CN112069927B - Element set processing method and device applied to modularized vision software - Google Patents
Element set processing method and device applied to modularized vision software Download PDFInfo
- Publication number
- CN112069927B CN112069927B CN202010838638.8A CN202010838638A CN112069927B CN 112069927 B CN112069927 B CN 112069927B CN 202010838638 A CN202010838638 A CN 202010838638A CN 112069927 B CN112069927 B CN 112069927B
- Authority
- CN
- China
- Prior art keywords
- node
- data
- program
- splitting
- collection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims description 6
- 238000000034 method Methods 0.000 claims abstract description 40
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 13
- 230000009191 jumping Effects 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 238000013480 data collection Methods 0.000 abstract 1
- 238000004519 manufacturing process Methods 0.000 description 7
- 239000013072 incoming material Substances 0.000 description 6
- 230000000007 visual effect Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 241001522296 Erithacus rubecula Species 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011010 flushing procedure Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Abstract
In the machine vision recognition, an image recognition program recognizes an image, a template or a line circle is searched for an element set in the image, a 'splitting-collecting' mode is used for constructing a cyclic body program to process the recognition task of the image, list data of the recognition task enters a data processing program after passing through a splitting node, the result of the data processing program is output through a collecting node, the data collection of the collecting node is not finished, the splitting node is returned to continuously execute the splitting node flow, the splitting and collecting are executed once each time, and the data is reserved. The method of the invention has the advantages of no data loss in circulation, high data freedom degree, reduced flow complexity, program writing difficulty and error probability of data flow circulation processing, easy use by hands, and easier debugging.
Description
Technical Field
The invention belongs to the technical field of computers, relates to robot vision, and discloses a container data processing method applied to modularized vision software.
Background
Industry 4.0 is vigorous in development, new requirements are put on the requirement universality and the freedom degree of an industrial field, and the robot is matched with vision operation, so that vision auxiliary software for assisting the robot operation is derived. The software collects field pictures, analyzes object characteristics and interacts with robot data. However, the object recognition by using the vision software has a problem of recognizing many interference items at a time, so that the mainstream vision software on the market usually solves the problem in a cyclic processing manner, i.e. a set of cyclic modules is designed to process the analysis result of the robot, and the data may be a set of positions or a set of lines.
At present, a general circulation processing mode is to create a circulation module comprising an output module and a processing module, select the associated data and the associated data size, and output the result after analysis and calculation. The following takes the visual software of the hikva ision as an example, as shown in fig. 1.
FIG. 1 shows a simple round robin case where the data stream processed in the round robin module is input into other flows. The case has obvious problems that elements output in each execution cycle in the cycle module cannot be reserved, and finally only the processing result of the last cycle execution can be output, and the rest elements are all lost.
The data flow point set of fig. 1 is disassembled on the basis of fig. 1 to obtain fig. 2, in which the data nodes of each module in and out cycle are marked. When the element enters the loop body for processing, the element is split into single data by the crossed nodes and then submitted to the flow in the loop body; when the loop body is directly ended, the last element is submitted to other processes through the data node.
Obviously, in the conventional loop processing mode, for the data node, the data node entering the loop can be accessed for a plurality of times, the access times are consistent with the loop times, and the node jumping out of the loop can be accessed only once. Because each execution of the flow only saves the current state of the flow, and the data nodes are not processed in the traditional vision software, the data processing has great limitation. For example, a data node entering a loop may execute multiple times, but essentially only the last time it is valid data, the intermediate data being lost in the next execution. The reason is that only the data of the cycle (the last data of the cycle) is useful to be output after the cycle is ended. This problem is an unavoidable disadvantage in the conventional art.
To solve the above, conventional vision software adds a concept of global data, as shown in fig. 3. The global data is not affected by any execution scope, and when the data is sent to the global data, the global data is reserved, so that the intermediate data is not lost.
However, the global data is not affected by the scope, and although the problem of data loss is solved, the situation of out-of-control is caused, for example, the global data cannot be emptied along with the end of loop execution, program execution is performed, and a large amount of error data can occur in the global data. This requires additional logic to implement global data flushing and other processing. The scheme clearly increases the complexity and the error probability of the program, meanwhile, the use experience of the software is poor, the production debugging needs to consume a large amount of time for processing, and global data outside the program is needed to host the data inside the program, so that the programming difficulty of the logic of the program is greatly increased, and the method is unfavorable for rapidly responding to the actual production requirements in the production environment.
