Intelligent factory safety early warning system based on artificial intelligence

Abstract

The invention discloses a smart factory safety early warning system based on artificial intelligence, and relates to the technical field of artificial intelligence, and the system comprises a monitoring center which is connected with a smart collection module, a factory processing module, an operation analysis module and a dynamic early warning module. Tense video data and comprehensive monitoring data are acquired through an intelligent acquisition module; performing code conversion on the comprehensive monitoring data in a factory processing module to obtain a local coefficient set, and performing conversion combination on temporal video data to obtain a target element sequence; a refinement criterion is set in the operation analysis module to carry out form calibration and feature selection on the target element sequence, and a feature target map sequence is obtained; the dynamic early warning module performs video matching on the feature target sequence according to the local coefficient set to obtain a matched video segment, and sends an early warning instruction to the warning response area through the monitoring center to obtain area warning information; the early warning accuracy and the early warning response speed are improved, and the factory safety accident rate is reduced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based smart factory safety early warning system. Background Art

[0002] Artificial intelligence is an interdisciplinary technical science that aims to study and develop theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. The goal of artificial intelligence is to enable machines to perform complex tasks.

[0003] A smart factory is a factory that uses modern information technologies such as the Internet of Things, big data analysis, artificial intelligence, and cloud computing to achieve automation, informatization, and intelligence in the production process. It uses highly integrated information systems and intelligent devices to monitor and optimize the production process in real time, improve production efficiency, reduce resource consumption, and enhance product quality. Due to the high degree of automation and fast equipment operation in smart factories, once a failure or abnormality occurs, it may quickly lead to a safety accident. Although automation reduces the direct participation of personnel, employees still need to work in the factory, and the safety warning system can protect employees from harm; therefore, real-time monitoring of smart factories, detection of factory operation status, early detection and warning of potential safety production risks, and thus taking preventive measures to reduce the occurrence of safety accidents are the problems we need to solve; for this reason, a smart factory safety warning system based on artificial intelligence is now provided. Summary of the invention

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] An artificial intelligence-based smart factory safety early warning system includes a monitoring center, wherein the monitoring center is connected to a smart acquisition module, a factory processing module, an operation analysis module, and a dynamic early warning module;

[0006] The process of collecting temporal video data and comprehensive monitoring data by the intelligent acquisition module includes:

[0007] Set up real-time capture end and intelligent collection end;

[0008] The real-time capture terminal is used to collect video of factory components to obtain temporal video data, and the collected temporal video data is time-stamped to obtain the capture time;

[0009] The factory components are collected at the edge through the intelligent collection terminal to obtain comprehensive monitoring data, and the collected comprehensive monitoring data is time-stamped to obtain the monitoring time;

[0010] The synthetic monitoring data is associated with the corresponding temporal video data based on the monitoring time.

[0011] The process of obtaining the regulation extraction coefficient segment includes:

[0012] Performing code conversion on the obtained comprehensive monitoring data to obtain a comprehensive monitoring code;

[0013] Set a coding extraction coefficient, compress and change the coding extraction coefficient to obtain a compression index, compress and extract the coding extraction coefficient according to the compression index to obtain a compression interval, regulate and allocate the coding extraction coefficient according to the compression index and the compression interval to obtain a regulation set number;

[0014] The coding extraction coefficients are intermittently truncated according to the number of control sets to obtain control extraction coefficient segments.

[0015] The process of obtaining matching video data includes:

[0016] Performing original segmentation on the comprehensive monitoring code according to the obtained control extraction coefficient segment to obtain the original coding coefficient segment;

[0017] The obtained original coding coefficient segments are sorted and counted based on the order of discontinuous truncation to obtain a local coefficient set;

[0018] The monitoring time corresponding to the local coefficient set is obtained, the capture time is time-matched according to the monitoring time, and the successfully matched temporal video data is marked as the matching video data.

[0019] The process of obtaining the target element sequence includes:

[0020] Set a key interval, perform target framing on the matching video data according to the key interval, obtain a target video frame, perform image-frame conversion on the target video frame, and obtain a target component image;

[0021] The obtained target component graphs are combined based on the matching video data to obtain a target component sequence, and the obtained target component sequence is associated with the corresponding temporal video data.

