CN117409369A - Construction safety management method, system and storage medium based on big data - Google Patents

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a construction safety management method, a construction safety management system and a storage medium based on big data, wherein the method comprises the steps of installing a plurality of shooting devices in different construction areas of a construction site, and periodically collecting video data shot by the plurality of shooting devices; sequentially carrying out portrait detection on a plurality of video data, detecting whether non-constructors exist in the plurality of video data, sending out early warning when the non-constructors exist, informing the non-constructors to leave a construction site, and recording information data of the non-constructors; and when detecting that a plurality of non-constructors exist in the video data, judging whether the plurality of non-constructors are the same person, and if so, improving the early warning level. The invention can help identify non-constructors at the construction site, and avoid the same non-constructors from staying at the construction site for a long time, thereby avoiding safety accidents at the construction site.

Description

Construction safety management method, system and storage medium based on big data

Technical Field

The invention belongs to the technical field of data processing, and particularly relates to a construction safety management method, a construction safety management system and a storage medium based on big data.

Background

The method is characterized in that a large data technology is utilized to realize image processing, the technical bottleneck of traditional image processing can be broken through, large-scale image data processing can be rapidly realized, a construction site is provided with a plurality of dangerous sources, a non-constructor can not know which places to avoid are scattered everywhere and can generate unnecessary accidents, in order to avoid the problems, a plurality of shooting devices are arranged in different areas of the construction site, whether external personnel enter the construction site by mistake is monitored in real time, video data are collected through the shooting devices arranged in the construction site, the video data are converted into image data, the images are processed based on the large data technology to obtain portrait characteristics, the portrait characteristics are used for carrying out portrait detection on the image data, non-constructors are identified from the portrait characteristics according to marks such as safety belts or safety caps, and management personnel inform the non-constructors of leaving the construction site to avoid the occurrence of safety accidents, so that the construction safety management method based on the large data is provided, and the non-constructor does not enter by mistake to cause unnecessary safety accidents are monitored in real time.

Similar prior art publication No. CN113434902B, the invention provides a blockchain-based construction safety monitoring management system and method, comprising: the data acquisition module is used for acquiring target monitoring data and information of the construction site; the data processing module processes the data and the information through a deep learning algorithm, acquires and broadcasts event key information, and then performs consistency verification and sequencing on the event key information; the supervision early warning module receives node information broadcast on the block chain, deploys intelligent contracts according to the accident tree model and the event key information, and performs early warning and feedback; and the remote server module records the original information of each device and the operators, and acquires the uploaded and stored node information in real time for inquiring and accident responsibility tracing. However, the invention does not consider the situation that the external non-constructors mistakenly enter the construction site to cause safety accidents.

The similar prior art also discloses a Chinese invention patent with publication number of CN109947806B, and discloses a case-based reasoning ultra-high-rise construction safety accident emergency auxiliary decision-making method, which comprises the following steps: step 1, inputting accident case information to form a sample case library; step 2, accident case retrieval and reasoning analysis are carried out, and a case index data set is established; step 3: similarity calculation and correction; step 4, generating and evaluating an emergency auxiliary decision scheme; according to the case-based reasoning ultra-high-rise building construction safety accident emergency auxiliary decision-making method disclosed by the invention, the problem of weak theory and strong experience in ultra-high-rise construction safety accident emergency rescue can be effectively solved, the defect that an emergency decision-making scheme lacks pertinence due to incomplete construction safety accident information can be effectively avoided, and a novel scientific thinking and supporting method can be provided for ultra-high-rise construction safety accident emergency decision-making. But the invention also does not consider the situation that the external non-constructors mistakenly enter the construction site to cause safety accidents. Thus, the invention provides a construction safety management method, a construction safety management system and a storage medium based on big data.

Disclosure of Invention

According to the invention, a plurality of shooting devices are installed in different construction areas of a construction site, and video data shot by the plurality of shooting devices are periodically collected; sequentially carrying out portrait detection on a plurality of video data, detecting whether non-constructors exist in the plurality of video data, sending out early warning when the non-constructors exist, informing the non-constructors to leave a construction site, and recording information data of the non-constructors; and when detecting that a plurality of non-constructors exist in the video data, judging whether the plurality of non-constructors are the same person, and if so, improving the early warning level. The invention can help identify non-constructors at the construction site, and avoid the same non-constructors from staying at the construction site for a long time, thereby avoiding safety accidents at the construction site.

