect the process in series, and finally generate a complete process model directed by arrows.

[0056] Specifically, by logging into the satellite simulation platform, the process construction module builds the process algorithm module by component dragging, and selects sequential, parallel, conditional and other logical structure modules and arrow directions to connect the process in series, and generates a complete process model directed by arrows.

[0057] A2: Convert the process model into a process structure tree using the Petri net structure model.

[0058] Specifically, using the Petri net structure model P=(P, T, F, W), P represents the task nodes in the process; T represents the logical nodes in the process; F represents the directed arcs between the task nodes and the logical nodes; W represents the dynamic objects of moving a task node to another task node in the process. Convert the process model into a multi-fork structure tree, with the root node being the mark represented by the sequential structure, the first layer of leaf nodes being the task units and the logical units, and then the logical units branching down to represent the task units contained in the logic. For example, the task units A and B in the parallel unit AND represent that A and B are in parallel structure, and the process continues until all leaf nodes in the structure tree are task nodes; create a structure tree set M.

[0059] B: generating a path table containing logical structure by using the process structure tree, collecting process parameter information and characteristic attributes, and constructing an attribute set F=(T R ,L I ,L P ,E) of all characteristics of the process object; wherein F is an attribute set; T R is a path table set generated according to the process structure tree; the structural relationship in the path table is represented by a symbol, and the algorithm module is represented by a code, and the objects represented by the symbol and the code are stored in a reference table E; L I is a set of process attributes contained in the process construction process; for example, a user writing direction and a suitable satellite, etc.

[0060] L P is a set of algorithm module parameters in the process construction process; E is a reference table; the reference table stores the true meaning represented by the symbol and the code in the path table.

[0061] The step B includes the following specific steps:

[0062] B1: obtaining a set of process structure trees, performing a pre-order traversal on the structure trees, and first observing whether the root node is a logical node;

[0063] It should be noted that if the root node is a task node, there is no subsequent child node, because logical nodes are needed to connect between task nodes, at this time, there is only one task node in the process model, and the process cannot be formed. Therefore, the root node can only be a logical node.

[0064] B2: when the root node is a logical node, the child nodes of the root node are observed; if there is a logical node in the node at this level, the identifier of the logical node is stored in the path table, and the start mark and the end mark are used to divide the process segment connected by the logical node. The child nodes under the logical node are the process segments connected by the logical node, so all the logical nodes are not leaf nodes, at this time, the start mark and the end mark of the logical node are stored in the path table; then the next layer of nodes is traversed, the start mark and the end mark of the logical node are stored in the path table, until the deepest layer of the structure tree is traversed;

[0065] B3: After reaching the deepest layer of the structure tree, traverse from the right child node of the parent node. When the right child node of the parent node exists, the right child node is in the same layer as the node, and the two nodes are concatenated by the logical node of the parent node. When the right child node of the parent node does not exist, the task node concatenated by the parent node has been concatenated completely, and the logical segment is exited. That is, an end marker is inserted at this position, indicating that the flow segment has been constructed completely. Since the end marker occupies one storage space, the storage position of the subsequent task unit needs to be moved one position backward. By analogy, if the layer number of the node is two less than the layer number of the previous node, two logical segments are exited, and the storage position of the corresponding task unit is after the two end markers.

[0066] B4: The path table stores the logical relationship in the flow, and the flow attributes and algorithm attributes in the flow construction process are stored in the task log text format, and the data are represented by symbols and codes. The correspondence table stores the actual meaning of the corresponding symbols and codes.

[0067] Path table T R Generation: Preorder traversal is adopted for storage. Simply, if the node is a leaf node, it is stored normally. If the node is a logical node, the start marker and the end marker of the logical node are stored in the corresponding positions at the same time. For example, Figure 2 The multi-ary structure tree of a certain flow shown in the figure, the final storage order of the nodes is: S-start, a1, A-start, a3, a4, A-end, a2, S-end.

[0068] Specifically, taking Figure 2 as an example, the specific process of step B is described:

[0069] 1: Define the flow structure tree set M; the root node of the flow structure tree P is P0, and the non-root node is P n (n>0), establish the two-dimensional path table R and the layer number table L.

[0070] 2: Scan M i , determine the node number n and the layer number table L i (L1, L2…L n ); perform preorder traversal;

[0071] 3: Conditional judgment: if P0 is a logical node "X" (Sequence, And or or), store "X-Start" and "X-End" in the two-dimensional path table R i [i][0] and R i [i][1] storage unit. Otherwise (P0 is a non-logical unit), the root node is a task node, indicating that there is only one task node in the flow, and the flow cannot be constructed.