[0022] The process of obtaining the feature target graph sequence includes:

[0023] Setting a refinement criterion, performing formal calibration on the target element sequence according to the refinement criterion, and obtaining a standard target graph sequence;

[0024] A filter kernel is selected according to the local coefficient set setting, and the selected filter kernel is uploaded to the standard target graph sequence. Feature selection is performed on the standard target graph sequence by selecting the filter kernel to obtain a feature target graph sequence, wherein the feature target graph sequence includes the standard target graph.

[0025] The process of feature selection by selecting and filtering the standard target graph sequence includes:

[0026] Matching and dividing the standard target image according to the local coefficient set to obtain the filtered matching area;

[0027] Set the link start point, upload the selected filter core to the link start point, capture and extract the filter matching area by selecting the filter core, and obtain the capture filter area;

[0028] Move the selected filter core to the next filter matching area, and capture and extract it with the filter matching area until all the filter matching areas are captured and extracted with the selected filter core, and replace the original filter matching area with the captured filter area in the standard target image to obtain a feature capture image;

[0029] The obtained feature capture graphs are sorted based on the standard target graph sequence to obtain a feature target graph sequence.

[0030] The process of obtaining a hazard warning sequence includes:

[0031] Set a capture identity point according to the feature capture map, upload a local coefficient set to a feature target sequence based on the capture identity point, perform terminal verification on the feature target sequence according to the local coefficient set, and obtain a terminal code set;

[0032] Perform digital-to-graphic conversion on the terminal code set to obtain a terminal monitoring curve chart, set a threshold comparison axis, perform threshold comparison on the terminal monitoring curve chart through the threshold comparison axis, and obtain a primary warning coefficient;

[0033] According to the primary warning coefficient, the characteristic target sequence is marked with warning marks to obtain a color warning map;

[0034] Danger determination is performed on the color warning image based on the characteristic target sequence to obtain a danger warning sequence, wherein the danger warning sequence includes continuous warning images.

[0035] The process of obtaining regional alert information includes:

[0036] Match the local coefficient set according to the continuous warning map to obtain the warning monitoring data;

[0037] Acquire temporal video data corresponding to the danger warning sequence, perform video matching on continuous warning images based on key intervals, and obtain matching video segments;

[0038] Acquire plant components corresponding to the danger warning sequence, and mark the acquired plant components as warning response areas;

[0039] The matching video segments and warning monitoring data are uploaded to the monitoring center, and a warning instruction is issued to the warning response area through the monitoring center. The warning response area performs strategy transformation on the matching video segments and warning monitoring data according to the obtained warning instruction to obtain regional alarm information.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The collected comprehensive monitoring data is encoded and converted to obtain a local coefficient set, and the collected temporal video data is converted, combined and feature selected to obtain a feature target graph sequence; video monitoring and production data monitoring are performed on the factory area of ​​the smart factory to increase the speed of safety warnings and quickly locate dangerous locations, reduce the incidence of factory safety accidents, and ensure production and employee safety;

[0042] Video matching is performed on the feature target sequence according to the local coefficient set to obtain the matching video segment. The monitoring center issues an early warning instruction to the warning response area to obtain regional alarm information. The monitoring data and video data are synchronously extracted to determine the dangerous data value and the video data corresponding to the dangerous equipment, which greatly improves the accuracy of the early warning and the response speed, allowing relevant personnel to quickly take preventive measures to reduce accident losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] like Figure 1 As shown, a smart factory safety early warning system based on artificial intelligence includes a monitoring center, and the monitoring center is connected to a smart acquisition module, a factory processing module, an operation analysis module and a dynamic early warning module;

[0047] The smart acquisition module is used to collect temporal video data and comprehensive monitoring data. The specific process includes:

[0048] Set up intelligent collection end and real-time capture end;

[0049] The video of the factory components is collected through the real-time capture terminal to obtain temporal video data, the time length of the temporal video data is marked to obtain the video duration, and the collected temporal video data is time-marked to obtain the capture time;

[0050] Further, the factory components include production equipment components, personnel performing production construction, and component areas obtained by dividing the free space in the smart factory, wherein "component areas obtained by dividing the free space" means that in the smart factory, the free positions other than production equipment and factory personnel are equidistantly divided to obtain areas of the same size, which are recorded as component areas, and one component area is recorded as a factory component; then "video acquisition of factory components" means video acquisition of all machines and factory areas controlled by the smart factory through a real-time capture terminal; the temporal video data means video monitoring of all areas in the smart factory, capturing video data corresponding to factory components, and obtaining temporal video data;