In order to achieve the above object, the present invention provides a construction safety management method based on big data as described below, mainly comprising the steps of:

s1, installing a plurality of shooting devices in different construction areas of a construction site, periodically collecting first video data shot by the first shooting device, collecting second video data shot by the second shooting device, and collecting N-th video data shot by the N-th shooting device, wherein N is a positive integer larger than 2;

S2, judging whether non-constructors exist in the first video data when the first video data are detected, sending out first-level early warning when the non-constructors exist, recording information data of the first non-constructors, detecting the second video data, recording information data of a second non-constructor when the non-constructors exist from the second video data, judging whether two non-constructors are the same person according to the recorded information data of the first non-constructors and the second non-constructors, and sending out second-level early warning when the two non-constructors are the same person;

and S3, continuously detecting third to N video data, recording information data corresponding to non-constructors every time one non-constructor is found, comparing the information data with the information data of the existing non-constructors to judge whether the information data is the same person, and sending out three-level early warning under the condition that the same person exists in the video data with the number larger than the preset number.

As a preferred technical solution of the present invention, the process of recording the information data of the first non-constructor includes the following steps:

s21, converting the first video data to generate a time sequence image, carrying out portrait detection on a first image of the time sequence image, detecting whether a non-constructor exists in the first image, continuously detecting the next image of the time sequence image under the condition that the non-constructor does not exist, until all images of the time sequence image are detected, and taking the non-constructor as a target object under the condition that the non-constructor exists;

S22, tracking the target object, extracting a portrait area of the target object, and performing enlargement or reduction processing on the portrait area according to a preset standard;

s23, analyzing the moving direction and the face orientation of a target object, and acquiring portrait areas of the target object in different directions according to the moving direction and the face orientation, wherein during the process of tracking the target object, portrait areas of the target object in the same direction at different moments are acquired;

s24, dividing the portrait area into a plurality of areas, extracting color characteristic values in each area, adding the color characteristic values into a color record table of the target object, and recording the color characteristic values in the same direction into the same color record table, wherein the color record table comprises portrait direction, time information, portrait areas and corresponding color characteristic values;

s25, after all images of the time sequence images are detected, calculating a weight coefficient of each region according to the color record table;

s26, after the weight coefficient is calculated, storing the information data of the target object, judging whether the information data exists in an information record table before the information data of the target object is stored, if the information data does not exist, distributing an object ID and an object tag to the target object, storing table names of all color record tables corresponding to the object ID, the object tag and the target object in the information record table as one information data, if the information data exists, indicating that a non-constructor is detected when the video data before detection exists, taking out the first information data in the information record table, comparing the first information data with the information data of the target object, judging whether the non-constructor corresponding to the first information data and the target object are the same person, if the non-constructor corresponding to the first information data and the target object are the same person, distributing the object tag same as the first information data to the target object, and storing the information data of the target object in the information record table, if the information data of the target object is not the same person, taking out the second information data in the information record table and comparing the information data of the target object until the information of the target object is compared.

As a preferred technical solution of the present invention, extracting a color feature value in each of the regions includes the steps of:

s241, acquiring HSL color values of all pixel points of the area in sequence, adding tone values of the HSL color values of all pixel points, and dividing the tone values by the number of all pixel points of the area to obtain a tone average value of the area;

s242, calculating the difference value between the hue value of each pixel point and the hue average value, and marking the pixel point as a salient pixel under the condition that the difference value is larger than a preset threshold value;

s243, obtaining RGB color values of all the salient pixels, calculating RGB average values of the RGB color values of all the salient pixels, and taking the RGB average values as color characteristic values of the area.

As a preferred technical solution of the present invention, calculating a weight coefficient of each area according to the color record table includes the following steps:

s251, obtaining color characteristic values of the same region at different moments, respectively calculating color change rates of two adjacent moments, adding the color change rates to obtain the sum of the color change rates, obtaining the number of the color change rates, dividing the sum of the color change rates by the number of the color change rates, and calculating to obtain the average color change rate of the current region;

S252, dividing one by the average color change rate, obtaining the reciprocal of the average color change rate of all areas, and taking the reciprocal as a weight coefficient of the corresponding area.

As a preferable technical scheme of the invention, the process for judging whether the non-constructor corresponding to the first piece of information data and the target object are the same person comprises the following steps:

s261, acquiring a first color record table of a first direction of the first information data, wherein each region in the color record table corresponds to color characteristic values of a plurality of moments, calculating the average color characteristic value of each region according to the color characteristic values of different moments, acquiring a second color record table of the first direction of the information data of the target object, calculating the average color characteristic value of each region in the second color record table, calculating the color similarity of each region according to the average color characteristic value of each same region, calculating the average weights of two weight coefficients corresponding to the same region in the first direction after calculating the color similarity of all regions, multiplying the color similarity of each region by the average weight of each region to obtain the region similarity of each region, and adding the region similarity of all regions to obtain the similarity of the first direction;

S262, calculating the similarity in other directions by using the method of S261, and calculating the average value of the similarity in all directions, wherein the average value is the overall similarity between the target object and the non-constructor;

s263, comparing the overall similarity with a preset threshold, wherein the non-constructor and the target object are the same person under the condition that the overall similarity is larger than the preset threshold, and the non-constructor and the target object are not the same person under the condition that the overall similarity is smaller than or equal to the preset threshold.

