Construction site personnel safety warning system based on image recognition

Abstract

The invention discloses a construction site personnel safety warning system based on image recognition, and relates to the technical field of construction site safety. The system comprises a monitoring unit used for acquiring a real-time video of a monitored area and transmitting the real-time video to an analysis unit; the analysis unit is used for carrying out faint analysis on the real-time video to obtain the to-be-rescued person marked as a faint state; good conditions are created for rescue deployment and control of subsequent to-be-rescued personnel, the rescue speed and effect of the to-be-rescued personnel are greatly improved, meanwhile, the panic phenomenon caused by gathering of constructors can be avoided, the site order is maintained, and the rescue efficiency is improved. And the problem of personal injury easily caused by random walking of constructors in the construction area can be avoided, and potential safety hazards are greatly reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of construction site safety, and in particular is a construction site personnel safety alarm system based on image recognition. Background Art

[0002] Construction site safety refers to a series of measures and activities implemented during the construction process to prevent accidents, protect the lives and property of construction workers and the public, and ensure the smooth progress of the construction process.

[0003] When construction workers are working in the mechanical operation area, they are prone to fainting due to the external temperature environment or their own conditions. Once they faint, the surrounding construction workers gather together. Due to the complex road conditions and the large number of mechanical equipment in the area, it is easy to cause personal injury and delay the best rescue time for the fainted person. It is difficult to judge the risk level of secondary injury of the fainted person in the existing technology, and it is difficult to match suitable investigation personnel to carry out rescue and control, which reduces the rescue effect. Therefore, in view of the above shortcomings, the present invention makes the following improvements. Summary of the invention

[0004] The present invention aims to solve the problems raised in the above background technology; to this end, the present invention proposes a construction site personnel safety warning system based on image recognition, comprising:

[0005] A monitoring unit, used to obtain real-time video of the monitoring area and transmit the real-time video to the analysis unit;

[0006] An analysis unit, used for performing fainting analysis on the real-time video to obtain persons to be rescued who are marked as fainted;

[0007] The processing unit sets a rescue coverage range according to the location information of the person to be rescued, obtains the total number of falling objects and running equipment in t statistical periods in the rescue coverage range, and calculates the average value of the falling objects and the total number of running equipment in t statistical periods, thereby obtaining the total risk value F of the person to be rescued and determining the risk level of the person to be rescued;

[0008] By obtaining the shortest distance between the construction workers and the people to be rescued within the coverage area of ​​the rescue call, as well as the scores of the construction workers in the first aid knowledge quiz, the inclusion value N of the construction workers is calculated, and the construction workers with the largest inclusion value N are selected as the search personnel.

[0009] Preferably, the specific analysis method in the analysis unit is:

[0010] Capture the video segments where human operations are present in the real-time video;

[0011] The video segment is imaged at a fixed period to obtain a number of images, which are marked as human body image data, and the fixed period is a set value;

[0012] The human body image data is compared one by one with the simulated fainting scene images in the image matching library. When the obtained human body image data matches any of the simulated fainting scene images, the construction worker is determined to be in a fainted state and is marked as a person to be rescued.

[0013] Preferably, the specific processing method in the processing unit is:

[0014] Get the location information of the person to be rescued, draw a circle with the center of gravity of the human body as the center coordinate and the preset value m as the radius, and mark the circular coverage area as the rescue coverage range;

[0015] The total number of falling objects within the coverage area of ​​the call for help in t statistical cycles C1, C2...Ct is obtained in sequence, and the discrete value S of falling objects within the coverage area of ​​the call for help in t statistical cycles is calculated by the formula. The specific calculation formula is as follows:

[0016] Where 1≤s≤t, and C is the average of the total number of falling objects within the coverage area of ​​the t statistical cycles;

[0017] When S≥U, the corresponding Cs values ​​are deleted in descending order according to |Cs-C| and the discrete value S of the remaining Cs values ​​is calculated, and S is compared with U again until S<U, where U is the discrete threshold, and the average value of the total number of remaining falling objects participating in the calculation of the discrete value S is marked as H;

[0018] Then, the total number of running devices within the coverage area of ​​the emergency call in t statistical cycles is obtained in turn, and the same processing is performed according to the above calculation method to obtain the average value of the total number of all remaining running devices participating in the discrete value calculation, and it is marked as K;

[0019] The total risk value F of the person to be rescued is calculated using the formula H+K=F, where when F>G, the person to be rescued is determined to be a high-risk person; when F<D, the person to be rescued is determined to be a low-risk person; when D≤F≤G, the person to be rescued is determined to be a medium-risk person, where D and G are both preset values.