[0051] The factory components are collected at the edge through the intelligent collection terminal to obtain comprehensive monitoring data, and the collected comprehensive monitoring data is time-stamped to obtain the monitoring time;

[0052] The comprehensive monitoring data includes environmental data, equipment status data, production data and personnel behavior data, wherein the environmental data includes temperature, humidity, harmful gas concentration, smoke concentration and light intensity, the equipment status data includes equipment operating temperature, equipment vibration, equipment pressure, motor current and voltage, equipment operating speed and equipment load, the production data includes production speed, production process and production quality inspection, and the personnel behavior data includes personnel safety equipment, personnel violations, personnel activities and personnel location;

[0053] The obtained comprehensive monitoring data is associated with the corresponding temporal video data based on the monitoring time, which means that the comprehensive monitoring data corresponding to the factory elements at the same time are associated with the temporal video data. That is, by obtaining the same monitoring time and capture time, the comprehensive monitoring data and temporal video data corresponding to the factory elements at the same time can be obtained.

[0054] The factory processing module is used to encode and convert the comprehensive monitoring data to obtain a local coefficient set, convert and combine the temporal video data to obtain a target element sequence. The specific process includes:

[0055] Performing code conversion on the obtained comprehensive monitoring data to obtain a comprehensive monitoring code;

[0056] The code conversion means converting the obtained comprehensive monitoring data into a signal form, that is, a comprehensive monitoring code;

[0057] Further, according to the environmental data, equipment status data, production data and personnel behavior data included in the comprehensive monitoring data, the comprehensive monitoring code obtained after the comprehensive monitoring data is coded and converted includes the environmental monitoring code, the equipment status code, the production monitoring code and the personnel behavior code;

[0058] Setting a code extraction coefficient, wherein the code extraction coefficient is expressed in a function form;

[0059] Compressing and changing the obtained coding extraction coefficient to obtain a compression index;

[0060] Furthermore, the compression change indicates controlling the scaling and translation of the coded extraction coefficients in the time dimension and the frequency dimension, and marking the distance of the scaling and translation to obtain a compression index;

[0061] According to the obtained compression index, the coded extraction coefficient is compressed and extracted to obtain a compression interval;

[0062] Furthermore, the compression extraction means marking the distance between two adjacent compression indexes in the coded extraction coefficient to obtain the compression spacing;

[0063] According to the obtained compression index and compression interval, the coding extraction coefficients are regulated and allocated to obtain the regulated set number, and the obtained regulated set number is marked as TK, where: , ZH represents the length of the integrated monitoring code, y represents the compression interval, m represents the number of compression intervals in the integrated monitoring code, a represents the compression index, β1 represents the rounding factor, and by adding β1, the result of the control set number is an integer, and when " When the decimal part of " is less than or equal to 0.5, β1 is negative. When the decimal part of " is greater than 0.5, β1 is a positive number; for example, " "The result is 6.5, so β1=-0.5, and TK=6.5+(-0.5)=6;

[0064] According to the obtained number of control sets, the coding extraction coefficients are intermittently intercepted to obtain control extraction coefficient segments;

[0065] Furthermore, the discontinuous truncation means that the coding extraction coefficients are equidistantly truncated according to the number of the obtained control sets to obtain control extraction coefficient segments of equal length;

[0066] Performing original segmentation on the comprehensive monitoring code according to the obtained control extraction coefficient segment to obtain the original coding coefficient segment;

[0067] It should be further explained that, in the specific implementation process, the original segmentation process includes:

[0068] According to the order of discontinuous interception, the obtained control extraction coefficients are uploaded to the comprehensive monitoring code in sequence, and the comprehensive monitoring code is locally convolved by the control extraction coefficients to obtain the original coding coefficient segment, wherein the local convolution means convolving the control extraction coefficients with the corresponding elements of the comprehensive monitoring code, and the convolution result obtained is the original coding coefficient segment;

[0069] According to the order of discontinuous interception, the next control extraction coefficient is uploaded to the comprehensive monitoring code, and the comprehensive monitoring code is locally convolved with the control extraction coefficient to obtain the original coding coefficient segment, until all the control extraction coefficients are locally convolved with the comprehensive monitoring code, and the obtained original coding coefficient segments are statistically analyzed;