The invention also provides a construction safety management system based on big data, which comprises the following modules:

the data collection unit is used for periodically collecting first video data shot by the first shooting equipment, collecting second video data shot by the second shooting equipment, collecting N-th video data shot by N-th shooting equipment, wherein N is a positive integer larger than 2, and storing the plurality of video data in the first memory;

the image detection unit is used for sequentially carrying out image detection on a plurality of video data, judging whether non-constructors exist in the first video data when the first video data are detected, continuously detecting second video data under the condition that the non-constructors do not exist, recording information data of the first non-constructors, continuously detecting the second video data, and recording information data of the second non-constructors under the condition that the non-constructors exist at the detection position of the second video data;

And the identification unit is used for judging whether the two non-constructors are the same person according to the recorded information data of the first non-constructor and the second non-constructor.

The present invention also provides a storage medium storing program instructions, wherein the program instructions, when executed, control a device in which the storage medium is located to perform any one of the methods described above.

Compared with the prior art, the invention has the following beneficial effects:

in the invention, a plurality of shooting devices are installed in different construction areas of a construction site, and video data shot by the plurality of shooting devices are periodically collected; sequentially carrying out portrait detection on a plurality of video data, detecting whether non-constructors exist in the plurality of video data, sending out early warning when the non-constructors exist, informing the non-constructors to leave a construction site, and recording information data of the non-constructors; and when detecting that a plurality of non-constructors exist in the video data, judging whether the plurality of non-constructors are the same person, and if so, improving the early warning level. The invention can help identify non-constructors at the construction site, and avoid the same non-constructors from staying at the construction site for a long time, thereby avoiding safety accidents at the construction site.

Drawings

FIG. 1 is a flow chart of steps of a construction safety management method based on big data of the present invention;

FIG. 2 is a block diagram of a construction safety management system based on big data according to the present invention;

FIG. 3 is a schematic illustration of a portrait area according to the present invention;

FIG. 4 is a front side color chart of the present invention;

shown in fig. 3: 100. a portrait area.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

It will be understood that the terms "first," "second," and the like, as used herein, may be used to describe various elements, but these elements are not limited by these terms unless otherwise specified. These terms are only used to distinguish one element from another element. For example, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script, without departing from the scope of the present application.

The invention provides a construction safety management method based on big data as shown in figure 1, which is realized mainly by executing the following steps:

S1, installing a plurality of shooting devices in different construction areas of a construction site, periodically collecting first video data shot by the first shooting device, collecting second video data shot by the second shooting device, and collecting N-th video data shot by the N-th shooting device, wherein N is a positive integer larger than 2;

s2, judging whether non-constructors exist in the first video data when the first video data are detected, sending out first-level early warning when the non-constructors exist, recording information data of the first non-constructors, detecting second video data, recording information data of the second non-constructors when the non-constructors exist from the second video data, judging whether the two non-constructors are the same person according to the recorded information data of the first non-constructors and the information data of the second non-constructors, and sending out second-level early warning when the two non-constructors are the same person;

and S3, continuously detecting third to N video data, recording information data corresponding to non-constructors every time one non-constructor is found, comparing the information data with the information data of the existing non-constructors to judge whether the information data is the same person, and sending out three-level early warning under the condition that the same person exists in the video data with the number larger than the preset number.