[0072] 4: If Pn is a child node, i.e., P n is a logical node "X", and insert "X-Start" and "X-End" into the path table R i [i][n] and R i in [i][n + 1] storage units; if P n is a leaf node, i.e., P n is a task node, then observe its layer table L. If ln ≥ ln-1, it means that the structure tree continues to traverse downward. At this time, insert P n into R i [i][n]. If ln < ln-1, it means that the structure tree has traversed to the deepest layer and needs to return to the parent node to traverse its right child node. Then insert P n into Ri[i][n + ln-1 - ln];

[0073] Step 5: Continue to traverse the second node of the structure tree, return B1, and repeat the process until the traversal is completed;

[0074] Step 6: Obtain the next structure tree in the structure tree set M, and repeat the above process until all the structure trees in the structure tree set M are converted into corresponding path tables. <000025​​​​​​​​​​​​​​​​​​​​​​​​Firstly, the root node is traversed, the root node is Sequence, indicating a sequential structure, and the start and end markers (-Start, -End) are stored in the path table; continue the pre-order traversal, the A node is a task node and is the second node, which is directly stored in the second storage unit of the path table; the third node is And, indicating a parallel structure, And-Start and And-End are inserted after the third storage unit; the fourth node is or, which is also stored after the fourth storage unit; the fifth and sixth nodes are B and C, which are task nodes and are directly stored in the fifth and sixth storage units; the seventh node is a task node D, but the marker in the layer table L is less than the marker of the previous node at this time, indicating that a structure fragment is completed, and the corresponding end marker needs to be jumped out, because the layer marker is less than the maximum layer marker by 1, the seventh node should be inserted in the 7+1=8th storage unit; similarly, the eighth node layer marker is less than the maximum layer marker by 2; should be inserted in the 10th storage unit. After all the traversal is completed, the path table is constructed.

[0082] C: convert all processes in the process library into a path table; take the path table of the process to be recommended and all path tables in the process library for similarity matching to generate a similarity matrix M based on logical structure Sim(r) .

[0083] Preferably, step C comprises the following specific steps:

[0084] C1: repeat the process of steps B1 to B4 to convert all process models in the process library and the process to be recommended into path tables, and obtain a set of path tables with different lengths;

[0085] C2: perform similarity calculation using a logical structure longest common sequence (LSL) method, and perform LSL-based sequence matching of the process to be recommended with all processes in the process library to obtain the longest common subsequence.

[0086] It should be noted that the longest common subsequence refers to a subsequence of two known sequences, and is the longest among all subsequences.

[0087] C3: multiply the length value of the obtained longest common subsequence by 2 as the numerator, and the sum of the lengths of the two sequences participating in matching as the denominator to calculate the similarity Sim(r):

[0088]

[0089] Wherein Sim(r) represents similarity, LSL(r1, r2) represents the longest common subsequence of two arrays r1, r2 containing logical structure; |r1| represents the number of elements in array r1; |r2| represents the number of elements in array r2;

[0090] Then similarity calculation is performed with the process in the process library respectively, and a similarity matrix M based on logical structure is obtained Sim(r) .

[0091] D: Obtain L I and L P in step B, and perform similarity calculation with the feature set of other processes in the process library, to generate a similarity matrix M based on feature attributes Sim(f) .

[0092] Preferably, step D comprises the following specific steps:

[0093] D1: Take the process attribute set L I , encapsulate its attributes in the text description file of the corresponding process, one process corresponds to one text description file; the attribute features of the process construction are stored in the text file in the form of nodes; for example, write user, research direction, applicable satellite, etc., and the character type is string; the specific process attribute setting is as follows:

[0094] UserName, username;

[0095] FlowTrend, process research direction, for example, Sensor (remote sensor), Optical (optical), etc.;

[0096] ItemCo, manufacturing unit;

[0097] SatelliteModel, satellite of interest;

[0098] UserIP, user IP address;

[0099] D2: Take the process parameter set L P , one process corresponds to one attribute set, multiple algorithms, and one algorithm corresponds to multiple parameter sets, L I ~nL A ~mL P . Store the data of each group of parameter sets nL P in the text description file of the corresponding attribute set, and the character type is floating point type, at this time the feature set L C of each process contains the data of attribute set and parameter set at the same time, as shown in Figure 4 .