[0070] Sorting and counting the original coding coefficient segments obtained based on the order of discontinuous truncation to obtain a local coefficient set, wherein the sorting and counting means arranging the statistical original coding coefficient segments according to the order of discontinuous truncation to obtain the local coefficient set;

[0071] Obtaining the monitoring time corresponding to the local coefficient set, performing time matching on the capture time according to the obtained monitoring time, and marking the successfully matched temporal video data as matching video data, wherein the time matching means obtaining the capture time that is the same as the monitoring time, and marking the temporal video data corresponding to the capture time as matching video data;

[0072] Setting a key interval, performing target framing on the matching video data according to the set key interval, and obtaining a target video frame;

[0073] Further, the key interval represents the interval time set for extracting frame pictures in the matching video data, for example, extracting a frame in the matching video data every n seconds, where the interval n is the key interval; the target framing represents extracting video frames at corresponding time intervals in the matching video data according to the key interval to obtain target video frames; in particular, the key interval is set so that the number of obtained target video frames is the same as the number of original coding coefficient segments in the local coefficient set;

[0074] Performing image-frame conversion on the obtained target video frame to obtain a target component image;

[0075] Furthermore, the image frame conversion means performing image conversion on the target video data obtained after the target frame division in the video processing tool, and converting the obtained target video frame into an image form through the video processing tool;

[0076] The obtained target element diagrams are combined based on the matching video data to obtain a target element sequence, and the obtained target element sequence is associated with the corresponding temporal video data; wherein the order of combination is sorted according to the order of key intervals corresponding to the matching video data.

[0077] The operation analysis module is used to set refinement criteria to perform form calibration and feature selection on the target element sequence to obtain a feature target graph sequence. The specific process includes:

[0078] Set refinement criteria, the refinement criteria include scale criteria and color criteria, the scale criteria include image length and image width, the image length represents the number of pixels in the horizontal direction, the image width represents the number of pixels in the vertical direction, and the color criteria represent the colors allowed to pass. For example, if the color criteria allows red, blue and green to pass, then red, blue and green can be identified;

[0079] Perform formal calibration on the target element sequence according to the set refinement criteria to obtain a standard target graph sequence;

[0080] Furthermore, the formal calibration means unifying the resolution and color clarity of the target element image by refining the criteria, that is, unifying the number of pixels in the horizontal and vertical directions of the image, and unifying the color, obtaining a standard target image, and forming a standard target sequence;

[0081] A selected filter kernel is set according to the obtained local coefficient set, wherein the selected filter kernel is a matrix composed of a plurality of elements, and the number of rows and columns of the matrix are equal, wherein the number of original coding coefficient segments in the local coefficient set is an integer multiple of the number of elements in the selected filter kernel;

[0082] The obtained selected filter kernel is uploaded to the standard target image sequence, and the feature selection is performed on the standard target image sequence by selecting the filter kernel to obtain a feature target image sequence;

[0083] It should be further explained that, in the specific implementation process, the feature selection process includes:

[0084] According to the obtained local coefficient set, the standard target image is matched and divided to obtain a filtered matching area, wherein the matching division means that according to the number of original coding coefficient segments in the local coefficient set, the standard target image is divided into equal areas to obtain areas of equal size, which are recorded as filtered matching areas;

[0085] A link starting point is set in the standard target graph, and the obtained selected filter kernel is uploaded to the link starting point of the standard target graph, wherein the link starting point represents the starting point of feature selection set in the standard target graph. In this embodiment, the filter matching area in the upper left corner of the standard target graph is used as the link starting point;

[0086] The capture and extraction of the filter-removal matching area is performed by selecting the filter-removal core to obtain the capture filter-removal area;

[0087] Further, the capture extraction means convolving the elements in the selected filter kernel with the elements in the filter matching area to obtain the capture filter area, wherein the convolution means convolving the elements in the selected filter kernel with the pixels in the filter matching area;

[0088] The obtained selected filter core is moved to the next filter matching area, and is captured and extracted with the filter matching area until all the filter matching areas are captured and extracted with the selected filter core, and the obtained capture filter areas are sorted, and the capture filter areas are replaced with the original filter matching areas in the standard target image to obtain a feature capture image;

[0089] The obtained feature capture graphs are sorted based on the standard target graph sequence to obtain a feature target graph sequence.