[0020] Preferably, the specific selection method of the investigator is:

[0021] Get the shortest distance between the construction worker and the person to be rescued, mark it as J, and get the score of the construction worker in the first aid knowledge quiz before taking up the post, mark it as Q. Here, you can log in to the first aid knowledge platform, enter the name or competition number of the construction worker to view the score, and calculate the inclusion value N of the construction worker by the formula. The specific calculation formula is:

[0022] 1 / (K1×J)+K2×Q=N;

[0023] For high-risk people waiting to be rescued, K1 takes the value of 0.8 and K2 takes the value of 0.2;

[0024] For medium-risk people waiting for rescue, K1 takes the value of 0.6 and K2 takes the value of 0.4;

[0025] For low-risk people waiting to be rescued, K1 takes the value of 0.3 and K2 takes the value of 0.7;

[0026] The remaining construction workers within the coverage area of ​​the rescue call are processed in the same way as above, and according to the risk level of the person to be rescued, the inclusion value N of all construction workers within the coverage area of ​​the rescue call at the corresponding risk level is obtained, and they are compared one by one, and the construction worker with the largest inclusion value N is selected as the investigation person.

[0027] Preferably, it also includes a path unit for obtaining the shortest route for the investigator to reach the person to be rescued, and using it as the walking path and sending it to the mobile terminal of the investigator.

[0028] Preferably, it also includes a rescue control unit, which is used for the detection personnel to provide simple rescue to the rescued persons and to place warning signs in the surrounding area, and to inform the rescuers of the conditions of the rescued persons and the surrounding environment through the mobile terminal, and coordinate the rescue work.

[0029] Preferably, another method for determining the person to be rescued is provided, and the specific determination method is as follows:

[0030] Construction workers wear smart helmets at construction sites, which have built-in acceleration sensors.

[0031] The acceleration JS of the construction workers at the construction site in r collection cycles is obtained, and r acceleration values ​​are obtained, which are marked as JSr, r = 1, 2, ..., r, where the collection cycle is a preset value, and the average value P of the total number of acceleration values ​​JS of the construction workers in r collection cycles is calculated by the formula. The specific formula is as follows:

[0032]

[0033] If P≤W, it means that the acceleration value JS of the construction worker has no obvious fluctuation in r acquisition cycles, then the state of the construction worker is judged to be relatively stable and he is marked as a normal person, where W is the preset critical value;

[0034] If P>W, it means that the acceleration value JS of the construction worker has obvious fluctuations in r acquisition cycles, then the state of the construction worker is determined to be relatively turbulent and he is marked as an abnormal person.

[0035] Preferably, for the abnormal personnel, the absolute value Hr-1 of the difference between the acceleration values ​​in two adjacent acquisition cycles is calculated in the order of JS1, JS2, ..., JSR, and the absolute value with the largest value in the absolute value Hr-1 is obtained and calibrated as CK;

[0036] If CK < PD, it is determined that the abnormal person is in a normal construction state and has no signs of fainting;

[0037] If CK≥PD, the abnormal person is determined to be a person in a fainted state and needs to be rescued, where PD is a preset value.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] Based on the hidden danger of secondary injury to the rescued persons, the rescue coverage range is obtained according to the location information of the rescued persons, and for the rescued persons of different risk levels, the appropriate construction workers can be matched within the rescue coverage range according to the distance between the construction workers and the rescued persons and the scores of the construction workers in the first aid knowledge quiz, and the construction workers can be judged as scouts, thus creating good conditions for the subsequent rescue and control of the rescued persons, and greatly improving the rescue speed and effect of the rescued persons.

[0040] At the same time, it can prevent construction workers from gathering in groups and causing panic, maintain order on site, and prevent construction workers from walking around freely in the construction area, which can easily cause personal injury, greatly reducing safety hazards. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the principle framework of the present invention;

[0042] Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0043] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Embodiment 1

[0045] See also Figure 1 and Figure 2 As shown, the present application provides a construction site personnel safety warning system based on image recognition, including:

[0046] The monitoring unit and the analysis unit. The monitoring unit is used to obtain real-time video of the monitoring area and transmit the real-time video to the analysis unit. The monitoring area refers to the mechanical operation area of ​​the construction site. The mechanical operation area refers to the area where equipment that causes injuries to personnel due to improper operation is located. The analysis unit is used to perform fainting analysis on the real-time video to obtain the personnel marked as fainted and waiting to be rescued. The specific analysis method is as follows:

[0047] Capture the video segments where human operations are present in the real-time video;

[0048] The video segment is imaged at a fixed period to obtain a number of images, which are marked as human body image data. The fixed period is a set value, which is generally 5 seconds.