Disclosure of Invention
The invention aims to solve the technical problems that: when the industrial products are classified and split-packed through machine vision, data loss is easy to occur in the processing process of visual identification data, and the complexity and the error probability of the existing scheme for solving the data loss are high.
The technical scheme of the invention is as follows: in the machine vision recognition, an image recognition program recognizes an image, finds a corresponding element set according to a recognition target, and processes the element set to obtain a recognition result; when the loop is executed once, the corresponding splitting and collecting are executed once, and in the executing process, the splitting node and the collecting node respectively store the output data of the node.
Further, the method comprises the following steps:
1) The loop body program obtains an element set;
2) The element set is taken as the input of the splitting node to split, the split element is the input element of the data processing program, when the data in the element set starts the splitting processing, the splitting node extracts the element from the beginning to the end of the data, and one element is put forward and output to the data processing program;
3) The data processing program is contained in a circulating body consisting of a splitting node and a collecting node, and when the data processing program in the circulating body obtains input elements from the splitting node, the processing program is executed to output processed new elements;
4) When the data processing program is executed to the collection node, the collection node uses the new element output by the data processing program in the last step as input, and stores the new element into the collection node to form a new element set which is used as the output of the node;
5) After the collection node finishes the collection execution once, checking split nodes in the loop body program, returning to the step 2) to continue splitting when the split execution of the split nodes on the element set is not finished, until all elements in the element set are extracted, wherein in the process, output elements in the split nodes and the collection node are reserved in the respective nodes, and the data processing program refreshes all elements in each loop process; otherwise, if the split node is already executed, executing the step 6);
6) And jumping out the loop, using the output element set of the collection node as an input element of other subsequent processes, and continuing to execute the other processes until the execution of the loop body program is finished.
The invention also provides an element set processing device applied to the modularized vision software, which is a device with data processing capability, wherein a software program is configured, and the software program realizes the element set processing method when being executed.
Based on the problems existing in the prior art, the invention provides that the node self data is managed in the circulated data node, and meanwhile, the complex modes of using global data assistance and the like are abandoned, namely, all places where the data processing exists must have data processing modules, and all the data processing must be carried out in the scope, the linear processing process and no irrelevant third party reference. The thinking logic flow for converting the field requirements is easily realized through the modules, the complexity of the whole flow, the programming difficulty and the error probability are greatly reduced, the method is easy to use, and meanwhile, the debugging is easier.
The invention has the following advantages:
1. under the method of the invention, the degree of freedom of the split data is high, and the realization of different calculations of multiple branches of the data is supported;
2. because each node is processing its own element, no data loss occurs in the loop;
3. the cyclic body program provided by the method greatly reduces the complexity of the program and is beneficial to the development work of designers.
Drawings
Fig. 1 is an example of a simple loop process of the hikvin visual software of the haven's vision.
FIG. 2 is an example of the addition of data nodes labeled each module's in and out loop based on FIG. 1.
Fig. 3 is an example of adding global data to existing visual software.
FIG. 4 is an illustration of the present invention for the split and collection of a common set of loops.
FIG. 5 is an example of the use of multiple splits to collect different types of data in the present invention.
Fig. 6 is a schematic flow chart of a split node according to the present invention.
Fig. 7 is a schematic flow chart of a collection node according to the present invention.
FIG. 8 is a schematic illustration of an incoming material site, in accordance with an embodiment of the present invention.
FIG. 9 is a flow chart of data processing according to an embodiment of the present invention.
FIG. 10 is a flow chart of the prior art for performing the data processing functions required in the embodiments of the present invention.
Detailed Description
The invention provides a data processing algorithm scheme, namely a split-collection mode, in vision auxiliary software.
The invention provides the following solution when the sorting and sub-packaging operation is needed for the raw materials in the industrial field aiming at the occasion of machine vision recognition of products.