[0090] The dynamic warning module is used to perform video matching on the feature target sequence according to the local coefficient set to obtain the matching video segment, issue a warning instruction to the warning response area through the monitoring center, and obtain regional alarm information. The specific process includes:

[0091] Setting a capture identity point for the obtained feature capture image, wherein the capture identity point is set at a central position within the feature capture image;

[0092] Uploading the obtained local coefficient set to the feature target sequence based on the capture identity point, that is, uploading the original coding coefficient segments in the local coefficient set to the capture identity point in the feature capture map in sequence until the local coefficient set covers each feature capture map of the feature target sequence;

[0093] Perform terminal check on the characteristic target sequence according to the obtained local coefficient set to obtain a terminal coding set;

[0094] It should be further explained that, in the specific implementation process, the terminal verification process includes:

[0095] The original coding coefficient segment in the feature capture image is terminally convolved with the feature capture image to obtain the terminal coding coefficient; wherein the terminal convolution means convolving the original coding coefficient segment with the corresponding element in the feature capture image, and the "element in the feature capture image" means the pixel at each pixel point, and all the elements after the convolution are summed to obtain the terminal coding coefficient;

[0096] Perform terminal convolution on all feature capture graphs based on the feature target sequence until all feature capture graphs complete terminal convolution with the corresponding original coding coefficient segments, sort the obtained terminal coding coefficients according to the order of the obtained feature target sequence, and obtain a terminal coding set;

[0097] Performing digital-to-graphic conversion on the obtained terminal code set to obtain a terminal monitoring curve graph;

[0098] Further, the digital-to-graph conversion means generating a coding coefficient curve according to the terminal coding set, constructing a two-dimensional rectangular coordinate system of time with respect to the terminal coding coefficient, uploading the obtained coding coefficient curve to the two-dimensional rectangular coordinate system, and obtaining a terminal monitoring curve graph;

[0099] Setting a threshold comparison axis for the obtained terminal monitoring curve graph, wherein the threshold comparison axis is a straight line parallel to the horizontal axis of the terminal monitoring curve graph, and the straight line represents a preset threshold value;

[0100] The threshold comparison axis is uploaded to the terminal monitoring curve chart, and the threshold comparison is performed on the terminal monitoring curve chart through the threshold comparison axis to obtain the primary warning coefficient and the safety range coefficient;

[0101] Further, the threshold comparison is represented in the terminal monitoring curve diagram, and the portion of the coding coefficient curve that is greater than the threshold comparison axis is recorded as the primary warning coefficient, wherein the "portion greater than the threshold comparison axis" represents the terminal coding coefficient greater than the threshold comparison axis; the portion less than or equal to the threshold comparison axis is recorded as the safety range coefficient, wherein the "portion less than or equal to the threshold comparison axis" represents the terminal coding coefficient less than or equal to the threshold comparison axis;

[0102] According to the primary warning coefficient, the characteristic target sequence is marked with warning marks to obtain a color warning map;

[0103] It should be further explained that, in the specific implementation process, the early warning marking process includes:

[0104] Obtain a feature capture map corresponding to the primary warning coefficient, record the obtained feature capture map as a primary warning map, mark the primary warning map with a warning color, and record the primary warning map marked with the warning color as a color warning map, wherein the warning color indicates that the primary warning map is marked with a warning color in the feature target sequence. In this embodiment, the marked warning color is red, then in the feature target sequence, the primary warning map is marked with red and recorded as a color warning map;

[0105] Based on the characteristic target sequence, the color warning image is subjected to danger determination to obtain a danger warning sequence;

[0106] Furthermore, the risk determination process includes:

[0107] Count the color warning images in the characteristic target sequence, obtain the number of adjacent color warning images, and record it as the number of continuous warnings;

[0108] Set a warning number threshold, where the warning number threshold is a positive integer;

[0109] The number of consecutive warnings is determined by the warning number threshold to obtain the danger warning degree;

[0110] Furthermore, when the number of continuous warnings is greater than or equal to the warning number threshold, the continuous color warning images corresponding to the continuous warning numbers are recorded as continuous warning images, and the characteristic target sequence containing the continuous warning images is recorded as a danger warning sequence;

[0111] When the number of consecutive warnings is less than the warning number threshold, the consecutive color warning images corresponding to the consecutive warning numbers are recorded as qualified images, and the feature target sequence containing the qualified images is recorded as a qualified warning sequence;