Specifically, construction personnel on a construction site generally wear safety measures such as a safety helmet, in order to avoid that external personnel enter the construction site by mistake or other reasons, but non-construction personnel generally do not wear safety measures such as the safety helmet and do not know which places on the construction site are dangerous and cannot approach, if the non-construction personnel walk on the construction site for a long time and in a large area, the possibility of safety accidents is improved, in order to avoid unnecessary safety accidents, a plurality of shooting devices are installed in different areas of the construction site, video data shot by the plurality of shooting devices are periodically collected, the video data are converted into image data, image detection is carried out based on the large data technology, the image processing is carried out by the large data technology, the information contained in the image can be fully utilized, the accuracy and the reliability of an image analysis result are greatly improved, the personnel on the construction site are identified, the personnel who do not wear the working clothes and do not wear safety marks are regarded as non-construction personnel, the possibility of the safety helmet can be improved in sequence, the machine learning model is used for identifying the personnel who do not wear the working and the safety helmet, the personnel are identified as the non-necessary safety helmet, the personnel can be prevented from leaving the construction site, the situation of the non-construction personnel can be easily read by the personnel on the construction site, the basis of the large-data, the non-requirement personnel can be easily read by the non-stop people on the construction site, and the condition of the non-stop people can be easily, and the danger personnel can be easily read by the non-stop on the construction site, and the condition of the non-stop people on the construction site, and the construction site through the condition can be well through the condition that is not and the detection of the voice personnel is not continuously can be well by the people on the construction site has the construction site, when judging that the same non-constructor exists in two video data, indicating that the non-constructor does not leave and at least walks to two areas of a construction site after hearing early warning information for the first time, sending out a secondary early warning to secondarily remind the non-constructor and improve reminding frequency and intensity, reminding the non-constructor to leave the construction site, and if judging that the same non-constructor exists in preset numbers such as three video data, indicating that the non-constructor stays in the construction site for at least three areas and for a long time, the possibility of occurrence of safety accidents is relatively increased, and at the moment, the non-constructor needs to be reminded of leaving the construction site by sending out a tertiary early warning, and informing an operator on duty to take the non-constructor away from the construction site so as to avoid the safety accidents.

Further, the process of recording the information data of the first non-constructor includes the following steps:

s21, converting the first video data to generate a time sequence image, carrying out portrait detection on the first image of the time sequence image, detecting whether a non-constructor exists in the first image, and continuously detecting the next image of the time sequence image under the condition that the non-constructor does not exist until all images of the time sequence image are detected, and taking the non-constructor as a target object under the condition that the non-constructor exists;

specifically, in order to detect whether a non-constructor exists in a construction site, video data shot by shooting equipment are respectively converted into time sequence images, a constructor can wear work clothes, wear safety helmets and other marked clothes, the constructor is regarded as the non-constructor for a person who does not wear the work clothes and wears the safety helmets, a machine learning model is used for carrying out image detection on a first image of the time sequence images, whether the non-constructor exists in the first image, the non-constructor does not exist in the first image, the non-constructor does not indicate that the non-constructor exists after all the time sequence images are detected, the non-constructor does not exist in a current area, the same detection is carried out on the next video data, if the non-constructor is detected from one image of the time sequence images, the non-constructor intrudes into an area covered by the current shooting equipment, and the detected non-constructor serves as a target object.

S22, tracking the target object, extracting a portrait area of the target object, and performing enlargement or reduction treatment on the portrait area according to a preset standard;

specifically, after a non-constructor is detected from a certain image of the time series image, the non-constructor is taken as a target object, when a subsequent image of the time series image is detected, the target object is tracked by a template matching method, the template matching method is not explained in the prior art, the human image areas of the non-constructor on different images in the time series image are obtained, the human image areas on different images correspond to the human image areas at different moments, as shown in fig. 3, the human image areas refer to the areas containing the whole human body in the image, as the human image areas shot by the shooting equipment are larger or smaller along with the movement of the non-constructor, for example, when the human image areas are relatively close to the shooting equipment, the proportion of the human image areas is larger, and when the human image areas are relatively far away from the shooting equipment, the proportion of the human image areas is smaller, the human image areas are amplified or reduced according to preset standards, and errors are reduced for subsequent judgment and calculation.

S23, analyzing the moving direction and the face orientation of the target object, and acquiring the portrait areas of the target object in different directions according to the moving direction and the face orientation, wherein the portrait areas of the target object in the same direction are acquired in the process of tracking the target object;

Specifically, in the process of tracking the target object, the moving direction and the face orientation of the target object are analyzed, when the target object is closer to the shooting device and the face is opposite to the shooting device, the situation that the person is opposite to the shooting device, that is, the front face image area can be acquired at the moment, when the target object is farther from the shooting device and the face is opposite to the shooting device, the situation that the person is opposite to the shooting device, that is, the back face image area is acquired at the moment, along with the movement of the target object, a plurality of image areas in the same direction but at different moments can be acquired, and as the target object moves, the color characteristics of the image areas in the same direction and at different moments can be different, and therefore the image areas in a plurality of moments can be acquired.

S24, dividing the portrait area into a plurality of areas, extracting color characteristic values in each area, adding the color characteristic values into a color record table of a target object, and recording the color characteristic values in the same direction into the same color record table, wherein the color record table comprises portrait direction, time information, portrait areas and corresponding color characteristic values;

specifically, after the image areas in different directions and at different moments are obtained, the image areas are equally divided into a plurality of areas from top to bottom, and the color characteristic values in each area are respectively extracted by taking five areas as an example, and the corresponding color characteristic values are added into a color record table of a corresponding target object, so that the colors in the same direction are recorded in the same color record table, as shown in fig. 4, the color record table in the front direction is the color record table in which all the areas in different moments in the same direction correspond to the color characteristic values.