[0049] The human image data is compared with the simulated fainting scene images in the image matching library one by one. When the obtained human image data matches any simulated fainting scene image, the construction worker is determined to be in a fainted state and marked as a person to be rescued. The technology here is realized with the help of a neural network model. The general method is to use a hybrid CNN-HMM human action recognition method to obtain several sample data and divide them into recognition data and test data;

[0050] The recognition data is the images of people with the fainted status marked. The test data contains two states of people: fainted and not fainted, and the results are marked for both.

[0051] The recognition data is used as input and entered into the recognition model for training. The trained recognition model is marked as the initial recognition model.

[0052] After the training is completed, the test data is entered into the initial state recognition model as input to obtain the accuracy of the initial state recognition model for the test data;

[0053] If the accuracy rate exceeds the set value Z1, the initial recognition model is automatically marked as a confirmed model. Otherwise, some recognition data is acquired again and the recognition model is retrained until the accuracy rate exceeds the set value Z1.

[0054] Get the confirmed model;

[0055] The analysis unit is used to transmit the persons to be rescued to the processing unit;

[0056] The processing unit is used to analyze the risk status of the rescue personnel. The specific method of risk status analysis is as follows:

[0057] Get the location information of the person to be rescued, draw a circle with the center of gravity of the human body as the center coordinate and the preset value m as the radius, and mark the circular coverage area as the rescue coverage range;

[0058] The total number of falling objects within the coverage area of ​​the call for help in t statistical cycles C1, C2...Ct is obtained in sequence. In this embodiment, t statistical cycles are based on the current statistical cycle as the starting point and trace back t statistical cycles in the past. The discrete value S of falling objects within the coverage area of ​​the call for help in t statistical cycles is calculated by a formula. The specific calculation formula is as follows:

[0059] Where 1≤s≤t, and C is the average of the total number of falling objects within the coverage area of ​​the t statistical cycles;

[0060] When S≥U, the corresponding Cs values ​​are deleted in descending order according to |Cs-C| and the discrete value S of the remaining Cs values ​​is calculated, and S is compared with U again until S<U, where U is the discrete threshold. The average value of the total number of remaining falling objects participating in the calculation of the discrete value S at this time is marked as H;

[0061] It should be noted that the above calculation steps are to eliminate isolated values ​​whose number of falling objects in the coverage area of ​​the t statistical cycles obviously deviates from the average value. The prominent method for isolated values ​​can also use outlier detection, encouraged deep forest algorithm, etc.

[0062] Then, the total number of running devices within the coverage of the emergency call in t statistical cycles is obtained in turn, and the same processing is performed according to the above calculation method to obtain the average value of the total number of all remaining running devices participating in the discrete value calculation, and it is marked as K. The running device refers to the machine under manual or automatic control in the operating state;

[0063] The total risk value F of the person to be rescued is calculated using the formula H+K=F, wherein when F>G, the person to be rescued is determined to be a high-risk person, when F<D, the person to be rescued is determined to be a low-risk person, and when D≤F≤G, the person to be rescued is determined to be a medium-risk person, wherein D and G are both preset values;

[0064] Take a construction worker within the rescue coverage area as an example, obtain the shortest distance between the construction worker and the person to be rescued, mark it as J, and obtain the score of the construction worker in the first aid knowledge quiz before taking up the job, mark it as Q. Here, you can log in to the first aid knowledge platform, enter the name or competition number of the construction worker to view the score, and calculate the inclusion value N of the construction worker by the formula. The specific calculation formula is:

[0065] 1 / (K1×J)+K2×Q=N, where K1 and K2 are preset weights used to highlight the different importance of different elements;

[0066] For high-risk people waiting to be rescued, K1 takes the value of 0.8 and K2 takes the value of 0.2;

[0067] For medium-risk people waiting for rescue, K1 takes the value of 0.6 and K2 takes the value of 0.4;

[0068] For low-risk people waiting to be rescued, K1 takes the value of 0.3 and K2 takes the value of 0.7;

[0069] The remaining construction workers within the coverage area of ​​the rescue call are processed in the same way as above, and according to the risk level of the person to be rescued, the inclusion value N of all construction workers within the coverage area of ​​the rescue call at the corresponding risk level is obtained, and they are compared one by one, and the construction worker with the largest inclusion value N is selected as the investigation person.