[0112] According to the continuous warning map, the local coefficient set is matched to obtain the warning coding coefficient, and the comprehensive monitoring data corresponding to the warning coding coefficient is obtained, which is recorded as the warning monitoring data, wherein the obtained warning coding coefficient is the original coding coefficient segment within the capture identity point corresponding to the continuous warning map, and the comprehensive monitoring data corresponding to the warning coding coefficient is matched according to the comprehensive monitoring code, that is, the warning monitoring data;

[0113] Acquire temporal video data corresponding to the danger warning sequence, perform video matching on continuous warning images based on key intervals, and obtain matching video segments corresponding to the continuous warning images;

[0114] Furthermore, the video matching means that the part of the video corresponding to the continuous warning images is recorded as a matching video segment according to the key interval in the temporal video data;

[0115] Acquire plant components corresponding to the danger warning sequence, and mark the acquired plant components as warning response areas;

[0116] Associating the obtained matching video segments and early warning monitoring data with corresponding warning response areas;

[0117] The obtained matching video segments and early warning monitoring data are uploaded to the monitoring center, and an early warning instruction is issued to the warning response area through the monitoring center. The warning response area performs strategy conversion on the matching video segments and early warning monitoring data according to the obtained early warning instruction to obtain regional alarm information;

[0118] Furthermore, the early warning instruction indicates an alarm for the warning response area, notifies the monitoring center to issue a safety warning for the factory components corresponding to the warning response area, and obtains the matching video segments and early warning monitoring data corresponding to the factory components for recording to obtain regional alarm information. The regional alarm information indicates the danger monitoring data corresponding to the warning response area at the corresponding monitoring time and the video data corresponding to the danger monitoring data, that is, the early warning monitoring data and the matching video segments. The matching video segments and early warning monitoring data corresponding to the warning response area are recorded as regional alarm information. According to the obtained regional alarm information, the safety inspection speed of the management personnel for the warning response area can be increased, and the location and fault inspection can be intuitively carried out through the matching video segments. At the same time, the early warning monitoring data and matching video segments of the warning response area are obtained. If there is the same early warning monitoring data at the next capture time, a safety alarm can be issued by the monitoring center, and a safety inspection can be carried out by the management personnel.

[0119] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A smart factory safety early warning system based on artificial intelligence, including a monitoring center, characterized in that: The monitoring center is connected to a smart acquisition module, a factory processing module, an operation analysis module and a dynamic early warning module; The smart acquisition module is used to collect temporal video data and comprehensive monitoring data; The factory processing module is used to perform code conversion on the comprehensive monitoring data, obtain the comprehensive monitoring code, set the code extraction coefficient for compression change and discontinuous interception, obtain the control extraction coefficient segment, perform original segmentation and sorting statistics on the comprehensive monitoring code according to the control extraction coefficient segment, obtain the local coefficient set, perform time pairing on the local coefficient set, obtain matching video data, set the key interval to perform target frame segmentation and conversion combination on the matching video data, and obtain the target element sequence; The operation analysis module is used to set refinement criteria to perform form calibration on the target element sequence to obtain a standard target graph sequence, set selection and filtering to perform feature selection on the standard target graph sequence to obtain a feature target graph sequence; The dynamic warning module is used to perform terminal verification on the characteristic target sequence according to the local coefficient set, obtain the terminal code set, perform digital-to-image conversion and danger judgment on the terminal code set, obtain the danger warning sequence, perform video matching on the danger warning sequence, obtain the matching video segment, upload the matching video segment and the warning monitoring data to the monitoring center, issue a warning instruction to the warning response area through the monitoring center and perform strategy conversion to obtain the regional alarm information.

2. According to claim 1, the intelligent factory safety early warning system based on artificial intelligence is characterized in that: The process of collecting temporal video data and comprehensive monitoring data by the intelligent acquisition module includes: Set up real-time capture end and intelligent collection end; The real-time capture terminal is used to collect video of factory components to obtain temporal video data, and the collected temporal video data is time-stamped to obtain the capture time; The factory components are collected at the edge through the intelligent collection terminal to obtain comprehensive monitoring data, and the collected comprehensive monitoring data is time-stamped to obtain the monitoring time; The synthetic monitoring data is associated with the corresponding temporal video data based on the monitoring time.

3. The intelligent factory safety early warning system based on artificial intelligence according to claim 2 is characterized in that: The process of obtaining the regulation extraction coefficient segment includes: Performing code conversion on the obtained comprehensive monitoring data to obtain a comprehensive monitoring code; Set a coding extraction coefficient, compress and change the coding extraction coefficient to obtain a compression index, compress and extract the coding extraction coefficient according to the compression index to obtain a compression interval, regulate and allocate the coding extraction coefficient according to the compression index and the compression interval to obtain a regulation set number; The coding extraction coefficients are intermittently truncated according to the number of control sets to obtain control extraction coefficient segments.

4. The artificial intelligence-based smart factory safety early warning system according to claim 3 is characterized in that: The process of obtaining matching video data includes: Performing original segmentation on the comprehensive monitoring code according to the obtained control extraction coefficient segment to obtain the original coding coefficient segment; The obtained original coding coefficient segments are sorted and counted based on the order of discontinuous truncation to obtain a local coefficient set; The monitoring time corresponding to the local coefficient set is obtained, the capture time is time-matched according to the monitoring time, and the successfully matched temporal video data is marked as the matching video data.

5. The artificial intelligence-based smart factory safety early warning system according to claim 4 is characterized in that: The process of obtaining the target element sequence includes: Set a key interval, perform target framing on the matching video data according to the key interval, obtain a target video frame, perform image-frame conversion on the target video frame, and obtain a target component image; The obtained target component graphs are combined based on the matching video data to obtain a target component sequence, and the obtained target component sequence is associated with the corresponding temporal video data.

6. The intelligent factory safety early warning system based on artificial intelligence according to claim 5 is characterized in that: The process of obtaining the feature target graph sequence includes: Setting a refinement criterion, performing formal calibration on the target element sequence according to the refinement criterion, and obtaining a standard target graph sequence; A filter kernel is selected according to the local coefficient set setting, and the selected filter kernel is uploaded to the standard target graph sequence. Feature selection is performed on the standard target graph sequence by selecting the filter kernel to obtain a feature target graph sequence, wherein the feature target graph sequence includes the standard target graph.

7. The intelligent factory safety early warning system based on artificial intelligence according to claim 6 is characterized in that: The process of feature selection by selecting and filtering the standard target graph sequence includes: Matching and dividing the standard target image according to the local coefficient set to obtain the filtered matching area; Set the link start point, upload the selected filter core to the link start point, capture and extract the filter matching area by selecting the filter core, and obtain the capture filter area; Move the selected filter core to the next filter matching area, and capture and extract it with the filter matching area until all the filter matching areas are captured and extracted with the selected filter core, and replace the original filter matching area with the captured filter area in the standard target image to obtain a feature capture image; The obtained feature capture graphs are sorted based on the standard target graph sequence to obtain a feature target graph sequence.

8. The artificial intelligence-based smart factory safety early warning system according to claim 7 is characterized in that: The process of obtaining a hazard warning sequence includes: Set a capture identity point according to the feature capture map, upload a local coefficient set to a feature target sequence based on the capture identity point, perform terminal verification on the feature target sequence according to the local coefficient set, and obtain a terminal code set; Perform digital-to-graphic conversion on the terminal code set to obtain a terminal monitoring curve chart, set a threshold comparison axis, perform threshold comparison on the terminal monitoring curve chart through the threshold comparison axis, and obtain a primary warning coefficient; According to the primary warning coefficient, the characteristic target sequence is marked with warning marks to obtain a color warning map; Danger determination is performed on the color warning image based on the characteristic target sequence to obtain a danger warning sequence, wherein the danger warning sequence includes continuous warning images.

9. The artificial intelligence-based smart factory safety early warning system according to claim 8 is characterized in that: The process of obtaining regional alert information includes: Match the local coefficient set according to the continuous warning map to obtain the warning monitoring data; Acquire temporal video data corresponding to the danger warning sequence, perform video matching on continuous warning images based on key intervals, and obtain matching video segments; Acquire plant components corresponding to the danger warning sequence, and mark the acquired plant components as warning response areas; The matching video segments and warning monitoring data are uploaded to the monitoring center, and a warning instruction is issued to the warning response area through the monitoring center. The warning response area performs strategy transformation on the matching video segments and warning monitoring data according to the obtained warning instruction to obtain regional alarm information.

Images