[0100] D3: Feature set L CThe feature set contains attribute feature γ and parameter feature i, each assigned a weight of 50%. The feature set of the process to be recommended is matched against the attribute feature γ of each process in the process library using a binary method. If the strings match, the match is successful; if they do not match, the match fails. The quotient of the number of successful matches to the total number of matches is used to obtain the attribute feature similarity matrix M. Sim(γ) ;

[0101] D4: Perform similarity matching between the feature set of the process to be recommended and the parameter feature i of each process in the process library. First, match the algorithm set L of the process to be recommended. A Other process algorithm sets L′ A The algorithms in the algorithm are matched. If they are of the same type, the number and meaning of their parameters are consistent; otherwise, there is no matching significance. The intersection L is obtained. A ∩L′ A The algorithms in the intersection are those that appear in both processes. The parameters of the algorithms in the intersection are matched using a binary method; a match is successful if the floating-point values ​​match. The quotient of the number of successful matches to the total number of matches is used as the parameter feature similarity result. After all calculations are completed, the parameter feature similarity matrix M is obtained. Sim(i) ;

[0102] D5: Based on the weights assigned in step D3, M Sim(γ) and M Sim(i) The values ​​are summed after multiplying each by its weight, as shown in the following formula:

[0103]

[0104] Where U1 represents the number of mutually matching elements in the user features, and U0 represents the number of non-matching elements in the user features; I1 represents the number of mutually matching elements in the algorithm features, and I0 represents the number of non-matching elements in the algorithm features.

[0105] Then, the similarity matrix M based on feature attributes is obtained. Sim(f) .

[0106] E: The similarity matrix M based on logical structure Sim(r) Similarity matrix M based on feature attributes Sim(f) Multiplying these results in a new similarity matrix, generating a Top matrix that has the highest similarity to the process to be recommended. n The recommendation process, Top n If the value is greater than 1, obtain the algorithm module with the recommendation position of length(α)+1 in the recommendation process as the recommendation result, return to step A, and repeat the process until the complete remote sensing process is constructed.

[0107] Preferably, step E includes the following specific steps:

[0108] E1: obtaining a similarity matrix M based on the logical structure of the process library Sim(r) and a similarity matrix M based on the feature attribute of the process library Sim(f) Numerical multiplication, the formula is as follows:

[0109]

[0110] Wherein, |r1| represents the number of elements in the array r1; |r2| represents the number of elements in the array r2;

[0111] Then the similarity value of the process to be recommended and other processes in the process library is obtained;

[0112] E2: obtaining the process construction length length (a) of the process to be recommended, and the recommended position length (a) + 1;

[0113] E3: generating the top n Similarity process of the process to be recommended according to the similarity value, and the algorithm of the recommended position length (a) + 1 in the process as the final recommended algorithm is recommended to the user;

[0114] E4: repeat the above process until the complete process is constructed;

[0115] E5: the structure information, attribute information and parameter information in the process construction are respectively stored in the database: the structure information is converted into a path table and stored in the path table set T R ; the attribute information is stored in the attribute set L I ; the parameter information is stored in the parameter set L P . The number and type of processes are constantly expanding.

[0116] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A remote sensing workflow recommendation method based on logical structure and attribute features, characterized in that, The method includes the following steps: A: First, select the required process components and algorithm modules, fill in the parameters, connect each algorithm module with arrows, generate a complete process model, and generate a process structure tree based on logical units from the complete process model; then generate an incomplete process model as the process to be recommended. B: Utilize the aforementioned process structure tree to generate a path table containing logical structures, and collect process parameter information and feature attributes to construct an attribute set for all features of the process object. ; in, A collection of attributes; This is a set of path tables generated based on the process structure tree. The set of process attributes included in the process construction process; E is a set of parameters for each algorithm module during the process of building the workflow; E is a lookup table. C: Convert all processes in the process library into path tables; perform similarity matching between the path table of the process to be recommended and all path tables in the process library to generate a similarity matrix based on logical structure. ; D: Obtain the information from step B and It then calculates the similarity between the feature sets of other processes in the process library and generates a similarity matrix based on feature attributes. ; E: The similarity matrix based on logical structure Similarity matrix based on feature attributes Multiplying them yields a new similarity matrix, generating the one with the highest similarity to the process to be recommended. A recommendation process, Greater than 1; the recommended position in the recommendation process is... The algorithm module is used as the recommendation result, and step A is returned. The process is repeated until a complete remote sensing workflow is constructed. Step B includes the following specific steps: B1: Obtain the process structure tree set, perform a preorder traversal of the structure tree, and first observe whether the root node is a logical node; B2: When the root node is a logical node, observe the child nodes of the root node; if there is a logical node in the layer node where the child node of the root node is located, store the identifier of the logical node in the path table, and use the start mark and end mark to divide the process segment connected by the logical node, and store the start mark and end mark of the logical node in the path table; then traverse the next layer of nodes, and store the start mark and end mark of the logical node in the path table, until the deepest layer of the structure tree is traversed; B3: After reaching the deepest level of the tree structure, traverse from the right child node of the parent node. If the right child node of the parent node exists, then the right child node is at the same level as the node. If the right child node of the parent node does not exist, then the task nodes connected by the parent node have been connected, and the logical segment is exited. If the level of the node is 2 less than the level of the previous node, then two logical segments are exited, and the storage location of the corresponding task unit is after the two end markers. B4: The path table stores the logical relationships in the process. The process attributes and algorithm attributes in the process construction process are stored in the task log text format, and their data are represented by symbols and codes. The lookup table stores the actual meaning of the corresponding symbols and codes.