[0070] Embodiment 2

[0071] Based on Example 1, please refer to Figure 1 and Figure 2 The specific improvements are as follows:

[0072] The path unit obtains the shortest route for the investigators to reach the rescued person through the navigation device, and sends it to the mobile terminal of the investigators as the walking path. The walking path is targeted at the areas where the construction site is more complex and obstructed by obstacles;

[0073] In the rescue control unit, after the survey personnel obtain the walking path, they will perform simple rescue after arriving at the person to be rescued, and set up warning signs around the person to be rescued. At the same time, the survey personnel will immediately inform the rescue personnel of the conditions of the person to be rescued and the surrounding environment through the mobile terminal, and coordinate the rescue work.

[0074] Embodiment 3

[0075] In the specific implementation process, compared with the first embodiment, this embodiment is different in that another method for determining the person to be rescued is provided, and the specific improvements are as follows:

[0076] Construction workers wear smart helmets at the construction site. The smart helmets have built-in acceleration sensors, and the model of the acceleration sensor is ADXL345.

[0077] The acceleration JS of the construction workers at the construction site in r collection cycles is obtained, and r acceleration values ​​are obtained, which are marked as JSr, r = 1, 2, ..., r, where the collection cycle is a preset value, and the average value P of the total number of acceleration values ​​JS of the construction workers in r collection cycles is calculated by the formula. The specific formula is as follows:

[0078]

[0079] If P≤W, it means that the acceleration value JS of the construction worker has no obvious fluctuation in r acquisition cycles, then the state of the construction worker is judged to be relatively stable and he is marked as a normal person, where W is the preset critical value;

[0080] If P>W, it means that the acceleration value JS of the construction worker has obvious fluctuations in r acquisition cycles, and the state of the construction worker is judged to be relatively turbulent, and he is marked as an abnormal person;

[0081] For abnormal personnel, the absolute value Hr-1 of the difference between the acceleration values ​​in two adjacent acquisition cycles is calculated in the order of JS1, JS2, ..., JSR, and the absolute value with the largest value in the absolute value Hr-1 is obtained and calibrated as CK;

[0082] If CK < PD, it is determined that the abnormal person is in a normal construction state and has no signs of fainting;

[0083] If CK≥PD, the abnormal person is determined to be a person in a fainted state and needs to be rescued, where PD is a preset value.

[0084] Embodiment 4

[0085] See also Figure 1 to Figure 2 As shown, this embodiment is obtained by combining embodiment 1 to embodiment 3.

[0086] Some of the data in the above formulas are dimensionless for numerical calculations, and the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0087] 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 construction site personnel safety warning system based on image recognition, characterized in that: include: A monitoring unit, used to obtain real-time video of the monitoring area and transmit the real-time video to the analysis unit; An analysis unit, used for performing fainting analysis on the real-time video to obtain persons to be rescued who are marked as fainted; The processing unit sets a rescue coverage range according to the location information of the rescued person, obtains the total number of falling objects and the total number of devices in operation in t statistical periods within the rescue coverage range, and then calculates the total risk value F of the rescued person to determine the risk level of the rescued person; By obtaining the shortest distance between the construction workers and the people to be rescued within the coverage area of ​​the rescue call, as well as the scores of the construction workers in the first aid knowledge quiz, the inclusion value N of the construction workers is calculated, and the construction workers with the largest inclusion value N are selected as the search personnel.

2. The construction site personnel safety warning system based on image recognition according to claim 1 is characterized in that: The specific analysis method in the analysis unit is: Capture the video segments where human operations are present in the real-time video; The video segment is imaged at a fixed period to obtain a number of images, which are marked as human body image data, and the fixed period is a set value; The human body image data is compared one by one with the simulated fainting scene images in the image matching library. When the obtained human body image data matches any of the simulated fainting scene images, the construction worker is determined to be in a fainted state and is marked as a person to be rescued.