The operation flow of the industrial site is as follows: photographing a raw material workpiece, analyzing the photograph to obtain a type data set, processing the data, processing the result, sending the robot, sorting by the robot and producing.
Firstly, transmitting incoming materials, photographing the incoming materials by using an industrial camera to obtain a high-resolution bitmap photo, and transmitting the high-resolution bitmap photo to an image recognition system.
The image system searches for the unit set according to the template or the line circle, and performs data processing. The image recognition program recognizes the image, searches for an element set in the image according to a template or a line circle, wherein the element set refers to a result found in vision, the result is various, such as a circle is found, the result of line search is a group of lines, wherein the single circle and line are called elements, the data are stored in corresponding processing nodes, and when a plurality of elements exist, the data are called element sets. And processing the element combination, for example, in the line searching, processing the line data to obtain intersection point data, and transmitting the intersection point data back to the robot.
The invention uses a 'split-collect' mode to construct a cyclic body program for an element set to process the identification task of an image, wherein the cyclic body program comprises a split node, a data processing program and a collection node, after the element set passes through the split node, split output elements enter the data processing program, the data processing program processes and outputs new elements, after the element set is collected by the collection node, the new element set is output, after the collection node executes, if the split of the split node is detected not to be finished, namely, the collection task is not finished, the cyclic body program returns to the split node to continuously execute the split node flow, and therefore a cyclic processing structure is formed; when the loop is executed once, the corresponding splitting and collecting are executed once, and in the executing process, the splitting node and the collecting node respectively store the output data of the node. The new mode of split-collect of the present invention focuses on describing the nature of data stream processing as opposed to traditional loop modes.
The following is an example of analysis.
The circulation flow obtained by using the technology of the invention is shown in fig. 4, and a circulation body is constructed by Split (Split) and collection (collection) together. In this mode, the boundary of data processing is to split and collect the flow of global data and auxiliary management global data of a third party, which is not needed any more, compared with the conventional loop flow of fig. 3, and the flow is to be executed sequentially, and has no redundant branches and is easy to write. When the program is executed, the splitting and collecting are executed once every time the loop is executed, and data are reserved. This ensures both the independence and scope of the processing modules and solves the problem of data loss. The method comprises the following specific steps:
1) The method comprises the steps that a loop body program obtains an element set Y, wherein in the program, elements are embodied as data arrays, and the element set is embodied as a data list formed by a plurality of data arrays;
2) The element set is taken as the input of the splitting node to split, the split element is the input element of the data processing program, when the data in the element set starts to split, the splitting node extracts the elements from the data list according to the sequence from beginning to end, and one element is put forward and output to the data processing program;
3) The data processing program is contained in a circulating body consisting of a splitting node and a collecting node, and when the data processing program in the circulating body obtains input elements from the splitting node, the processing program is executed to output processed new elements;
4) When the data processing program is executed to the collection node, the collection node uses the new element output by the data processing program in the last step as input, and stores the new element into the collection node to form a new element set Y' which is used as output of the node;
5) After the collection node finishes the collection execution once, checking split nodes in the loop body program, returning to the step 2) to continue splitting when the split execution of the split nodes on the element set is not finished, until all elements in the element set are extracted, wherein in the process, output elements in the split nodes and the collection node are reserved in the respective nodes, and the data processing program refreshes all elements in each loop process; otherwise, if the split node is already executed, executing the step 6);
6) And jumping out the loop, using the output element set Y' of the collection node as an input element of other subsequent processes, and continuing to execute the other processes until the execution of the loop body program is finished.
More complex data processing can be done on the loop because the nodes of the data processing are directly exposed. For example, the collection node may terminate execution by using a termination condition of the split node, and in the cyclic body program, set a plurality of split nodes to split element sets of different types, and the collection node checks an execution condition of the corresponding split node according to an input element type, and further executes different cyclic paths to collect element data of different types. Meanwhile, the effect that the collected data is different from the original data can be achieved, conditions are set between the data processing program and the collection nodes, and the output of the data processing program is divided into different collection nodes according to the conditions, so that different collection results are obtained. As in fig. 5, the two collected data sets are clearly not uniform in size, but both are desirable.
Fig. 6 shows a specific execution flow of the split node.
Fig. 7 shows a specific implementation flow of the collection node.
