Video monitoring system for wind power plant production place

Abstract

The invention discloses a video monitoring system for a wind power plant production place, and relates to the technical field of wind power plant monitoring, the video monitoring system comprises a control center, the control center is connected with a database, a video data acquisition module, a video data analysis module, a video data transmission module and an early warning module; the database is used for storing an abnormal video, a conventional video and an abnormal judgment model; the video data acquisition module is used for acquiring monitoring video data; the video data analysis module is used for judging whether the monitoring video data is abnormal or not through an abnormality judgment model, if the monitoring video data is abnormal, the monitoring video data is an abnormal video, and if the monitoring video data is not abnormal, the monitoring video data is a conventional video; and the video data transmission module is used for submitting the abnormal video and the conventional video to a database for storage, and in the process of submitting the conventional video, performing frame extraction processing on the conventional video, performing expansive duplicate removal on the conventional video after the frame extraction processing, and generating a second-order conventional video.

Description

Technical Field

[0001] The invention relates to the technical field of wind farm monitoring, and in particular to a video monitoring system used in a wind farm production site. Background Art

[0002] Wind energy is a clean, pollution-free, renewable energy source that has been used by people for a long time. Wind energy reserves are huge, so it is increasingly valued by countries around the world. Wind farms are important places for converting wind energy into electrical energy, and their safety and operating conditions have always been valued by people.

[0003] When monitoring the production site of a wind farm to understand its safety and operating conditions, a large amount of surveillance video data is generated by the monitoring. How to better store the surveillance video data, how to store more surveillance video data in a limited storage space, and improve the utilization of storage resources are issues that we currently need to consider. Summary of the invention

[0004] In order to solve the above problems, an object of the present invention is to provide a video monitoring system for a wind farm production site.

[0005] The object of the present invention can be achieved by the following technical solutions: A video monitoring system for a wind farm production site comprises a control center, wherein the control center is communicatively connected to a database, a video data acquisition module, a video data analysis module, a video data transmission module and an early warning module;

[0006] The database is used to store abnormal videos, regular videos and abnormal judgment models;

[0007] The video data acquisition module is used to obtain monitoring video data of the wind farm production site;

[0008] The video data analysis module is used to determine whether the monitoring video data has an abnormality through the abnormality judgment model. If an abnormality exists, the monitoring video data is an abnormal video and generates corresponding warning information. If no abnormality exists, the monitoring video data is a regular video.

[0009] The video data transmission module is used to submit abnormal videos and regular videos to the database for storage, and in the process of submitting regular videos, perform frame extraction processing on the regular videos, perform expansion deduplication on the regular videos after frame extraction processing, and generate second-order regular videos;

[0010] The early warning module is used to process early warning information.

[0011] Furthermore, the production site of the wind farm is divided into a number of monitoring areas, and the monitoring areas are numbered; there are overlapping areas between the monitoring areas, and the numbers are unique; the monitoring areas are monitored by the monitoring device.

[0012] Further, the database is used to store abnormal videos, regular videos, and abnormal judgment models;

[0013] The database includes an abnormal unit, a regular unit, and a model unit;

[0014] The abnormal unit includes abnormal video data and an abnormal number, and is used to store abnormal videos; the abnormal number is the monitoring area number corresponding to the abnormal video data;

[0015] The regular unit includes regular video data and a regular number, and is used to store regular videos; the regular number is the monitoring area number corresponding to the regular video data;

[0016] The model unit includes a model unit and a self-learning unit; the model unit is used to store abnormal judgment models; the self-learning unit generates new abnormal judgment models according to the abnormal video data through a self-supervised learning method, and submits the new abnormal judgment models to the model unit for storage.

[0017] Further, the process of obtaining the monitoring video data of the wind farm production site includes:

[0018] The video data acquisition module is composed of several monitoring devices;

[0019] The monitoring device includes a camera, an anemometer, and a microprocessor;

[0020] Set the acquisition period T1, the acquisition period T2, and the acquisition period T3;

[0021] Set the wind speed value w1 and the wind speed value w2, and w1 < w2;

[0022] Obtain the current wind speed through the anemometer in the monitoring device, and mark the wind speed as w;

[0023] When w < w1, the monitoring device generates the monitoring video data of the wind farm production site according to the acquisition period T1;

[0024] When w1 ≤ w < w2, the monitoring device generates the monitoring video data of the wind farm production site according to the acquisition period T2;

[0025] When w ≥ w2, the monitoring device generates the monitoring video data of the wind farm production site according to the acquisition period T3.

[0026] Further, the process of judging whether there is an abnormality in the monitoring video data according to the abnormal judgment model includes:

[0027] The abnormality judgment model generates result information according to input data; the input data is video data, and the result information includes the presence of abnormality and the absence of abnormality;

[0028] The monitoring video data is input into the abnormality judgment model, and the abnormality judgment model generates corresponding result information according to the input monitoring video data;

[0029] If the result information indicates that an abnormality exists, the monitoring video data is an abnormal video, and the monitoring area number corresponding to the abnormal video is obtained to generate an early warning message, and then the early warning message is submitted to the early warning module, and the early warning module makes corresponding processing;

[0030] If the result information shows that there is no abnormality, the surveillance video data is a regular video.

[0031] Furthermore, the process of submitting abnormal videos and regular videos to the database for storage includes:

[0032] Obtaining the monitoring area number corresponding to the abnormal video, and submitting the abnormal video and the monitoring area number to the database for storage;

[0033] The monitoring area number corresponding to the regular video is obtained, and the regular video and the monitoring area number are submitted to the database for storage.

[0034] Furthermore, the process of performing frame extraction processing on a conventional video includes:

[0035] Split a regular video into several images in frames and sort them according to time to generate a video image set;

[0036] Obtaining the time corresponding to the first frame image and the last frame image of the regular video, and generating a valid time zone of the regular video according to the time;

[0037] If the time corresponding to the first frame of a regular video is time a, and the time corresponding to the last frame of a regular video is time b, then the effective time zone of the regular video is [time a, time b];

[0038] The frame extraction is to delete a certain frame of image;

[0039] The video image set = {P1, P2, ..., P n}, where n is a positive integer;

[0040] Step a1: obtaining two images with adjacent subscripts, comparing the subscripts, marking the image corresponding to the smaller subscript as the left-phase image, and marking the image corresponding to the larger subscript as the right-phase image;

[0041] Step a2: Obtain the similarity between the left-phase image and the right-phase image;

[0042] Step a3: setting a similarity threshold. If the similarity is greater than or equal to the similarity threshold, a similar image pair is generated according to the left-phase image and the right-phase image; otherwise, no operation is required; the similar image pair = {left-phase image, right-phase image};

[0043] Step a4: Repeat steps a1 to a3 until no similar image pairs can be generated;

[0044] Step a5: Get similar image pairs in the video image set, and randomly delete one image in the similar image pair to complete the frame extraction operation; if there is a similar image pair A{P i-1 , P i} and similar image pairs B{P i , P i+1}, if the same image exists in the similar image pair A and the similar image pair B, then the similar image pair A and the similar image pair B delete different images when performing the frame extraction operation, that is, the similar image pair A and the similar image pair B cannot delete the image P at the same time i ;

[0045] Step a5: sort the video image set after the frame extraction operation according to time, generate a new video image set, and overwrite the original video image set with the new video image set.

[0046] Furthermore, the process of obtaining the similarity between the left-phase image and the right-phase image includes:

[0047] Establishing a three-dimensional rectangular coordinate system, wherein the three-dimensional rectangular coordinate system includes an x-axis, a y-axis, and a z-axis;

[0048] Construct an xy-axis two-dimensional coordinate system according to the x-axis and the y-axis, construct an xz-axis two-dimensional coordinate system according to the x-axis and the z-axis, and construct a yz-axis two-dimensional coordinate system according to the y-axis and the z-axis;

[0049] Map the left phase image to the yz axis two-dimensional coordinate system, and map the right phase image to the xz axis two-dimensional coordinate system, so that the pixels in the left phase image and the right phase image have their corresponding coordinates;

[0050] Grayscale processing is performed on the left-phase image and the right-phase image;

[0051] Obtain the grayscale values ​​and the corresponding pixel numbers in the left-phase image to generate a left-phase grayscale set; the left-phase grayscale set = {L_Gray1 = L_num1, L_Gray2 = L_num2, ..., L_Gray n =L_num n}, where n is a positive integer, L_Gray i Refers to the grayscale value in the left phase image, L_num i Refers to L_Grayi The corresponding number of pixels;

[0052] Get the grayscale values ​​and the corresponding pixel numbers in the right phase image to generate a right phase grayscale set; the right phase grayscale set = {R_Gray1 = R_num1, R_Gray2 = R_num2, ..., R_Gray n =R_num n}, where n is a positive integer, R_Gray i Refers to the grayscale value in the right phase image, R_num i Refers to R_Gray i The corresponding number of pixels;

[0053] Obtain the overlap between the left-phase grayscale set and the right-phase grayscale set;

[0054] Set the coincidence threshold;

[0055] If L_Gray exists in the left grayscale set a =L_num a , R_Gray exists in the right phase grayscale concentration b =R_num b , L_Gray a With R_Gray b Consistent, and |L_num a -R_num b | is less than the overlap threshold, then L_Gray a =L_num a and R_Gray b =R_num b Mark as coincident factor; otherwise, mark as non-coincident factor;

[0056] Get the number of coincidence factors in the left-phase grayscale concentration, marked as LNUM_Y; get the number of non-coincidence factors in the left-phase grayscale concentration, marked as LNUM_N;

[0057] Get the number of coincidence factors in the right phase grayscale concentration, marked as RNUM_Y; get the number of non-coincidence factors in the right phase grayscale concentration, marked as RNUM_N;

[0058] Obtain the degree of overlap between the left-phase grayscale set and the right-phase grayscale set according to the number of overlap factors and the number of non-overlap factors;

[0059]

[0060] Among them, the SMALLER(A, B) function is used to return the smaller value between A and B;

[0061] Obtaining grayscale values ​​in the left-phase grayscale set in sequence, and marking the grayscale values ​​as target grayscale values;

[0062] Set the derivative value to generate the derivative gray value according to the target gray value and the derivative value;

[0063] If the target gray value is tarGray, and the derivative value der is set, the derivative gray value derGray is generated according to the target gray value and the derivative value, and the derivative gray value derGray∈[tarGray-der, tarGray+der], if tarGray-der is less than 0, tarGray-der is set to 0, and if tarGray+der is greater than 255, tarGray+der is set to 255;

[0064] Obtaining pixel coordinates corresponding to target grayscale values ​​in the left-phase image, marking the coordinates as left-phase coordinates, and generating a left-phase coordinate set;

[0065] The left phase coordinate set = {left phase coordinate L1 , left phase coordinate a2 , ..., left phase coordinate Ln};

[0066] Obtaining pixel coordinates corresponding to the derived grayscale values ​​in the right phase image, marking the coordinates as right phase coordinates, and generating a right phase coordinate set;

[0067] The right phase coordinate set = {right phase coordinate R1 , right phase coordinate R2 , ..., right phase coordinate Rn};

[0068] Generate a number of grayscale vectors according to the left phase coordinate set and the right phase coordinate set; the grayscale vectors include grayscale vectors L1，R1 , grayscale vector L1，R2 , ..., grayscale vector L1，Rn , ..., grayscale vector Li，R1 , grayscale vector Li，R2 , ..., grayscale vector Li，Rn , ..., grayscale vector Ln，R1 , grayscale vector Ln，R2 , ..., grayscale vector Ln，Rn ;

[0069] Obtaining grayscale vectors in sequence, marking the grayscale vectors as target vectors, and marking other grayscale vectors except the target vector as non-target vectors;

[0070] Obtain a non-target vector parallel to the target vector, obtain the number of the non-target vectors, and mark the number as a parallel value para of the target vector;

[0071] Get the parallel values ​​corresponding to each grayscale vector and generate the parallel set PARA;

[0072] The parallel set PARA = {para1, para2, ..., para n};

[0073] Set the parallel error paraDev, and obtain the quantity difference interval according to the parallel error and the left phase grayscale set;

[0074]

[0075] Set weight coefficient θ1 and weight coefficient θ2, and θ1<θ2;

[0076] According to the weight coefficient θ1, the weight coefficient θ2, the coincidence degree, the parallel set and the quantity difference interval, the similarity between the left phase image and the right phase image is obtained;

[0077]

[0079] Among them, if there is a value in A that is consistent with that in B in the EXIXST(A, B) function, it returns 1, otherwise it returns 0.

[0080] Furthermore, the process of performing expansion deduplication on the regular video after the frame extraction process to generate a second-order regular video includes:

[0081] If there is an overlapping area between monitoring area A and monitoring area B, then the regular videos corresponding to monitoring area A and monitoring area B have a regional relationship;

[0082] The expansive deduplication is to stitch together a number of regular videos that have a regional relationship within the same effective time zone;

[0083] Obtain a regular video with a consistent effective time zone, and mark the regular video as a same-domain video;

[0084] According to the overlapping areas between the monitoring areas, it is determined whether there is a regional relationship between the videos in the same domain;

[0085] If there is no regional relationship between the videos in the same domain, the monitoring area numbers corresponding to the videos in the same domain are obtained respectively, and the videos in the same domain and their corresponding monitoring area numbers are packaged and sent to the conventional unit for storage;

[0086] If there is a regional relationship between some of the same-domain videos and no regional relationship between the other part of the same-domain videos, the monitoring area numbers corresponding to the same-domain videos without regional relationship are obtained, the same-domain videos and their corresponding monitoring area numbers are packaged and sent to the regular unit for storage, and then the same-domain videos with regional relationship are obtained for expansive deduplication to generate secondary regular videos, and the secondary regular videos and their corresponding monitoring area numbers are sent to the regular unit for storage;

[0087] If all the same-domain videos have regional relationships, then the same-domain videos with regional relationships are expansively deduplicated to generate secondary regular videos, and the secondary regular videos and their corresponding monitoring area numbers are sent to the regular unit for storage;

[0088] The process of stitching together a plurality of conventional videos having regional relationships within the same effective time zone includes:

[0089] It is assumed that there is a regional relationship between the same-domain video A and the same-domain video B;

[0090] Obtain video image set A corresponding to same-domain video A, and obtain video image set B corresponding to same-domain video B;

[0091] The video image set A={P a1 , P a2 , ..., P an}, where n is a positive integer;

[0092] The video image set B={P b1 , P b2 , ..., P bn}, where n is a positive integer;

[0093] Obtaining a monitoring area corresponding to the same-domain video A, and marking the monitoring area as monitoring area A;

[0094] Obtaining a monitoring area corresponding to the same-domain video B, and marking the monitoring area as monitoring area B;

[0095] Obtaining an overlapping area between monitoring area A and monitoring area B, and marking the overlapping area as overlapping area AB;

[0096] The frame numbers corresponding to the video image set A and the video image set B are obtained respectively and compared, and the video image set with more frames is marked as the basic image set, and the video image set with less frames is marked as the interpolated frame image set; if the frame numbers are the same, one video image set is randomly marked as the basic image set, and the other video image set is marked as the interpolated frame image set;

[0097] Generate a number of second-order image pairs according to the basic image set and the complemented frame image set;

[0098] The second-order image pair = {second-order image i , the second-order image j}, where the second-order image i and the second-order image j In no particular order, and second-order images i and the second-order image j The images are opposite to each other;

[0099] Preset interpolation frame image set = {P a1 , P a2 , P a3 , P a4}, basic image set = {P b1 , P b2 , P b3 , P b4 , P b5 , P b6};

[0100] Get the time corresponding to each image in the interpolated frame image set and record it as T a1 , T a2 , T a3 and T a4 , where T a1 <T a2 <T a3 <T a4 , and a1 <a2<a3<a4;

[0101] Get the time corresponding to each image in the basic image set and record it as T b1 , T b2 , T b3 , T b4 , T b5 and T b6 , T b1 <T b2 <T b3 <T b4 <T b5 <T b6 , and b1 <b2<b3<b4<b5<b6;

[0102] Preset time axis T a1 <T b1 <T b2 <T a2 =T b3 <T b4 <T b5 <T b6 <T a3 <T a4 ;

[0103] According to the time axis, each image in the basic image set and the image in the complementary frame image set closest to its own time form a second-order image pair; if the number of images in the basic image set that obtain the complementary frame image set closest to its own time is 2, then the image in the basic image set and the image in the complementary frame image set that is larger than its own time and has the smallest subscript form a second-order image pair;

[0104] If there is an image in the supplementary frame image set that does not form a second-order image pair with any image in the basic image set, then according to the principle that each image in the basic image set forms a second-order image pair with an image in the supplementary frame image set, the image in the supplementary frame image set and the image in the basic image set closest to itself form a second-order image pair;

[0105] That is, generate a second-order image pair 1{second-order image b1 , the second-order image a1}、Second-order image pair 2{Second-order image b2 , the second-order image a2}、Second-order image pair 3{Second-order image b3 , the second-order image a2}、Second-order image pair 4{Second-order image b4 , the second-order image a2}、Second-order image pair 5{Second-order image b5 , the second-order image a3}、Second-order image pair 6{Second-order image b6 , the second-order image a3}、Second-order image pair 7{Second-order image a4 , the second-order image b6};

[0106] generating a plurality of final frame images according to a plurality of second-order image pairs;

[0107] Acquire any second-order image in each second-order image pair to generate an image set to be stitched;

[0108] Sequentially acquire images in the image set to be stitched, and mark the images as stitched images;

[0109] Obtain pixel points corresponding to the overlapped area AB in the stitched image, mark the pixel points as pixels to be covered, and mark other pixel points as stitched pixel points;

[0110] Establish a two-dimensional coordinate system A, map the stitched image to the two-dimensional coordinate system A, make each pixel point have its corresponding coordinate, and generate the coordinate relationship between the stitched pixel point and the pixel point to be covered;

[0111] Obtain an opposite side image corresponding to the stitched image, obtain pixel points corresponding to the overlapped area AB in the opposite side image, mark the pixel points as top-package pixel points, and mark other pixel points as opposite side pixel points;

[0112] Establish a two-dimensional coordinate system B, map the opposite edge image into the two-dimensional coordinate system B, so that each pixel has its corresponding coordinate, and generate the coordinate relationship between the opposite edge pixel and the top package pixel;

[0113] According to the overlapped area AB, the two-dimensional coordinate system A and the two-dimensional coordinate system B, the top-pack pixel points and the to-be-top-pack pixel points are matched, so that the top-pack pixel points and the to-be-top-pack pixel points are in a one-to-one correspondence, that is, one top-pack pixel point corresponds to and only corresponds to one to-be-top-pack pixel point;

[0114] Obtain the pixels to be covered in sequence, and mark the pixels to be covered as target pixels;

[0115] Obtain the top pixel point corresponding to the target pixel point, and mark the top pixel point as a bidding pixel point;

[0116] Get the RGB value of the target pixel and the RGB value of the bidding pixel;

[0117] Compare the RGB value of the target pixel and the RGB value of the bidding pixel, and obtain the pixel corresponding to the larger RGB value as the expected pixel; if the RGB value of the target pixel is consistent with the RGB value of the bidding pixel, randomly obtain one of the pixels as the expected pixel;

[0118] Generate a corresponding final frame image according to the coordinate relationship between the stitched pixel points and the pixel points to be covered, the coordinate relationship between the opposite edge pixel points and the pixel points to be covered, and the expected pixel points;

[0119] According to the time sequence of the basic image set, a plurality of final frame images are sorted once; according to the time sequence of the supplementary frame image set, a plurality of final frame images are sorted twice to generate a second-order conventional video.

[0120] Furthermore, the process of processing the warning information includes:

[0121] The monitoring area number in the warning information is sent to the user, informing the user that an abnormality exists in the monitoring area corresponding to the monitoring area number, and the user takes corresponding measures.

[0122] Compared with the prior art, the present invention has the following beneficial effects:

[0123] 1. By performing frame extraction operation on surveillance video data and deleting similar images, the size of surveillance video data can be reduced;

[0124] 2. Stitch surveillance videos that are located in the same time zone and have a certain overlap in surveillance content, and stitch multiple surveillance videos into a new surveillance video, which increases the content volume of the surveillance video and improves the utilization rate of storage space. BRIEF DESCRIPTION OF THE DRAWINGS

[0125] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0126] like Figure 1 As shown, a video monitoring system for a wind farm production site includes a control center, and the control center is connected to a database, a video data acquisition module, a video data analysis module, a video data transmission module and an early warning module;

[0127] The database is used to store abnormal videos, regular videos and abnormal judgment models;

[0128] The video data acquisition module is used to obtain monitoring video data of the wind farm production site;

[0129] The video data analysis module is used to determine whether the monitoring video data has an abnormality through the abnormality judgment model. If an abnormality exists, the monitoring video data is an abnormal video and generates corresponding warning information. If no abnormality exists, the monitoring video data is a regular video.

[0130] The video data transmission module is used to submit abnormal videos and regular videos to the database for storage, and in the process of submitting regular videos, perform frame extraction processing on the regular videos, perform expansion deduplication on the regular videos after frame extraction processing, and generate second-order regular videos;

[0131] The warning module is used to process warning information;

[0132] It should be further explained that, in the specific implementation process, the production site of the wind farm is divided into several monitoring areas, and the several monitoring areas are numbered; there are overlapping areas between the monitoring areas, and the numbers are unique; the monitoring areas are monitored by the monitoring device;

[0133] It should be further explained that, in the specific implementation process, the database is used to store abnormal videos, regular videos and abnormal judgment models;

[0134] The database includes abnormal units, regular units and model units;

[0135] The abnormal unit includes abnormal video data and an abnormal number, which is used to store abnormal videos; the abnormal number is the monitoring area number corresponding to the abnormal video data;

[0136] The conventional unit includes conventional video data and a conventional number, and is used to store conventional videos; the conventional number is the monitoring area number corresponding to the conventional video data;

[0137] The model unit includes a model unit and a self-learning unit; the model unit is used to store an anomaly judgment model; the self-learning unit generates a new anomaly judgment model according to the anomaly video data through a self-supervised learning method, and submits the new anomaly judgment model to the model unit for storage;

[0138] It should be further noted that in the specific implementation process, the process of obtaining the monitoring video data of the wind farm production site includes:

[0139] The video data acquisition module is composed of several monitoring devices;

[0140] The monitoring device includes a camera, a wind speed sensor and a microprocessor;

[0141] Set the acquisition period T1, the acquisition period T2 and the acquisition period T3;

[0142] Set the wind speed value w1 and the wind speed value w2, and w1 < w2;

[0143] Obtain the current wind speed through the wind speed sensor in the monitoring device, and mark the wind speed as w;

[0144] When w < w1, the monitoring device generates the monitoring video data of the wind farm production site according to the acquisition period T1;

[0145] When w1 ≤ w < w2, the monitoring device generates the monitoring video data of the wind farm production site according to the acquisition period T2;

[0146] When w ≥ w2, the monitoring device generates the monitoring video data of the wind farm production site according to the acquisition period T3;

[0147] It should be further noted that in the specific implementation process, the process of judging whether the monitoring video data is abnormal according to the anomaly judgment model includes:

[0148] The anomaly judgment model generates result information according to the input data; the input data is video data, and the result information includes abnormal and normal;

[0149] Input the monitoring video data into the anomaly judgment model, and the anomaly judgment model generates corresponding result information according to the input monitoring video data;

[0150] If the result information is abnormal, then the monitoring video data is abnormal video, and obtain the monitoring area number corresponding to the abnormal video to generate a warning information, and then submit the warning information to the warning module, and the warning module makes corresponding processing;

[0151] If the result information is normal, then the monitoring video data is normal video;

[0152] It should be further explained that, in the specific implementation process, the process of submitting abnormal videos and regular videos to the database for storage includes:

[0153] Obtaining the monitoring area number corresponding to the abnormal video, and submitting the abnormal video and the monitoring area number to the database for storage;

[0154] Obtaining the monitoring area number corresponding to the regular video, and submitting the regular video and the monitoring area number to the database for storage;

[0155] It should be further explained that, in the specific implementation process, the process of performing frame extraction processing on a conventional video includes:

[0156] Split a regular video into several images in frames and sort them according to time to generate a video image set;

[0157] Obtaining the time corresponding to the first frame image and the last frame image of the regular video, and generating a valid time zone of the regular video according to the time;

[0158] If the time corresponding to the first frame of a regular video is time a, and the time corresponding to the last frame of a regular video is time b, then the effective time zone of the regular video is [time a, time b];

[0159] The frame extraction is to delete a certain frame of image;

[0160] The video image set = {P1, P2, ..., P n}, where n is a positive integer;

[0161] Step a1: obtaining two images with adjacent subscripts, comparing the subscripts, marking the image corresponding to the smaller subscript as the left-phase image, and marking the image corresponding to the larger subscript as the right-phase image;

[0162] Step a2: Obtain the similarity between the left-phase image and the right-phase image;

[0163] Step a3: setting a similarity threshold. If the similarity is greater than or equal to the similarity threshold, a similar image pair is generated according to the left-phase image and the right-phase image; otherwise, no operation is required; the similar image pair = {left-phase image, right-phase image};

[0164] Step a4: Repeat steps a1 to a3 until no similar image pairs can be generated;

[0165] Step a5: Get similar image pairs in the video image set, and randomly delete one image in the similar image pair to complete the frame extraction operation; if there is a similar image pair A{P i-1 , P i} and similar image pairs B{P i , Pi+1}, if the same image exists in the similar image pair A and the similar image pair B, then the similar image pair A and the similar image pair B delete different images when performing the frame extraction operation, that is, the similar image pair A and the similar image pair B cannot delete the image P at the same time. i ;

[0166] Step a5: sorting the video image set after the frame extraction operation according to time to generate a new video image set, and overwriting the original video image set with the new video image set;

[0167] The process of obtaining the similarity between the left-phase image and the right-phase image includes:

[0168] Establishing a three-dimensional rectangular coordinate system, wherein the three-dimensional rectangular coordinate system includes an x-axis, a y-axis, and a z-axis;

[0169] Construct an xy-axis two-dimensional coordinate system according to the x-axis and the y-axis, construct an xz-axis two-dimensional coordinate system according to the x-axis and the z-axis, and construct a yz-axis two-dimensional coordinate system according to the y-axis and the z-axis;

[0170] Map the left phase image to the yz axis two-dimensional coordinate system, and map the right phase image to the xz axis two-dimensional coordinate system, so that the pixels in the left phase image and the right phase image have their corresponding coordinates;

[0171] Grayscale processing is performed on the left-phase image and the right-phase image;

[0172] Obtain the grayscale values ​​and the corresponding pixel numbers in the left-phase image to generate a left-phase grayscale set; the left-phase grayscale set = {L_Gray1 = L_num1, L_Gray2 = L_num2, ..., L_Gray n =L_num n}, where n is a positive integer, L_Gray i Refers to the grayscale value in the left phase image, L_num i Refers to L_Gray i The corresponding number of pixels;

[0173] Get the grayscale values ​​and the corresponding pixel numbers in the right phase image to generate a right phase grayscale set; the right phase grayscale set = {R_Gray1 = R_num1, R_Gray2 = R_num2, ..., R_Gray n =R_num n}, where n is a positive integer, R_Gray i Refers to the grayscale value in the right phase image, R_num i Refers to R_Gray i The corresponding number of pixels;

[0174] Obtain the overlap between the left-phase grayscale set and the right-phase grayscale set;

[0175] Set the coincidence threshold;

[0176] If L_Gray exists in the left grayscale set a =L_num a , R_Gray exists in the right phase grayscale concentration b =R_num b , L_Gray a With R_Gray b Consistent, and |L_num a -R_num b | is less than the overlap threshold, then L_Gray a =L_num a and R_Gray b =R_num b Mark as coincident factor; otherwise, mark as non-coincident factor;

[0177] Get the number of coincidence factors in the left-phase grayscale concentration, marked as LNUM_Y; get the number of non-coincidence factors in the left-phase grayscale concentration, marked as LNUM_N;

[0178] Get the number of coincidence factors in the right phase grayscale concentration, marked as RNUM_Y; get the number of non-coincidence factors in the right phase grayscale concentration, marked as RNUM_N;

[0179] Obtain the degree of overlap between the left-phase grayscale set and the right-phase grayscale set according to the number of overlap factors and the number of non-overlap factors;

[0180]

[0181] Among them, the SMALLER(A, B) function is used to return the smaller value between A and B;

[0182] Obtaining grayscale values ​​in the left-phase grayscale set in sequence, and marking the grayscale values ​​as target grayscale values;

[0183] Set the derivative value to generate the derivative gray value according to the target gray value and the derivative value;

[0184] If the target gray value is tarGray, and the derivative value der is set, the derivative gray value derGray is generated according to the target gray value and the derivative value, and the derivative gray value derGray∈[tarGray-der, tarGray+der], if tarGray-der is less than 0, tarGray-der is set to 0, and if tarGray+der is greater than 255, tarGray+der is set to 255;

[0185] Obtaining pixel coordinates corresponding to target grayscale values ​​in the left-phase image, marking the coordinates as left-phase coordinates, and generating a left-phase coordinate set;

[0186] The left phase coordinate set = {left phase coordinate L1 , left phase coordinate L2 , ..., left phase coordinate Ln};

[0187] Obtaining pixel coordinates corresponding to the derived grayscale values ​​in the right phase image, marking the coordinates as right phase coordinates, and generating a right phase coordinate set;

[0188] The right phase coordinate set = {right phase coordinate R1 , right phase coordinate R2 , ..., right phase coordinate Rn};

[0189] Generate a number of grayscale vectors according to the left phase coordinate set and the right phase coordinate set; the grayscale vectors include grayscale vectors L1，R1 , grayscale vector L1，R2 , ..., grayscale vector L1，Rn , ..., grayscale vector Li，R1 , grayscale vector Li，R2 , ..., grayscale vector Li，Rn , ..., grayscale vector Ln，R1 , grayscale vector Ln，R2 , ..., grayscale vector Ln，Rn ;

[0190] Obtaining grayscale vectors in sequence, marking the grayscale vectors as target vectors, and marking other grayscale vectors except the target vector as non-target vectors;

[0191] Obtain a non-target vector parallel to the target vector, obtain the number of the non-target vectors, and mark the number as a parallel value para of the target vector;

[0192] Get the parallel values ​​corresponding to each grayscale vector and generate the parallel set PARA;

[0193] The parallel set PARA = {para1, para2, ..., para n};

[0194] Set the parallel error paraDev, and obtain the quantity difference interval according to the parallel error and the left phase grayscale set;

[0195]

[0196] Set weight coefficient θ1 and weight coefficient θ2, and θ1<θ2;

[0197] According to the weight coefficient θ1, the weight coefficient θ2, the coincidence degree, the parallel set and the quantity difference interval, the similarity between the left phase image and the right phase image is obtained;

[0198]

[0199] Among them, if the value in A in the EXIXST(A, B) function is consistent with that in B, it returns 1, otherwise, it returns 0;

[0200] It should be further explained that, in the specific implementation process, the process of performing expansion deduplication on the conventional video after frame extraction to generate the second-order conventional video includes:

[0201] If there is an overlapping area between monitoring area A and monitoring area B, then the regular videos corresponding to monitoring area A and monitoring area B have a regional relationship;

[0202] The expansive deduplication is to stitch together a number of regular videos that have a regional relationship within the same effective time zone;

[0203] Obtain a regular video with a consistent effective time zone, and mark the regular video as a same-domain video;

[0204] According to the overlapping areas between the monitoring areas, it is determined whether there is a regional relationship between the videos in the same domain;

[0205] If there is no regional relationship between the videos in the same domain, the monitoring area numbers corresponding to the videos in the same domain are obtained respectively, and the videos in the same domain and their corresponding monitoring area numbers are packaged and sent to the conventional unit for storage;

[0206] If there is a regional relationship between some of the same-domain videos and no regional relationship between the other part of the same-domain videos, the monitoring area numbers corresponding to the same-domain videos without regional relationship are obtained, the same-domain videos and their corresponding monitoring area numbers are packaged and sent to the regular unit for storage, and then the same-domain videos with regional relationship are obtained for expansive deduplication to generate secondary regular videos, and the secondary regular videos and their corresponding monitoring area numbers are sent to the regular unit for storage;

[0207] If all the same-domain videos have regional relationships, then the same-domain videos with regional relationships are expansively deduplicated to generate secondary regular videos, and the secondary regular videos and their corresponding monitoring area numbers are sent to the regular unit for storage;

[0208] The process of stitching together a plurality of conventional videos having regional relationships within the same effective time zone includes:

[0209] It is assumed that there is a regional relationship between the same-domain video A and the same-domain video B;

[0210] Obtain video image set A corresponding to same-domain video A, and obtain video image set B corresponding to same-domain video B;

[0211] The video image set A={P a1 , P a2 , ..., P an}, where n is a positive integer;

[0212] The video image set B={P b1 , P b2 , ..., P bn}, where n is a positive integer;

[0213] Obtaining a monitoring area corresponding to the same-domain video A, and marking the monitoring area as monitoring area A;

[0214] Obtaining a monitoring area corresponding to the same-domain video B, and marking the monitoring area as monitoring area B;

[0215] Obtaining an overlapping area between monitoring area A and monitoring area B, and marking the overlapping area as overlapping area AB;

[0216] The frame numbers corresponding to the video image set A and the video image set B are obtained respectively and compared, and the video image set with more frames is marked as the basic image set, and the video image set with less frames is marked as the interpolated frame image set; if the frame numbers are the same, one video image set is randomly marked as the basic image set, and the other video image set is marked as the interpolated frame image set;

[0217] Generate a number of second-order image pairs according to the basic image set and the complemented frame image set;

[0218] The second-order image pair = {second-order image i , the second-order image j}, where the second-order image i and the second-order image j In no particular order, and second-order images i and the second-order image j The images are opposite to each other;

[0219] Preset interpolation frame image set = {P a1 , P a2 , P a3 , P a4}, basic image set = {P b1 , P b2 , P b3 , P b4 , P b5 , P b6};

[0220] Get the time corresponding to each image in the interpolated frame image set and record it as T a1, T a2 , T a3 and T a4 , where T a1 <T a2 <T a3 <T a4 , and a1 <a2<a3<a4;

[0221] Get the time corresponding to each image in the basic image set and record it as T b1 , T b2 , T b3 , T b4 , T b5 and T b6 , T b1 <T b2 <T b3 <T b4 <T b5 <T b6 , and b1 <b2<b3<b4<b5<b6;

[0222] Preset time axis T a1 <T b1 <T b2 <T a2 =T b3 <T b4 <T b5 <T b6 <T a3 <T a4 ;

[0223] According to the time axis, each image in the basic image set and the image in the complementary frame image set closest to its own time form a second-order image pair; if the number of images in the basic image set that obtain the complementary frame image set closest to its own time is 2, then the image in the basic image set and the image in the complementary frame image set that is larger than its own time and has the smallest subscript form a second-order image pair;

[0224] If there is an image in the supplementary frame image set that does not form a second-order image pair with any image in the basic image set, then according to the principle that each image in the basic image set forms a second-order image pair with an image in the supplementary frame image set, the image in the supplementary frame image set and the image in the basic image set closest to itself form a second-order image pair;

[0225] That is, generate a second-order image pair 1{second-order image b1 , the second-order image a1}、Second-order image pair 2{Second-order image b2 , the second-order image a2}、Second-order image pair 3{Second-order image b3 , the second-order image a2}、Second-order image pair 4{Second-order imageb4 , the second-order image a2}、Second-order image pair 5{Second-order image b5 , the second-order image a3}、Second-order image pair 6{Second-order image b6 , the second-order image a3}、Second-order image pair 7{Second-order image a4 , the second-order image b6};

[0226] generating a plurality of final frame images according to a plurality of second-order image pairs;

[0227] Obtain any second-order image in each second-order image pair to generate an image set to be stitched;

[0228] Sequentially acquire images in the image set to be stitched, and mark the images as stitched images;

[0229] Obtain pixel points corresponding to the overlapped area AB in the stitched image, mark the pixel points as pixels to be covered, and mark other pixel points as stitched pixel points;

[0230] Establish a two-dimensional coordinate system A, map the stitched image to the two-dimensional coordinate system A, make each pixel point have its corresponding coordinate, and generate the coordinate relationship between the stitched pixel point and the pixel point to be covered;

[0231] Obtain an opposite side image corresponding to the stitched image, obtain pixel points corresponding to the overlapped area AB in the opposite side image, mark the pixel points as top-package pixel points, and mark other pixel points as opposite side pixel points;

[0232] Establish a two-dimensional coordinate system B, map the opposite edge image into the two-dimensional coordinate system B, so that each pixel has its corresponding coordinate, and generate the coordinate relationship between the opposite edge pixel and the top package pixel;

[0233] According to the overlapped area AB, the two-dimensional coordinate system A and the two-dimensional coordinate system B, the top-pack pixel points and the to-be-top-pack pixel points are matched, so that the top-pack pixel points and the to-be-top-pack pixel points are in a one-to-one correspondence, that is, one top-pack pixel point corresponds to and only corresponds to one to-be-top-pack pixel point;

[0234] Obtain the pixels to be covered in sequence, and mark the pixels to be covered as target pixels;

[0235] Obtain the top pixel point corresponding to the target pixel point, and mark the top pixel point as a bidding pixel point;

[0236] Get the RGB value of the target pixel and the RGB value of the bidding pixel;

[0237] Compare the RGB value of the target pixel and the RGB value of the bidding pixel, and obtain the pixel corresponding to the larger RGB value as the expected pixel; if the RGB value of the target pixel is consistent with the RGB value of the bidding pixel, randomly obtain one of the pixels as the expected pixel;

[0238] Generate a corresponding final frame image according to the coordinate relationship between the stitched pixel points and the pixel points to be covered, the coordinate relationship between the opposite edge pixel points and the pixel points to be covered, and the expected pixel points;

[0239] According to the time sequence of the basic image set, a plurality of final frame images are sorted once; according to the time sequence of the supplementary frame image set, a plurality of final frame images are sorted twice to generate a second-order conventional video;

[0240] It should be further explained that, in the specific implementation process, the process of handling early warning information includes:

[0241] The monitoring area number in the warning information is sent to the user, informing the user that an abnormality exists in the monitoring area corresponding to the monitoring area number, and the user takes corresponding measures;

[0242] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A video monitoring system for a wind farm production site, comprising a control center, characterized in that: The control center is communicatively connected to a database, a video data acquisition module, a video data analysis module, a video data transmission module, and an early warning module; The database is used to store abnormal videos, regular videos, and an abnormal judgment model; The video data acquisition module is used to obtain the monitoring video data of the production site of the wind farm; The video data analysis module is used to judge whether there is an abnormality in the monitoring video data through the abnormal judgment model. If there is an abnormality, the monitoring video data is an abnormal video, and corresponding early warning information is generated. If there is no abnormality, the monitoring video data is a regular video; The video data transmission module is used to submit the abnormal videos and regular videos to the database for storage, and during the process of submitting the regular videos, perform frame extraction on the regular videos, and perform expansion deduplication on the frame-extracted regular videos to generate second-order regular videos; The early warning module is used to process the early warning information.

2. A video monitoring system for a wind farm production site according to claim 1, characterized in that: The production site of the wind farm is divided into several monitoring areas, and the several monitoring areas are numbered; there are overlapping areas between the monitoring areas, and the numbers are unique; the monitoring areas are monitored by monitoring devices.

3. A video monitoring system for a wind farm production site according to claim 2, characterized in that: The database includes an abnormal unit, a regular unit, and a model unit; The abnormal unit includes abnormal video data and an abnormal number, and is used to store abnormal videos; the abnormal number is the monitoring area number corresponding to the abnormal video data; The regular unit includes regular video data and a regular number, and is used to store regular videos; the regular number is the monitoring area number corresponding to the regular video data; The model unit includes a model unit and a self-learning unit; the model unit is used to store the abnormal judgment model; the self-learning unit generates a new abnormal judgment model according to the abnormal video data through a self-supervised learning method, and submits the new abnormal judgment model to the model unit for storage.

4. A video monitoring system for a wind farm production site according to claim 3, characterized in that: The process of obtaining the monitoring video data of the production site of the wind farm includes: The video data acquisition module is composed of several monitoring devices; The monitoring device includes a camera, a wind speed sensor, and a microprocessor; Set the acquisition period T1, the acquisition period T2, and the acquisition period T3; Set the wind speed value w1 and the wind speed value w2, and w1 < w2; Obtain the current wind speed through the wind speed sensor in the monitoring device, and mark the wind speed as w; When w < w1, the monitoring device generates the monitoring video data of the production site of the wind farm according to the acquisition period T1; when w1 ≤ w < w2, the monitoring device generates the monitoring video data of the production site of the wind farm according to the acquisition period T2; when w ≥ w2, the monitoring device generates the monitoring video data of the production site of the wind farm according to the acquisition period T3.

5. A video monitoring system for a wind farm production site according to claim 4, characterized in that: The process of judging whether there is an abnormality in the monitoring video data according to the abnormal judgment model includes: The abnormal judgment model generates result information according to the input data; the input data is video data, and the result information includes the existence of an abnormality and the non-existence of an abnormality; Input the monitoring video data into the abnormal judgment model, and the abnormal judgment model generates corresponding result information according to the input monitoring video data; If the result information indicates that an abnormality exists, the monitoring video data is an abnormal video, and the monitoring area number corresponding to the abnormal video is obtained to generate an early warning message, and then the early warning message is submitted to the early warning module, and the early warning module makes corresponding processing; If the result information shows that there is no abnormality, the surveillance video data is a regular video.

6. A video monitoring system for a wind farm production site according to claim 5, characterized in that: The process of submitting abnormal videos and regular videos to the database for storage includes: The monitoring area number corresponding to the abnormal video is obtained, and the abnormal video and the monitoring area number are submitted to the database for storage; the monitoring area number corresponding to the normal video is obtained, and the normal video and the monitoring area number are submitted to the database for storage.

7. A video monitoring system for a wind farm production site according to claim 6, characterized in that: The process of frame extraction for regular video includes: A conventional video is split into a number of images in units of frames, and the images are sorted according to time to generate a video image set; the video image set = {P1, P2, ..., P n }, where n is a positive integer; Obtaining the time corresponding to the first frame image and the last frame image of the regular video, and generating a valid time zone of the regular video according to the time; The frame extraction is to delete a certain frame of image; Step a1: obtaining two images with adjacent subscripts, comparing the subscripts, marking the image corresponding to the smaller subscript as the left-phase image, and marking the image corresponding to the larger subscript as the right-phase image; Step a2: Obtain the similarity between the left-phase image and the right-phase image; Step a3: setting a similarity threshold. If the similarity is greater than or equal to the similarity threshold, a similar image pair is generated based on the left-phase image and the right-phase image. Otherwise, no operation is required. Step a4: Repeat steps a1 to a3 until no similar image pairs can be generated; Step a5: obtaining similar image pairs in the video image set, and randomly deleting one image in the similar image pair to complete the frame extraction operation; if there are similar image pairs A and similar image pairs B, and the same image exists in the similar image pairs A and B, then the different images are deleted from the similar image pairs A and B when performing the frame extraction operation; Step a5: sort the video image set after the frame extraction operation according to time, generate a new video image set, and overwrite the original video image set with the new video image set.

8. A video monitoring system for a wind farm production site according to claim 7, characterized in that: The process of obtaining the similarity between the left-phase image and the right-phase image includes: Establishing a three-dimensional rectangular coordinate system, wherein the three-dimensional rectangular coordinate system includes an xy axis two-dimensional coordinate system, an xz axis two-dimensional coordinate system, and a yz axis two-dimensional coordinate system; Map the left phase image to the yz axis two-dimensional coordinate system, and map the right phase image to the xz axis two-dimensional coordinate system, so that the pixels in the left phase image and the right phase image have their corresponding coordinates; Grayscale processing is performed on the left-phase image and the right-phase image; Obtain the grayscale values ​​and the corresponding pixel numbers in the left-phase image to generate a left-phase grayscale set; the left-phase grayscale set = {L_Gray1 = L_num1, L_Gray2 = L_num2, ..., L_Gray n =L_num n }, where n is a positive integer, L_Gray i Refers to the grayscale value in the left phase image, L_num i Refers to L_Gray i The corresponding number of pixels; Get the grayscale values ​​and the corresponding pixel numbers in the right phase image to generate a right phase grayscale set; the right phase grayscale set = {R_Gray1 = R_num1, R_Gray2 = R_num2, ..., R_Gray n =R_num n }, where n is a positive integer, R_Gray i Refers to the grayscale value in the right phase image, R_num i Refers to R_Gray i The corresponding number of pixels; Set the coincidence threshold; If there is L_Gray in the left phase grayscale set a =L_num a , R_Gray exists in the right phase grayscale concentration b =R_num b , L_Gray a With R_Gray b Consistent, and |L_num a -R_num b | is less than the overlap threshold, then L_Gray a =L_num a and R_Gray b =R_num b Mark as coincident factor; otherwise, mark as non-coincident factor; Get the number of coincidence factors in the left-phase grayscale concentration; get the number of non-coincidence factors in the left-phase grayscale concentration; Get the number of coincidence factors in the right phase grayscale concentration; Get the number of non-coincidence factors in the right phase grayscale concentration; Obtain the degree of overlap between the left-phase grayscale set and the right-phase grayscale set according to the number of overlap factors and the number of non-overlap factors; Obtaining grayscale values ​​in the left-phase grayscale set in sequence, and marking the grayscale values ​​as target grayscale values; Set the derivative value to generate the derivative gray value according to the target gray value and the derivative value; Obtaining pixel coordinates corresponding to target grayscale values ​​in the left-phase image, marking the coordinates as left-phase coordinates, and generating a left-phase coordinate set; Obtaining pixel coordinates corresponding to the derived grayscale values ​​in the right phase image, marking the coordinates as right phase coordinates, and generating a right phase coordinate set; Generate a number of grayscale vectors according to the left phase coordinate set and the right phase coordinate set; Obtaining grayscale vectors in sequence, marking the grayscale vectors as target vectors, and marking other grayscale vectors except the target vector as non-target vectors; Acquire a non-target vector parallel to the target vector, acquire the number of the non-target vectors, and mark the number as a parallel value of the target vector; Get the parallel values ​​corresponding to each grayscale vector and generate a parallel set; Set the parallel error, and obtain the quantity difference interval according to the parallel error and the left phase grayscale set; Set weight coefficient θ1 and weight coefficient θ2, and θ1<θ2; The similarity between the left-phase image and the right-phase image is obtained according to the weight coefficient θ1, the weight coefficient θ2, the coincidence degree, the parallel set and the quantity difference interval.

9. A video monitoring system for a wind farm production site according to claim 8, characterized in that: The process of performing expansion deduplication on the regular video after frame extraction to generate the second-order regular video includes: If there is an overlapping area between monitoring area A and monitoring area B, then the regular videos corresponding to monitoring area A and monitoring area B have a regional relationship; The expansive deduplication is to stitch together a number of regular videos that have a regional relationship within the same effective time zone; Obtain a regular video with a consistent effective time zone, and mark the regular video as a same-domain video; According to the overlapping areas between the various monitoring areas, determine whether there is a regional relationship between the videos in the same domain; if there is no regional relationship between the videos in the same domain, respectively obtain the monitoring area numbers corresponding to the videos in the same domain, and package the videos in the same domain and their corresponding monitoring area numbers and send them to the conventional unit for storage; if there is a regional relationship between some videos in the same domain and there is no regional relationship between the other videos in the same domain, obtain the monitoring area numbers corresponding to the videos in the same domain without the regional relationship, package the videos in the same domain and their corresponding monitoring area numbers and send them to the conventional unit for storage, then obtain the videos in the same domain with the regional relationship for expansive deduplication, generate secondary conventional videos, and send the secondary conventional videos and their corresponding monitoring area numbers to the conventional unit for storage; if there is a regional relationship between the videos in the same domain, perform expansive deduplication on the videos in the same domain with the regional relationship, generate secondary conventional videos, and send the secondary conventional videos and their corresponding monitoring area numbers to the conventional unit for storage; The process of stitching together a plurality of conventional videos having regional relationships within the same effective time zone includes: It is assumed that there is a regional relationship between the same-domain video A and the same-domain video B; Obtain a video image set A corresponding to a same-domain video A, and obtain a video image set B corresponding to a same-domain video B; obtain a monitoring area corresponding to the same-domain video A, and mark the monitoring area as monitoring area A; obtain a monitoring area corresponding to the same-domain video B, and mark the monitoring area as monitoring area B; Obtaining an overlapping area between monitoring area A and monitoring area B, and marking the overlapping area as overlapping area AB; The frame numbers corresponding to the video image set A and the video image set B are obtained respectively and compared, and the video image set with more frames is marked as the basic image set, and the video image set with less frames is marked as the interpolated frame image set; if the frame numbers are the same, one video image set is randomly marked as the basic image set, and the other video image set is marked as the interpolated frame image set; Generate a number of second-order image pairs according to the basic image set and the complemented frame image set; generating a plurality of final frame images according to a plurality of second-order image pairs; The second-order image pair = {second-order image i , the second-order image j }, where the second-order image i and the second-order image j In no particular order, and second-order images i and the second-order image j The images are opposite to each other; Obtain any second-order image in each second-order image pair to generate an image set to be stitched; Sequentially acquire images in the image set to be stitched, and mark the images as stitched images; Obtain pixel points corresponding to the overlapped area AB in the stitched image, mark the pixel points as pixels to be covered, and mark other pixel points as stitched pixel points; Establish a two-dimensional coordinate system A, map the stitched image to the two-dimensional coordinate system A, make each pixel point have its corresponding coordinate, and generate the coordinate relationship between the stitched pixel point and the pixel point to be covered; Obtain an opposite side image corresponding to the stitched image, obtain pixel points corresponding to the overlapped area AB in the opposite side image, mark the pixel points as top-package pixel points, and mark other pixel points as opposite side pixel points; Establish a two-dimensional coordinate system B, map the opposite edge image into the two-dimensional coordinate system B, so that each pixel has its corresponding coordinate, and generate the coordinate relationship between the opposite edge pixel and the top package pixel; According to the overlapped area AB, the two-dimensional coordinate system A and the two-dimensional coordinate system B, the top-pack pixel points and the to-be-top-pack pixel points are matched, so that the top-pack pixel points and the to-be-top-pack pixel points are in a one-to-one correspondence, that is, one top-pack pixel point corresponds to and only corresponds to one to-be-top-pack pixel point; Obtain the pixels to be covered in sequence, and mark the pixels to be covered as target pixels; Obtain the top pixel point corresponding to the target pixel point, and mark the top pixel point as a bidding pixel point; Get the RGB value of the target pixel and the RGB value of the bidding pixel; Compare the RGB value of the target pixel and the RGB value of the bidding pixel, and obtain the pixel corresponding to the larger RGB value as the expected pixel; if the RGB value of the target pixel is consistent with the RGB value of the bidding pixel, randomly obtain one of the pixels as the expected pixel; Generate a corresponding final frame image according to the coordinate relationship between the stitched pixel points and the pixel points to be covered, the coordinate relationship between the opposite edge pixel points and the pixel points to be covered, and the expected pixel points; Generate a second-order regular video based on several final frame images.

10. A video monitoring system for a wind farm production site according to claim 9, characterized in that: The process of handling warning information includes: The monitoring area number in the warning information is sent to the user, informing the user that an abnormality exists in the monitoring area corresponding to the monitoring area number, and the user takes corresponding measures.

Images