3. The construction site personnel safety warning system based on image recognition according to claim 1 is characterized in that: The specific processing method in the processing unit is: Get the location information of the person to be rescued, draw a circle with the center of gravity of the human body as the center coordinate and the preset value m as the radius, and mark the circular coverage area as the rescue coverage range; The total number of falling objects within the coverage area of ​​the call for help in t statistical cycles C1, C2...Ct is obtained in sequence, and the discrete value S of falling objects within the coverage area of ​​the call for help in t statistical cycles is calculated by the formula. The specific calculation formula is as follows: Where 1≤s≤t, and C is the average of the total number of falling objects within the coverage area of ​​the t statistical cycles; When S≥U, the corresponding Cs values ​​are deleted in descending order according to |Cs-C| and the discrete value S of the remaining Cs values ​​is calculated, and S is compared with U again until S<U, where U is the discrete threshold, and the average value of the total number of remaining falling objects participating in the calculation of the discrete value S is marked as H; Then, the total number of running devices within the coverage area of ​​the emergency call in t statistical cycles is obtained in turn, and the same processing is performed according to the above calculation method to obtain the average value of the total number of all remaining running devices participating in the discrete value calculation, and it is marked as K; The total risk value F of the person to be rescued is calculated using the formula H+K=F, where when F>G, the person to be rescued is determined to be a high-risk person; when F<D, the person to be rescued is determined to be a low-risk person; when D≤F≤G, the person to be rescued is determined to be a medium-risk person, where D and G are both preset values.

4. The construction site personnel safety warning system based on image recognition according to claim 3 is characterized in that: The specific method for selecting investigators is as follows: Get the shortest distance between the construction worker and the person to be rescued, mark it as J, and get the score of the construction worker in the first aid knowledge quiz before taking up the post, mark it as Q. Here, you can log in to the first aid knowledge platform, enter the name or competition number of the construction worker to view the score, and calculate the inclusion value N of the construction worker by the formula. The specific calculation formula is: 1 / (K1×J)+K2×Q=N; For high-risk people waiting to be rescued, K1 takes the value of 0.8 and K2 takes the value of 0.2; For medium-risk people waiting for rescue, K1 takes the value of 0.6 and K2 takes the value of 0.4; For low-risk people waiting to be rescued, K1 takes a value of 0.3 and K2 takes a value of 0.7; The remaining construction workers within the coverage area of ​​the rescue call are processed in the same way as above, and according to the risk level of the person to be rescued, the inclusion value N of all construction workers within the coverage area of ​​the rescue call at the corresponding risk level is obtained, and they are compared one by one, and the construction worker with the largest inclusion value N is selected as the investigation person.

5. The construction site personnel safety warning system based on image recognition according to claim 1 is characterized in that: It also includes a path unit, which is used to obtain the shortest route for the investigator to reach the person to be rescued, and use it as a walking path and send it to the mobile terminal of the investigator.

6. The construction site personnel safety warning system based on image recognition according to claim 5 is characterized in that: It also includes a rescue control unit, which is used for detection personnel to provide simple rescue to the people to be rescued and to place warning signs in the surrounding area, and to inform rescue personnel of the conditions of the people to be rescued and the surrounding environment through mobile terminals to coordinate the rescue work.

7. The construction site personnel safety warning system based on image recognition according to claim 1 is characterized in that: Another method for determining the persons to be rescued is provided, and the specific determination method is as follows: Construction workers wear smart helmets at construction sites, which have built-in acceleration sensors. The acceleration JS of the construction workers at the construction site in r collection cycles is obtained, and r acceleration values ​​are obtained, which are marked as JSr, r = 1, 2, ..., r, where the collection cycle is a preset value, and the average value P of the total number of acceleration values ​​JS of the construction workers in r collection cycles is calculated by the formula. The specific formula is as follows: If P≤W, it means that the acceleration value JS of the construction worker has no obvious fluctuation in r acquisition cycles, then the state of the construction worker is judged to be relatively stable and he is marked as a normal person, where W is the preset critical value; If P>W, it means that the acceleration value JS of the construction worker has obvious fluctuations in r acquisition cycles, then the state of the construction worker is determined to be relatively turbulent and he is marked as an abnormal person.

8. The construction site personnel safety warning system based on image recognition according to claim 7 is characterized in that: For abnormal personnel, in the order of JS1, JS2, ..., JSR, calculate the absolute value Hr-1 of the difference between the acceleration values ​​in two adjacent acquisition cycles, obtain the largest absolute value in the absolute value Hr-1, and calibrate it as CK; If CK < PD, it is determined that the abnormal person is in a normal construction state and has no signs of fainting; If CK≥PD, the abnormal person is determined to be a person in a fainted state and needs to be rescued, where PD is a preset value.

Images