2. The remote sensing workflow recommendation method based on logical structure and attribute features according to claim 1, characterized in that, Step A includes the following specific steps: A1: On the process building platform, process components and algorithm modules are built by dragging and dropping components, and sequential, parallel, and conditional logic structure modules are selected to connect the processes, finally generating a complete process model guided by arrows. A2: Use the Petri net structure model to convert the process model into a process structure tree.

3. The remote sensing workflow recommendation method based on logical structure and attribute features according to claim 1, characterized in that, Step C includes the following specific steps: C1: Repeat steps B1 to B4 to convert all process models and processes to be recommended in the process library into path tables and obtain path table sets of varying lengths. C2: Similarity is calculated using the longest common subsequence method based on logical structure. The process to be recommended is matched with all processes in the process library based on LSL sequence matching to obtain the longest common subsequence. C3: Multiply the length of the longest common subsequence obtained by 2 as the numerator, and use the sum of the lengths of the two sequences involved in the matching as the denominator to calculate the similarity. : in To represent similarity, LSL(r1, r2) represents the longest common subsequence of logical structure between arrays r1 and r2; |r1| represents the number of elements in array r1; |r2| represents the number of elements in array r2. Then, similarity calculations are performed on each process in the process library to obtain a similarity matrix based on logical structure. .

4. The remote sensing workflow recommendation method based on logical structure and attribute features according to claim 1, characterized in that, Step D includes the following specific steps: D1: Retrieve process attribute set The attributes are encapsulated in the corresponding text description file of the process, with one text description file for each process; the attribute features during process construction are stored in the text file in the form of nodes. D2: Retrieve process parameter set One process corresponds to one attribute set and multiple algorithms, and one algorithm corresponds to multiple parameter sets. The data of each parameter set is stored in the corresponding attribute set's text description file, with the character type being floating-point. At this point, the feature set of each process... It contains data that includes both attribute sets and parameter sets; D3: Feature Set Includes attribute features and parameter features Each of these two features is assigned a weight of half. The feature set of the process to be recommended is compared with the attribute features of each process in the process library. Similarity matching is performed using a binary method. If the strings match, the match is successful; if the strings do not match, the match fails. The quotient of the number of successful matches to the total number of matches is used to obtain the attribute feature similarity matrix. ; D4: Compare the feature set of the process to be recommended with the parameter features of each process in the process library. To perform similarity matching, the algorithm set of the stream to be recommended is first... Other process algorithm sets The algorithms in the algorithm are matched. If they are of the same type, the number and meaning of their parameters are consistent; otherwise, there is no matching significance. The intersection is obtained. The parameters in the algorithm within the intersection are matched using a binary method. A successful match is indicated by a matching of floating-point values. The quotient of the number of successful matches to the total number of matches is used as the parameter feature similarity result. After all calculations are completed, the parameter feature similarity matrix is ​​obtained. ; D5: Based on the weights assigned in step D3, and The values ​​are summed after multiplying each by its weight, as shown in the following formula: in, This indicates the number of mutually matching elements in the user characteristics. Indicates the number of non-matching elements in the user characteristics; This represents the number of mutually matching elements in the algorithm's features. This indicates the number of non-matching elements in the algorithm's features; Then, a similarity matrix based on feature attributes is obtained. .

5. The remote sensing workflow recommendation method based on logical structure and attribute features according to claim 1, characterized in that, Step E includes the following specific steps: E1: Similarity matrix based on logical structure Similarity matrix based on feature attributes The formula for multiplying numerical values ​​is as follows: Where |r1| represents the number of elements in array r1; |r2| represents the number of elements in array r2; Then, the similarity value between the process to be recommended and other processes in the process library is obtained; E2: Get the process construction length of the process to be recommended. Recommended location E3: Generate the process with the highest similarity to the one to be recommended based on the similarity value. The recommendation process has a specific location within it. The algorithm is used as the final recommendation algorithm to recommend to the user.

Images