In the above process, when the splitting and collecting are not completed, the node data will not be emptied although the node data will still jump, which is different from other modules in the program.
After all sorting is finished, the data are formatted and sent to the corresponding robots. And the robot sorts the incoming materials according to the data, and finally enters a production link.
The method of the invention is embodied as a software program, and therefore the invention correspondingly proposes an element set processing device applied to modularized vision software, the device being a device with data processing capability, wherein the software program is configured, and the software program is executed to realize the element set processing method of the invention.
The invention is aimed at gear screening of an automobile assembly production line, and the working flow is described below.
Firstly, we use ID to mark different gear specifications in the template, when a new batch of gears comes, the camera takes a picture, and then enters the identification stage.
And (3) completing identification of a group of workpieces, formatting the identified gear model information in an (X, Y, A, ID) format, sending the formatted character string to a robot for identification, and screening and carrying gears to different production lines, such as a transmission, an engine and the like, by the robot according to the workpiece information. Where X, Y denotes the X-coordinate and Y-coordinate of the workpiece position, A denotes the rotation angle of the workpiece (angle of the template is 0), and ID denotes the number of the gear template. X, Y, A, ID are integers, and the formatted string is styled as (179,105,0,1). It should be noted that the gears are different in size and shape, and the number of gears carried by each batch is variable, as shown in fig. 8, which is a typical situation of the incoming material site.
There are three different gears in the scenario of fig. 8, each gear is also different in number, a new batch of gears is delivered in at intervals, and the time for next gear delivery, and whether the gear is screened out when delivered are unknown. Therefore, constant screening is required, and all information is formatted and sent to the robot each time. The flow of fig. 9 can deal with this scenario problem, which is a linear flow without any third party references and complex branches. The detailed description steps are as follows:
1. executing a template matching module to obtain four groups of data, namely X < [ ] (position abscissa), Y < [ ] (position ordinate), A < [ ] (angle), ID < [ ] (type number):
X[]＝{78,352,179,321,55,170,306}
Y[]＝{68,81,105,208,229,265,299}
A[]＝{0,90,0,90,20,20,20}
ID[]＝{1,4,1,4,2,2,2}
2. executing the splitting module, wherein the starting position is 1, and the ending position is the array length. After splitting, four data are obtained each time, namely X (position abscissa), Y (position ordinate), A (angle) and ID (type number), the array length is 7, the splitting is needed for 7 times, and the data obtained by each splitting are respectively:
3. after splitting the data, formatting all four data referenced each time, wherein the number of formatting execution times is the same as the splitting number in the circulation, and the formatting result outputs 7 groups of character strings which are respectively:
1)(78,68,0,1)
2)(352,81,90,4)
3)(179,105,0,1)
4)(321,208,90,4)
5)(55,229,20,2)
6)(170,265,20,2)
7)(306,299,20,2)
here, the result of the formatting is merged by using the merge string function, and the merge interval uses ",", after 7 times of execution are completed, the string data inside the collection node is: (78,68,0,1), (352,81,90,4), (179,105,0,1), (321,208,90,4), (55,229,20,2), (170,265,20,2), (306,299,20,2). Before the splitting is finished, the collection is not finished, when the splitting position reaches 7, the splitting is finished, the collection is finished when the splitting is finished, and the step 4 is continued to be executed.
4. The combined string in the collection node is formatted, plus the protocol head-to-tail flag bits, into the final result, i.e., [ (78,68,0,1), (352,81,90,4), (179,105,0,1), (321,208,90,4), (55,229,20,2), (170,265,20,2), (306,299,20,2) ].
5. And sending the formatted character string to the robot through a data sending module.
6. And the robot sorts the incoming materials according to the formatted information and enters the subsequent production junction stage.
The flow ends up here.
The process of fig. 9 is executed circularly, so that the requirement of circular processing can be met. In contrast to fig. 10, if the above functional requirements are completed by using the conventional visual software, the complexity of the flow is greatly increased, and the data logic is required to be additionally maintained due to the addition of the third party global data which is difficult to control, which clearly increases the complexity of the program, and in addition, whether the data emptying and adding timing is proper or not needs to be noted.