S25, after all images of the time sequence images are detected, calculating a weight coefficient of each region according to a color record table;

specifically, after all images of the time series images are detected, color record tables of the front side and the back side are obtained by taking two directions as examples, according to the change condition of color characteristic values of each region in each color record table at different moments, the weight coefficient corresponding to each region is calculated, because the color characteristic values of the same region can be different at different moments due to the fact that certain parts of a body are shielded by articles held by a person along with the movement in the moving process, the reason leads to regions with large color change of the same region, if the influence degree of all regions on the subsequent calculation of the color similarity is observed equally, the deviation of a judging result is larger, the weight coefficient of the corresponding region is calculated through the change degree of the same region at different moments, the larger the change of the same region at different moments is, the weight coefficient of the region is smaller, the weight coefficient of the same region is larger at different moments is, the algorithm of the specific weight coefficient is interpreted later, and the weight coefficient corresponding to each region is added into the color record table after the weight coefficient is calculated.

S26, after the weight coefficient is calculated, storing information data of the target object, judging whether the information data exists in an information record table or not before the information data of the target object is stored, if the information data does not exist, distributing an object ID and an object label to the target object, storing table names of all color record tables corresponding to the object ID, the object label and the target object in the information record table as one piece of information data, if the information data exists, indicating that a non-constructor is detected when the video data before detection exists, taking out the first piece of information data in the information record table, comparing the first piece of information data with the information data of the target object, judging whether the non-constructor corresponding to the first piece of information data and the target object are the same person, if the non-constructor corresponding to the first piece of information data is the same person, distributing the target object label identical to the first piece of information data to the target object, storing the information data of the target object in the information record table, and if the information data of the target object is not the same person, and comparing the second piece of information data in the information record table with the information data of the target object until all the information data in the information record table are compared.

Specifically, after the weight coefficient is calculated, acquiring color record tables of two directions of a target object, acquiring table names of the two color record tables of the target object, judging whether information records exist in the information record table, if no information records exist, indicating that no non-constructors are detected in the previous video data detection, distributing an object ID and an object tag to the target object, storing the table names of the color record tables corresponding to the object ID, the object tag and the target object as one piece of information data in the information record table, and waiting until a new non-constructor is detected later, and taking the information data in the information record table as a comparison object; under the condition that information data exists, it is indicated that a non-constructor is found when video data before detection exists, for example, an information data exists, the non-constructor corresponding to the information data is regarded as a comparison object, firstly, the table name of the color record table of the comparison object is obtained, two color record tables are obtained through the table name, whether the comparison object and the target object are the same person or not is judged through the color record table of the comparison object and the color record table of the target object, under the condition that the comparison object and the target object are the same person, it is indicated that the target object does not leave a construction site by itself after receiving an early warning notification, the person is detected in the current video data, at the moment, the target object, namely, the non-constructor, sends a second early warning, meanwhile, the target object is assigned with the same object label as the comparison object, information of the target object is stored in the information record table, under the condition that the color record table of the target object is not the same person, and information data of the target object is stored in the information record table, and it is indicated that the non-constructor corresponding to the comparison object possibly leaves the construction site after receiving the first early warning, or is not in the area covered by the current video data, and if the information table exists in the current video data, and the method is continued, and the information is recorded after the comparison data is recorded, if the information is stored in the other pieces.

Further, extracting the color feature value in each region includes the following steps:

s241, sequentially obtaining HSL color values of all pixel points of the area, adding tone values of the HSL color values of all pixel points, and dividing the tone values by the number of all pixel points of the area to obtain a tone average value of the area;

s242, calculating the difference value between the hue value and the hue average value of each pixel point, and marking the pixel point as a salient pixel under the condition that the difference value is larger than a preset threshold value;

s243, obtaining RGB color values of all the salient pixels, calculating RGB average values of the RGB color values of all the salient pixels, and taking the RGB average values as color characteristic values of the region.

Specifically, in order to extract the color characteristic value in each area as whether the subsequent judgment is to make data preparation for the same person, firstly, the HSL color value of each pixel point in the area is obtained, the hue values of the HSL color values of all the pixel points are added, the number of the added hue values in all the pixel points is used for obtaining the average value of the color drop of the current area, then the difference value between the hue value and the average value of the hue of each pixel point is calculated sequentially, the pixel points with the difference value larger than a preset threshold value are marked as salient pixels, the RGB color values of all the salient pixels are obtained, the RGB average values of the RGB color values of all the salient pixels are calculated, the average value of each of the three color values of the RGB color values is calculated respectively, the combination of the average values of the three color values is the color characteristic value of the current area, and the color characteristic values of other all the areas are calculated by the same method.

Further, the weight coefficient of each region is calculated according to the color record table, and the method comprises the following steps:

s251, obtaining color characteristic values of the same region at different moments, respectively calculating color change rates of two adjacent moments, adding the color change rates to obtain the sum of the color change rates, obtaining the number of the color change rates, dividing the sum of the color change rates by the number of the color change rates, and calculating to obtain the average color change rate of the current region;

s252, dividing the average color change rate by one to obtain the reciprocal of the average color change rate, obtaining the reciprocal of the average color change rate of all the areas, and taking the reciprocal as the weight coefficient of the corresponding area.

Specifically, in order to calculate the weight coefficient of each region, obtain the color characteristic value of the same region at different moments, for example, the color characteristic value of the region at the moment one is RGB (109,221,34), the color characteristic value of the region at the moment two is RGB (109,221,34), the color characteristic value of the region at the moment three is RGB (109,221,34), the color characteristic value of the region at the moment four is RGB (100,211,42), then the first color change rate of the region at the moment two and the moment one is obtained through calculation, the second color change rate of the region at the moment three and the second color change rate of the region at the moment two, the third color change rate of the region at the moment four and the moment three is obtained, the color change rate of the whole region is the first color change rate, the second color change rate and the third color change rate are added and then divided by three, wherein the color change rate calculating method can utilize the existing color difference formula, the color difference formula is that in the prior art has various calculating formulas, for example, CIE Lab, CIEDE2000, CIUV and the like, and the color change rate of the region needs to be small to illustrate the movement of the time or the person, so the corresponding weight coefficient should be bigger, and the weight coefficient should be obtained.

Further, the process of judging whether the comparison object and the target object are the same person according to the color record table of the comparison object and the target object comprises the following steps:

s261, acquiring a first color record table of a first direction of first information data, wherein each region in the color record table corresponds to color characteristic values of a plurality of moments, calculating the average color characteristic value of each region according to the color characteristic values of different moments, acquiring a second color record table of the first direction of the information data of a target object, calculating the average color characteristic value of each region in the second color record table, calculating the color similarity of each region according to the average color characteristic value of each same region, calculating the average weight of two weight coefficients corresponding to the same region in the first direction after calculating the color similarity of all regions, multiplying the color similarity of each region by the average weight of each region to obtain the region similarity of each region, and adding the region similarity of all regions to obtain the similarity of the first direction;

s262, calculating the similarity in other directions by using the method of S261, and calculating the average value of the similarity in all directions, wherein the average value is the overall similarity of the target object and the non-constructors;

S263, comparing the overall similarity with a preset threshold, wherein under the condition that the overall similarity is larger than the preset threshold, the non-constructor and the target object are the same person, and under the condition that the overall similarity is smaller than or equal to the preset threshold, the non-constructor and the target object are not the same person.

Specifically, in order to judge whether the non-constructor and the target object are the same person according to the color record table, a first color record table of the front direction of the non-constructor and a second color record table of the target object are respectively obtained, for example, the technical method is more clearly explained by taking dividing a portrait area into three areas as an example, the number of the areas is not limited actually, specifically, the more similarity calculation of the area is more accurate, firstly, the average color characteristic value of each area I is calculated, the average color characteristic value calculating method is to obtain the color characteristic value of each moment, the three color values are respectively added and divided by the quantity of the moment to calculate, the first average color characteristic value of the area I of the first color record table and the second average color characteristic value of the area I of the second color record table are respectively calculated, the color similarity of the area I is calculated according to the first average color characteristic value and the second average color characteristic value, the color similarity calculation method may be calculated by using a euclidean distance algorithm, which is not explained in the prior art, in which the weight of the first color record table region and the weight of the first region in the second color record table are added and divided by two to obtain an average weight of the first region, the color similarity of the first region is multiplied by the average weight of the first region to obtain a region similarity of the first region, the region similarity of the second region and the region three is calculated by using the same method, the three region similarities are added to obtain a direction similarity of the front direction, the direction similarity of the back direction is calculated by the same method, the sum of the direction similarity of the front direction and the direction similarity of the back direction is divided by two to obtain an overall similarity, which is exemplified by the two directions of the front and the back of the human body, in practice, the method is not limited to the front direction and the back direction, and other more directions such as a left side surface direction and a right side surface direction can be provided, after the overall similarity is calculated, the overall similarity is compared with the size of a preset threshold, under the condition that the overall similarity is larger than the preset threshold, the non-constructor and the target object are the same person, and under the condition that the overall similarity is smaller than the preset threshold, the non-constructor and the target object are not the same person.

According to another aspect of the embodiment of the present invention, referring to fig. 2, there is further provided a construction safety management system based on big data, including a data collection unit, a portrait detection unit and an identification unit, for implementing a construction safety management method based on big data as described above, the specific functions of each unit are as follows:

the data collection unit is used for periodically collecting first video data shot by the first shooting equipment, collecting second video data shot by the second shooting equipment, collecting N-th video data shot by N-th shooting equipment, wherein N is a positive integer larger than 2, and storing a plurality of video data in the first memory;

the image detection unit is used for sequentially carrying out image detection on the plurality of video data, judging whether non-constructors exist in the first video data when the first video data are detected, continuously detecting the second video data when the non-constructors do not exist, recording the information data of the first non-constructors when the non-constructors exist, continuously detecting the second video data, and recording the information data of the second non-constructors when the non-constructors exist from the detection position of the second video data;

And the identification unit is used for judging whether the two non-constructors are the same person according to the recorded information data of the first non-constructor and the second non-constructor.

According to another aspect of the embodiment of the present invention, there is also provided a storage medium storing program instructions, where the program instructions, when executed, control a device in which the storage medium is located to perform the method of any one of the above.

In summary, according to the construction safety management method, the system and the storage medium based on big data, the method is that a plurality of shooting devices are installed in different construction areas of a construction site, and video data shot by the plurality of shooting devices are collected periodically; carrying out portrait detection on the plurality of video data in sequence, detecting whether non-constructors exist in the plurality of video data, sending out early warning when the non-constructors exist, informing the non-constructors to leave a construction site, and recording information data of the non-constructors; when a plurality of non-constructors are detected from the plurality of video data, whether the plurality of non-constructors are the same person or not is judged, and when the same person is judged, the early warning level is improved. The invention can help identify non-constructors at the construction site, and avoid the same non-constructors from staying at the construction site for a long time, thereby avoiding safety accidents at the construction site.

It should be understood that, although the steps in the flowcharts of the embodiments of the present invention are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in various embodiments may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor do the order in which the sub-steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of the sub-steps or stages of other steps or other steps.

Those skilled in the art will appreciate that implementing all or part of the above-described methods may be accomplished by way of computer programs, which may be stored on a non-transitory computer readable storage medium, and which, when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.

The technical features of the above embodiments may be arbitrarily combined, and for brevity, all of the possible combinations of the technical features of the above embodiments are not described, however, they should be considered as the scope of the description of the present specification as long as there is no contradiction between the combinations of the technical features.

The foregoing examples have been presented to illustrate only a few embodiments of the invention and are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims (7)

1. The construction safety management method based on big data is characterized by comprising the following steps:

S1, installing a plurality of shooting devices in different construction areas of a construction site, periodically collecting first video data shot by the first shooting device, collecting second video data shot by the second shooting device, and collecting N-th video data shot by the N-th shooting device, wherein N is a positive integer larger than 2;

s2, judging whether non-constructors exist in the first video data when the first video data are detected, sending out first-level early warning when the non-constructors exist, recording information data of the first non-constructors, detecting the second video data, recording information data of a second non-constructor when the non-constructors exist from the second video data, judging whether two non-constructors are the same person according to the recorded information data of the first non-constructors and the second non-constructors, and sending out second-level early warning when the two non-constructors are the same person;

and S3, continuously detecting third to N video data, recording information data corresponding to non-constructors every time one non-constructor is found, comparing the information data with the information data of the existing non-constructors to judge whether the information data is the same person, and sending out three-level early warning under the condition that the same person exists in the video data with the number larger than the preset number.

2. The construction safety management method based on big data according to claim 1, wherein the process of recording the information data of the first non-constructor comprises the steps of:

s21, converting the first video data to generate a time sequence image, carrying out portrait detection on a first image of the time sequence image, detecting whether a non-constructor exists in the first image, continuously detecting the next image of the time sequence image under the condition that the non-constructor does not exist, until all images of the time sequence image are detected, and taking the non-constructor as a target object under the condition that the non-constructor exists;

s22, tracking the target object, extracting a portrait area of the target object, and performing enlargement or reduction processing on the portrait area according to a preset standard;

s23, analyzing the moving direction and the face orientation of a target object, and acquiring portrait areas of the target object in different directions according to the moving direction and the face orientation, wherein during the process of tracking the target object, portrait areas of the target object in the same direction at different moments are acquired;

S24, dividing the portrait area into a plurality of areas, extracting color characteristic values in each area, adding the color characteristic values into a color record table of the target object, and recording the color characteristic values in the same direction into the same color record table, wherein the color record table comprises portrait direction, time information, portrait areas and corresponding color characteristic values;

s25, after all images of the time sequence images are detected, calculating a weight coefficient of each region according to the color record table;

s26, after the weight coefficient is calculated, storing the information data of the target object, judging whether the information data exists in an information record table before the information data of the target object is stored, if the information data does not exist, distributing an object ID and an object tag to the target object, storing table names of all color record tables corresponding to the object ID, the object tag and the target object in the information record table as one information data, if the information data exists, indicating that a non-constructor is detected when the video data before detection exists, taking out the first information data in the information record table, comparing the first information data with the information data of the target object, judging whether the non-constructor corresponding to the first information data and the target object are the same person, if the non-constructor corresponding to the first information data and the target object are the same person, distributing the object tag which is the same as the first information data to the target object, and storing the information data of the target object in the information record table if the information data of the target object is not the same person, and taking out the second information data in the information record table until the information of the information record table is compared with the information of the target object is compared.

3. The construction safety management method based on big data according to claim 2, wherein extracting the color feature value in each of the areas comprises the steps of:

s241, acquiring HSL color values of all pixel points of the area in sequence, adding tone values of the HSL color values of all pixel points, and dividing the tone values by the number of all pixel points of the area to obtain a tone average value of the area;

s242, calculating the difference value between the hue value of each pixel point and the hue average value, and marking the pixel point as a salient pixel under the condition that the difference value is larger than a preset threshold value;

s243, obtaining RGB color values of all the salient pixels, calculating RGB average values of the RGB color values of all the salient pixels, and taking the RGB average values as color characteristic values of the area.

4. The construction safety management method based on big data according to claim 2, wherein calculating the weight coefficient of each area according to the color record table comprises the steps of:

s251, obtaining color characteristic values of the same region at different moments, respectively calculating color change rates of two adjacent moments, adding the color change rates to obtain the sum of the color change rates, obtaining the number of the color change rates, dividing the sum of the color change rates by the number of the color change rates, and calculating to obtain the average color change rate of the current region;

S252, dividing one by the average color change rate, obtaining the reciprocal of the average color change rate of all areas, and taking the reciprocal as a weight coefficient of the corresponding area.

5. The construction safety management method based on big data according to claim 2, wherein the process of judging whether the non-constructor corresponding to the first piece of information data and the target object are the same person comprises the steps of:

s261, acquiring a first color record table of a first direction of the first information data, wherein each region in the color record table corresponds to color characteristic values of a plurality of moments, calculating the average color characteristic value of each region according to the color characteristic values of different moments, acquiring a second color record table of the first direction of the information data of the target object, calculating the average color characteristic value of each region in the second color record table, calculating the color similarity of each region according to the average color characteristic value of each same region, calculating the average weights of two weight coefficients corresponding to the same region in the first direction after calculating the color similarity of all regions, multiplying the color similarity of each region by the average weight of each region to obtain the region similarity of each region, and adding the region similarity of all regions to obtain the similarity of the first direction;

S262, calculating the similarity in other directions by using the method of S261, and calculating the average value of the similarity in all directions, wherein the average value is the overall similarity between the target object and the non-constructor;

s263, comparing the overall similarity with a preset threshold, wherein the non-constructor and the target object are the same person under the condition that the overall similarity is larger than the preset threshold, and the non-constructor and the target object are not the same person under the condition that the overall similarity is smaller than or equal to the preset threshold.

6. A big data based construction safety management system for implementing the method according to any of claims 1-5, comprising the following modules:

the data collection unit is used for periodically collecting first video data shot by the first shooting equipment, collecting second video data shot by the second shooting equipment, collecting N-th video data shot by N-th shooting equipment, wherein N is a positive integer larger than 2, and storing the plurality of video data in the first memory;

the image detection unit is used for sequentially carrying out image detection on a plurality of video data, judging whether non-constructors exist in the first video data when the first video data are detected, continuously detecting second video data under the condition that the non-constructors do not exist, recording information data of the first non-constructors, continuously detecting the second video data, and recording information data of the second non-constructors under the condition that the non-constructors exist at the detection position of the second video data;

And the identification unit is used for judging whether the two non-constructors are the same person according to the recorded information data of the first non-constructor and the second non-constructor.

7. A storage medium storing program instructions, wherein the program instructions, when executed, control a device in which the storage medium is located to perform the method of any one of claims 1-5.