Obviously, the cyclic processing mode provided by the invention at least simplifies half of traditional industrial production processing logic, and linear logic is easy to understand and write.
Claims (6)
1. In the machine vision recognition, an image recognition program recognizes an image, finds a corresponding element set according to a recognition target, and processes the element set to obtain a recognition result; when the loop is executed once, the corresponding splitting and collecting are executed once, and in the executing process, the splitting node and the collecting node respectively store the output data of the node.
2. The method for processing the element set applied to the modularized vision software according to claim 1, wherein the method comprises the following steps:
1) The loop body program obtains an element set;
2) The element set is taken as the input of the splitting node to split, the split element is the input element of the data processing program, when the data in the element set starts the splitting processing, the splitting node extracts the element from the beginning to the end of the data, and one element is put forward and output to the data processing program;
3) The data processing program is contained in a circulating body consisting of a splitting node and a collecting node, and when the data processing program in the circulating body obtains input elements from the splitting node, the processing program is executed to output processed new elements;
4) When the data processing program is executed to the collection node, the collection node uses the new element output by the data processing program in the last step as input, and stores the new element into the collection node to form a new element set which is used as the output of the node;
5) After the collection node finishes the collection execution once, checking split nodes in the loop body program, returning to the step 2) to continue splitting when the split execution of the split nodes on the element set is not finished, until all elements in the element set are extracted, wherein in the process, output elements in the split nodes and the collection node are reserved in the respective nodes, and the data processing program refreshes all elements in each loop process; otherwise, if the split node is already executed, executing the step 6);
6) And jumping out the loop, using the output element set of the collection node as an input element of other subsequent processes, and continuing to execute the other processes until the execution of the loop body program is finished.
3. The method for processing element sets applied to modularized vision software according to claim 1 or 2, wherein in a cyclic body program, a plurality of splitting nodes are arranged to correspondingly split element sets of different types, and a collecting node checks the execution condition of the corresponding splitting nodes according to the input element types so as to execute different cyclic paths and collect element data of different types.
4. The method for processing the element set applied to the modularized vision software according to claim 1 or 2, wherein conditions are set between the data processing program and the collection nodes, and the output of the data processing program is divided into different collection nodes according to the conditions, so that different collection results are obtained.
5. A method for processing an element set applied to modular vision software according to claim 3, wherein conditions are set between the data processing program and the collection nodes, and the output of the data processing program is divided into different collection nodes according to the conditions to obtain different collection results.
6. An element set processing device for use in modular vision software, characterized in that the device is a device having data processing capabilities, in which a software program is arranged, which software program, when executed, implements the element set processing method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010838638.8A CN112069927B (en) | 2020-08-19 | 2020-08-19 | Element set processing method and device applied to modularized vision software |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010838638.8A CN112069927B (en) | 2020-08-19 | 2020-08-19 | Element set processing method and device applied to modularized vision software |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112069927A CN112069927A (en) | 2020-12-11 |
| CN112069927B true CN112069927B (en) | 2023-10-24 |
Family
ID=73662298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010838638.8A Active CN112069927B (en) | 2020-08-19 | 2020-08-19 | Element set processing method and device applied to modularized vision software |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112069927B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113468770B (en) * | 2021-09-02 | 2021-11-12 | 成都新西旺自动化科技有限公司 | Method and system for generating machine vision formula |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101387958A (en) * | 2008-10-20 | 2009-03-18 | 东软集团股份有限公司 | Image data processing method and apparatus |
| CN109254946A (en) * | 2018-08-31 | 2019-01-22 | 郑州云海信息技术有限公司 | Image characteristic extracting method, device, equipment and readable storage medium storing program for executing |
| CN110766671A (en) * | 2019-10-21 | 2020-02-07 | 深圳市涌固精密治具有限公司 | Image processing method based on machine vision software |
| CN111178431A (en) * | 2019-12-20 | 2020-05-19 | 北京邮电大学 | Network node role identification method based on neural network and multi-dimensional feature extraction |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7801924B2 (en) * | 2006-12-28 | 2010-09-21 | Infosys Technologies Ltd. | Decision tree construction via frequent predictive itemsets and best attribute splits |
| JP4952317B2 (en) * | 2007-03-16 | 2012-06-13 | 富士通株式会社 | Saved data discrimination method, saved data discrimination program, and saved data discrimination device |
| US9031317B2 (en) * | 2012-09-18 | 2015-05-12 | Seiko Epson Corporation | Method and apparatus for improved training of object detecting system |
| US9489639B2 (en) * | 2013-11-13 | 2016-11-08 | Microsoft Technology Licensing, Llc | Memory facilitation using directed acyclic graphs |
| US10303757B2 (en) * | 2016-08-19 | 2019-05-28 | Cognex Corporation | Apparatuses, systems, and methods for providing a visual program for machine vision systems |
-
2020
- 2020-08-19 CN CN202010838638.8A patent/CN112069927B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101387958A (en) * | 2008-10-20 | 2009-03-18 | 东软集团股份有限公司 | Image data processing method and apparatus |
| CN109254946A (en) * | 2018-08-31 | 2019-01-22 | 郑州云海信息技术有限公司 | Image characteristic extracting method, device, equipment and readable storage medium storing program for executing |
| CN110766671A (en) * | 2019-10-21 | 2020-02-07 | 深圳市涌固精密治具有限公司 | Image processing method based on machine vision software |
| CN111178431A (en) * | 2019-12-20 | 2020-05-19 | 北京邮电大学 | Network node role identification method based on neural network and multi-dimensional feature extraction |
Non-Patent Citations (2)
| Title |
|---|
| 基于InfiniBand集群并行全景图拼接系统设计;陆应通;李文强;吴志红;;四川大学学报(自然科学版)(第06期);1247-1254 * |
| 特征点法视觉SLAM中的点特征提取算法研究对比;张光耀;黄通交;;信息系统工程(第06期);105-106 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112069927A (en) | 2020-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| DE60318722T2 (en) | A PROGRAMMABLE RULE PROCESSING DEVICE FOR HIGH-SPEED CONTEXT SEARCHING AND RECOGNITION OF PATTERNS IN DATA | |
| CN109685110B (en) | Training method of image classification network, image classification method and device, and server | |
| CN102521838B (en) | Image searching/matching method and system for same | |
| CN112926641A (en) | Three-stage feature fusion rotating machine fault diagnosis method based on multi-modal data | |
| US20190244091A1 (en) | Acceleration of neural networks using depth-first processing | |
| CN112069927B (en) | Element set processing method and device applied to modularized vision software | |
| CN103136364A (en) | Cluster database system and data query processing method thereof | |
| CN112860692B (en) | Database table structure conversion method and device and electronic equipment thereof | |
| CN116187391A (en) | Neural network model processing method and device | |
| CN115423084A (en) | Systolic array, systolic array system, method and apparatus for computing systolic array system, and storage medium | |
| CN109472484A (en) | A kind of production process exception record method based on flow chart | |
| CN114374708B (en) | Intelligent factory data automatic subscription method based on collaborative filtering and MQTT | |
| CN114372097B (en) | Efficient connection comparison implementation method and device for data set serialization | |
| Mao et al. | Complex Event Processing on uncertain data streams in product manufacturing process | |
| Joan Pep Company-Corcoles et al. | LiPo-LCD: Combining Lines and Points for Appearance-based Loop Closure Detection. | |
| CN114115023A (en) | Method for realizing identification and operation of AVC configuration file | |
| CN100483402C (en) | Programmable rule processing apparatus for conducting high speed contextual searches & characterzations of patterns in data | |
| CN104598205A (en) | Sorting system and method for dataflow of function block diagram | |
| CN111460747A (en) | Standard unit tracking method for integrated circuit design | |
| Ito et al. | Realization of a neural network model neocognitron on a hypercube parallel computer | |
| US20200311021A1 (en) | Systems and methods for reconfigurable systolic arrays | |
| Lam et al. | Analyzing equivalences of UML statechart diagrams by structural congruence and open bisimulations | |
| Chao et al. | Incomplete Contrastive Multi-View Clustering with High-Confidence Guiding | |
| TWI801836B (en) | Method for transforming and optimizing codes between different programing languages | |
| Yang et al. | Searching cycle-disjoint graphs |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |