An unmanned aerial vehicle monitoring system and method for unsafe behavior at a construction site

Abstract

The application relates to the technical field of unmanned aerial vehicle monitoring, and discloses an unmanned aerial vehicle monitoring system and method for unsafe behaviors in a construction site, which comprises the following steps: collecting relevant data of stacked articles in a target area by an unmanned aerial vehicle, judging whether the stacked articles are normal, judging whether the stacked articles have a prompt effect, and timely performing abnormal alarm for the abnormal stacked articles without the prompt effect and the situation that articles are removed in an abnormal area. The unmanned aerial vehicle monitoring system and method for unsafe behaviors in a construction site can judge whether the position of stacked building materials in the construction site is appropriate, judge whether the stacked building materials have risks such as collapse, judge whether the stacked building materials have a certain prompt effect, timely perform alarm and prompt for the unreasonable stacked building materials, and reduce dangerous situations such as casualties caused by material collapse due to improper stacking of building materials.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) monitoring technology, specifically to a UAV monitoring system and method for unsafe behaviors at construction sites. Background Technology

[0002] Construction sites are rife with unsafe behaviors, such as workers not wearing safety helmets, carelessly discarding tools, and illegally working at heights. These behaviors not only affect construction efficiency but can also lead to serious safety accidents. Traditional safety monitoring methods are often labor-intensive and have limited coverage. Therefore, there is an urgent need for an efficient and intelligent monitoring system. By adopting a drone monitoring system, the monitoring efficiency and accuracy of unsafe behaviors at construction sites can be effectively improved, providing strong protection for construction safety. With the continuous development of drone technology, the application prospects of this system are broad and it is expected to be promoted in more fields.

[0003] The drones patrol the construction site along a pre-set path, regularly capturing videos and photos to obtain real-time images of the site. Through machine learning algorithms and image processing technology, the images captured by the drones are analyzed to identify specific unsafe behaviors. The system can continuously optimize the identification algorithm through training data to improve detection accuracy. After identifying an unsafe behavior, the drones will send the information back to the central control system in real time through the data transmission module. The central system analyzes the risk based on the set thresholds and automatically generates alarms or notifications to on-site management personnel. The system can generate detailed monitoring reports, including the time, location, and type of unsafe behavior, which facilitates tracking and statistical analysis. Management personnel can view historical records through the user interface to analyze unsafe behaviors and formulate improvement measures.

[0004] Existing drone monitoring systems and methods for unsafe behaviors at construction sites cannot determine whether the stacking of building materials is appropriate, whether there is a risk of collapse, or whether the stacking serves any warning purpose. They also cannot provide timely alarms for improper stacking of building materials, which can easily lead to material collapse and cause injuries or fatalities. Therefore, their practicality is limited. Summary of the Invention

[0005] This invention provides an unmanned aerial vehicle (UAV) monitoring system and method for unsafe behaviors at construction sites, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs), comprising:

[0007] Obtain the current monitoring area of ​​the target device;

[0008] Acquire the height data of various locations within the current monitoring area and define them as the monitoring height;

[0009] Obtain all monitoring heights within the current monitoring area and form a height set;

[0010] Obtain reference height data;

[0011] If the monitored height equals the baseline height data, then it is determined that no object exists. Move the target device, update the current monitoring area, and repeat the above steps.

[0012] If the monitored height is not equal to the reference height data, it is determined that an object exists, and all the sampling edge points of the object are obtained to form a set of sampling points;

[0013] Obtain the analysis edge point corresponding to each element in the collection point set to form the analysis point set;

[0014] If the number of elements in the analysis point set is not equal to the number of elements in the collection point set, then the target device is determined to have error data.

[0015] If the number of elements in the analysis point set equals the number of elements in the collection point set, then the target device is determined to have no error data.

[0016] Obtain the judgment region;

[0017] If all elements in the set of analysis points are within the decision region, then there are fewer unsafe factors.

[0018] If there are elements in the set of analysis points that are not within the decision region, then there are many unsafe factors.

[0019] As described in this invention, the method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs) includes: acquiring all edge points of the object; specifically:

[0020] A1. Obtain all initial images of the item;

[0021] A2. Capture all the edge lines of the object and form a set of edge lines;

[0022] A3. Identify each element in the edge line set as the target edge line in turn;

[0023] A4. Obtain all data points on the target edge line to form a data point set;

[0024] A5. Obtain the coordinates of each element in the data point set and define them as data coordinates;

[0025] A6. Match each data coordinate with each element in the data point set to form a coordinate set;

[0026] A7. Match each set of coordinates with each element in the set of edge lines to form an edge dataset;

[0027] A8. Extract each element from the edge dataset sequentially and identify it as the analysis element;

[0028] A9. Define the edge dataset after removing the analysis elements as the analysis dataset;

[0029] A10. Compare the element to be analyzed with each element in the dataset in turn;

[0030] If there is no element in the dataset that is the same as the element being analyzed, then it is determined that the element being analyzed does not have any overlapping elements.

[0031] If the analysis dataset contains an element that is identical to the element being analyzed, then the analysis element is determined to have overlapping elements.

[0032] A11. Analyzed elements that are determined to have overlapping elements are identified as edge points for data collection.

[0033] As described in this invention, the method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs) includes: acquiring the analysis edge points corresponding to each element in the collection point set, specifically as follows:

[0034] Get all initial images of the item;

[0035] Acquire the device data when the target device acquires each initial image, and define it as the target device data;

[0036] Obtain the location of the item within the current monitoring area and designate it as the comparison location;

[0037] Obtain the restriction data of the target device;

[0038] Based on the constraint data, comparison location, and data from each target device, adjust the images captured by the target devices for the items, and designate them as comparison images;

[0039] Based on the comparison location and target device data, adjust the image captured by the target device on the item and set it as the comparison image;

[0040] Execute A2-A10;

[0041] Analytical elements that are found to have overlapping elements are identified as analytical edge points.

[0042] As described in this invention, a drone-based method for monitoring unsafe behaviors at construction sites includes: if the target device is determined to have error data, then error analysis is performed on the collected data.

[0043] Get all comparison images of the item;

[0044] Extract the number of obstacles in the comparison images;

[0045] Acquire the footprint data for each obstacle, including volume data and area data;

[0046] Calculate the volume impact data in the comparison image. The volume impact data = the sum of the volume data of each obstacle under the number of obstacles.

[0047] Calculate the area impact data in the comparison image. The area impact data = the sum of the area data of each obstacle under the number of obstacles.

[0048] If the volume impact data is greater than or equal to the area impact data × 15%, then the placement of the item is considered reliable, and there are relatively few unsafe factors.

[0049] If the volume impact data is less than 15% of the area impact data, the item placement is deemed unreliable, and an alarm is triggered.

[0050] As described in this invention, a drone-based method for monitoring unsafe behaviors at construction sites includes: if the number of unsafe factors is determined to be small, then a stacking analysis of the items is performed.

[0051] Acquire all items within the current monitoring area and form an item set;

[0052] Retrieve the type data for each item in the item collection;

[0053] If all items in the item set have the same type data, then any item extracted from the item set is identified as the target item.

[0054] Obtain the center point of the target item and set it as the target center point;

[0055] Retrieve all items in the item set that are adjacent to the target item, forming an analysis item set;

[0056] Each element in the set of items to be analyzed is identified as an item to be analyzed in turn;

[0057] Obtain all surfaces of the target object to form a first set of surfaces;

[0058] Extract the surface from the first set of surfaces that is closest to the object being analyzed, and define it as the first plane.

[0059] Obtain all surfaces of the analyzed object to form a second set of surfaces;

[0060] Extract the surface from the second set that is closest to the target object and define it as the second plane;

[0061] Several data analysis points are set on the first plane;

[0062] Obtain the vertical distance between each data analysis point and the second plane, and define it as the analysis distance;

[0063] Obtain all analyzed distances and form a distance set;

[0064] Set the judgment distance;

[0065] If the analysis distance is less than or equal to the judgment distance × (1 + 50%), then the judgment distance interval is small.

[0066] If the analysis distance is greater than the judgment distance × (1 + 50%), then the judgment distance interval is large;

[0067] If the distance intervals of all elements in the set are considered small, then the item intervals are considered small.

[0068] If the distance between the elements in the set is large, then the distance between the items is large.

[0069] If the analysis of all elements in the set of items determines that the intervals between items are small, then the items are considered stable.

[0070] If the analysis shows that there are elements in the set of items with large intervals between them, then the items are considered unstable.

[0071] Get the number of elements in the item set that are considered stable, and define it as the stable quantity;

[0072] Get the number of elements in the item collection, and define it as the item quantity;

[0073] If the stable quantity is greater than or equal to 90% of the item quantity, the item stacking is considered normal and no abnormal alarm is triggered.

[0074] If the stable quantity is less than 90% of the item quantity, the item stacking is determined to be abnormal, and an abnormal alarm is triggered.

[0075] As a method for monitoring unsafe behaviors at construction sites according to the present invention, if the number of unsafe factors is small, a stacking analysis of the items is performed, and the method further includes:

[0076] Acquire all items within the current monitoring area and form an item set;

[0077] Retrieve the type data for each item in the item collection;

[0078] If the type data of all items in the item collection are not of the same type, then each type data will be identified as the target type in turn;

[0079] Extract all items of the target type within the item set to form a decision set;

[0080] Randomly extract any item from the judgment set and designate it as the judgment item;

[0081] The set of decisions to remove the decision items is defined as the decision analysis set;

[0082] Each item in the decision analysis set is sequentially identified as a scoring item;

[0083] The number of items between the judgment item and the scoring item is defined as the judgment quantity;

[0084] If the number of judgments is 0, then the item being judged is judged to be a continuous item;

[0085] If the number of judgments is not equal to 0, then all items in the interval between the judgment item and the scoring item are extracted and defined as the interval items;

[0086] Obtain the type data for each interval item and define it as the analysis type;

[0087] If the analysis type equals the target type, then the items in between are determined to be the same item;

[0088] If the analysis type is not the same as the target type, then the items in between are determined to be different items.

[0089] If all the interval items are the same item, then the item being judged is considered a consecutive item;

[0090] If there are different items among all the interval items, the item being judged is determined to be a discontinuous item;

[0091] If all elements in the analysis set are consecutive items, then the item set is determined.

[0092] If there are elements in the analysis set that are discontinuous items, then the items are determined to be scattered.

[0093] If all elements in the item set are determined to be in a stacked state, then the item stacking is considered normal, and no abnormal alarm is triggered.

[0094] If the items in the collection are found to be scattered, then the items are considered to be stacked abnormally, and an alarm is triggered.

[0095] As described in this invention, a drone-based method for monitoring unsafe behaviors at construction sites includes: if multiple unsafe factors are identified, an environmental analysis of the objects is performed.

[0096] B1. Obtain the target area;

[0097] B2. Obtain data for all regions within the target area;

[0098] B3. Obtain all items within the current monitoring area and form an item set;

[0099] B4. Obtain the spatial coordinates of each item in the item collection and define them as item coordinates;

[0100] B5. Extract the coordinates of any item and designate it as the target coordinates;

[0101] B6. Obtain the distance between the coordinates of each item and the target coordinates, and define it as the coordinate distance;

[0102] B7. Select the coordinates of the item with the smallest numerical distance and consider them as the selected coordinates;

[0103] B8. Connect the target coordinates and the selected coordinates to form a reference connection line;

[0104] B9. Remove the target coordinates from the item collection and update the item collection;

[0105] B10. Update the selected coordinates to the target coordinates;

[0106] B11. Repeat B6-B11 until there are no more elements in the updated item set, and then take the selection coordinates corresponding to the last formed reference connection line as the termination coordinates.

[0107] B12. Use the target coordinates in B5 as the initial coordinates;

[0108] B13. Connect the initial coordinates and the final coordinates to form an analysis region;

[0109] B14. Obtain regional data within the analysis area and designate it as the analysis region data;

[0110] The regional data includes general regional data and special regional data;

[0111] B15. If the analyzed area data is ordinary area data, it is determined that there is no prompt for item stacking and an abnormal alarm is triggered;

[0112] If the data analyzed is from a special region, an alert will be issued indicating that there is an accumulation of items, and the analyzed area will be marked as an abnormal area for focused monitoring.

[0113] As described in this invention, a drone-based method for monitoring unsafe behaviors at construction sites includes: focusing on monitoring abnormal areas, including:

[0114] Obtain the monitoring interval duration of the target device;

[0115] Obtain the device condition data for the target area and designate it as the target condition data;

[0116] The device condition data includes both present conditions and non-present conditions;

[0117] If the target condition data is an existing condition, then the monitoring interval duration is cancelled;

[0118] Select any target device to monitor the abnormal area;

[0119] If the target condition data is non-existent, then calculate the adjustment interval duration: adjustment interval duration = monitoring interval duration × 20%;

[0120] Adjust the monitoring interval of the target device to monitor the abnormal area.

[0121] An alarm will be triggered if an item is removed from the abnormal area.

[0122] The present invention also discloses a system for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs), comprising:

[0123] Data Acquisition Module: Collects relevant data on the stacking of items within the target area using drones;

[0124] Analysis module: Analyzes areas within the target area where items are stacked to determine whether the stacking is normal;

[0125] Judgment module: Analyzes abnormal item stacking situations to determine whether the item stacking serves as a prompt.

[0126] Alarm module: It will promptly issue alarms for abnormal stacking of items without any prompts, as well as for the removal of items from abnormal areas, to prevent accidents.

[0127] The present invention has the following beneficial effects:

[0128] 1. The drone monitoring system and method for unsafe behaviors at the construction site acquires height data from various locations within the construction site to determine if there is any stacking of items. By comparing the data captured in the initial image of the items with the data captured in the currently collected comparison image, it determines whether there is any missing data, thereby determining whether there is any error in the data collected by the drone. If there is no error, it determines whether the placement of the items is reliable and whether there is any risk. It promptly issues alarms for unreasonable stacking of building materials to reduce the risk of dangerous situations such as material collapse and personal injury caused by improper stacking of building materials, and to prevent the stacking of building materials from affecting the operation of construction equipment, and to prevent the operation of construction equipment from causing the collapse of building materials and resulting in danger.

[0129] 2. The drone monitoring system and method for unsafe behaviors at the construction site: If the data collected by the drone has errors, the system can determine whether the items are stacked in a suitable area by obtaining the location of the items. If they are in a suitable area, the system can determine whether there are safety hazards in the stacking of items by obtaining data such as the material, size, and spacing between items, i.e., whether there is a risk of the stacking of items collapsing. If there are safety hazards, an alarm will be issued in time. If there are no safety hazards, monitoring can continue, thereby reducing the risk of dangerous situations such as personnel injury or death caused by the collapse of building materials due to improper stacking.

[0130] 3. The drone monitoring system and method for unsafe behaviors at the construction site: If items are not stacked in a suitable area, the system acquires terrain data of the area and spatial coordinates of each item to determine if the stacking serves as a warning. Specifically, it determines whether personnel are stacking items to form a closed area, enclosing special terrain features such as elevator shafts within this closed area. If a warning is present, the closed area is marked and monitored closely. If no warning is present, an alarm is triggered promptly. This reduces the risk of material collapse and personal injury caused by improper stacking of building materials, prevents workers from falling due to neglect of the terrain during transportation and work, and avoids situations where some personnel fail to promptly notify others after removing stacked items, leading to dangerous situations due to delayed information transmission and lack of environmental awareness. Attached Figure Description

[0131] Figure 1 This is a flowchart of the drone monitoring method for unsafe behaviors at construction sites according to the present invention;

[0132] Figure 2 This is a system block diagram of the UAV monitoring method for unsafe behaviors at construction sites according to the present invention. Detailed Implementation

[0133] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0134] Example 1: A method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs). (See attached document.) Figure 1 ,include:

[0135] Obtain the current monitoring area of ​​the target device, which is a drone;

[0136] Acquire the height data of various locations within the current monitoring area and define it as the monitoring height. The height data is the height distance between a certain point and the ground.

[0137] Obtain all monitoring heights within the current monitoring area and form a height set;

[0138] Obtain reference height data, where the reference height data is 0, i.e., the ground;

[0139] If the monitored height equals the baseline height data, then it is determined that no object exists. Move the target device, update the current monitoring area, and repeat the above steps.

[0140] If the monitored height is not equal to the reference height data, it is determined that an object exists, and all the sampling edge points of the object are obtained to form a set of sampling points;

[0141] Obtain the analysis edge point corresponding to each element in the collection point set to form the analysis point set;

[0142] If the number of elements in the analysis point set is not equal to the number of elements in the collection point set, then the target device is determined to have error data.

[0143] If the number of elements in the analysis point set equals the number of elements in the collection point set, then the target device is determined to have no error data.

[0144] Obtain the determination area, which is a pre-planned area where items can be placed;

[0145] If all elements in the set of analysis points are within the judgment area, then there are fewer unsafe factors, meaning that the items are all placed in the pre-planned area where items can be placed.

[0146] If there are elements in the set of analysis points that are not within the judgment area, then there are many unsafe factors, that is, there are items that are not placed within the judgment area.

[0147] Specifically, acquiring all the edge points of the item means:

[0148] A1. Acquire all initial images of the item, wherein the initial images are images of the item captured by the drone when the item is placed in the current monitoring area and is located in an area where the drone can collect complete data of the item;

[0149] A2. Capture all the edge lines of the object and form a set of edge lines;

[0150] A3. Identify each element in the edge line set as the target edge line in turn;

[0151] A4. Obtain all data points on the target edge line to form a data point set;

[0152] A5. Obtain the coordinates of each element in the data point set and define them as data coordinates;

[0153] A6. Match each data coordinate with each element in the data point set to form a coordinate set;

[0154] A7. Match each coordinate set with each element in the edge line set to form an edge dataset. Since the object is in a static state when all initial images of the object are collected, all coordinates of the object collected at this time are within the same coordinate axis, that is, the data is referenceable and has small error.

[0155] A8. Extract each element from the edge dataset sequentially and identify it as the analysis element;

[0156] A9. Define the edge dataset after removing the analysis elements as the analysis dataset;

[0157] A10. Compare the element to be analyzed with each element in the dataset in turn;

[0158] If there is no element in the dataset that is the same as the element being analyzed, then it is determined that the element being analyzed does not have any overlapping elements.

[0159] If the analysis dataset contains an element that is identical to the element being analyzed, then the analysis element is determined to have overlapping elements.

[0160] A11. Analyzed elements that are determined to have overlapping elements are identified as edge points for data collection.

[0161] This embodiment also provides that the analysis edge point corresponding to each element in the collection point set is specifically as follows:

[0162] Get all initial images of the item;

[0163] The device data acquired when the target device collects each initial image is defined as the target device data. The device data includes parameters such as the relative position and distance between the drone and the object when collecting the target image.

[0164] Obtain the location of the item within the current monitoring area and designate it as the comparison location;

[0165] Acquire the limiting data of the target device. The limiting data refers to the limiting conditions that prevent the drone from normally acquiring images, such as the presence of obstacles that prevent the drone from reaching the predetermined location to acquire data.

[0166] Based on the constraint data, comparison location, and data of each target device, the image of the object collected by the target device is adjusted and designated as the comparison image. The comparison image is the image of the object in the current monitoring area after the drone adjusts its attitude and position to the data when collecting the target image, ensuring that the comparison image and the target image are two images with basically consistent data and can be compared.

[0167] Based on the comparison location and target device data, the image of the object collected by the target device is adjusted and designated as the comparison image. The comparison image is the image of the object in the current monitoring area after the drone adjusts its attitude and position to the data when collecting the target image, ensuring that the comparison image and the target image are two images with basically consistent data and can be compared.

[0168] Execute A2-A10;

[0169] Analytical elements that are found to have overlapping elements are identified as analytical edge points.

[0170] By employing the methods described above, machine vision technology is used to analyze and process images and videos from the construction site, identifying non-compliant operational behaviors and potential hazards. A real-time early warning system is then constructed. Combining the results of hazardous behavior identification in construction and the analysis of construction worker behavior, the system can promptly detect dangerous behaviors and send early warning information. This information will be disseminated through alarms, vibrations or whistles from matching safety wristbands, or by connecting to the construction worker's personal mobile phone for timely reminders. The system can automatically identify non-compliant operations and potential hazards, monitor the behavior of construction workers in real time, and provide accurate early warning information so that timely and effective measures can be taken to ensure the safety of the construction site.

[0171] Example 2 is an improvement upon Example 2. This method for monitoring unsafe behaviors at construction sites using drones involves analyzing the collected data of the target equipment if error data is detected.

[0172] Get all comparison images of the item;

[0173] Extract the number of obstacles in the comparison images;

[0174] Acquire the footprint data for each obstacle, which includes volume data and area data. The footprint data is the volume of the obstacle and the area it occupies. The volume data is the volume of the obstacle and the area data is the area it occupies. This is used to determine the type of obstacle, such as a wall having a large volume and a large footprint, while a tower crane has a large volume and a small footprint.

[0175] Calculate the volume impact data in the comparison image. The volume impact data = the sum of the volume data of each obstacle under the number of obstacles.

[0176] Calculate the area impact data in the comparison image. The area impact data = the sum of the area data of each obstacle under the number of obstacles.

[0177] If the volume impact data is greater than or equal to the area impact data × 15%, then the placement of the item is considered reliable and there are few unsafe factors. That is, if the building materials are placed near fixed objects such as walls, the probability of the items colliding or being dangerous is low. As long as the building materials are stacked reasonably, the building materials can be placed in this location.

[0178] If the volume impact data is less than the area impact data × 15%, the placement of the item is deemed unreliable, and an abnormal alarm is triggered. This means that if building materials are placed near equipment such as tower cranes, there is a risk of collision or other dangers when the equipment is in operation. The building materials need to be moved to a suitable location as soon as possible.

[0179] Example 3 is an improvement on Example 3. In this example, if the number of unsafe factors is determined to be small, a stacking analysis of the items is performed:

[0180] Acquire all items within the current monitoring area and form an item set;

[0181] Obtain the type data of each item in the item collection. The type data includes the material, size, and other data of the item, in order to determine whether the stacked items in the area are the same type of item.

[0182] If all items in the item set have the same type data, then any item extracted from the item set is identified as the target item.

[0183] Obtain the center point of the target item and set it as the target center point;

[0184] Retrieve all items in the item set that are adjacent to the target item, forming an analysis item set;

[0185] Each element in the set of items to be analyzed is identified as an item to be analyzed in turn;

[0186] Obtain all surfaces of the target object to form a first set of surfaces;

[0187] Extract the surface from the first set of surfaces that is closest to the object being analyzed, and define it as the first plane.

[0188] Obtain all surfaces of the analyzed object to form a second set of surfaces;

[0189] Extract the surface from the second set that is closest to the target object and define it as the second plane;

[0190] Several data analysis points are set on the first plane;

[0191] Obtain the vertical distance between each data analysis point and the second plane, and define it as the analysis distance;

[0192] Obtain all analyzed distances and form a distance set;

[0193] A judgment distance is set, which is the gap distance between the top and bottom items when the items are stacked normally. If the gap distance between the top, bottom, left and right items and the items is different when the items are stacked, it is easy to cause the center of gravity to be unstable, which can easily lead to dangerous situations such as the items collapsing.

[0194] If the analysis distance is less than or equal to the judgment distance × (1 + 50%), then the judgment distance interval is small.

[0195] If the analysis distance is greater than the judgment distance × (1 + 50%), then the judgment distance interval is large;

[0196] If the distance intervals of all elements in the set are considered small, then the item intervals are considered small.

[0197] If the distance between the elements in the set is large, then the distance between the items is large.

[0198] If the analysis of all elements in the set of items determines that the intervals between items are small, then the items are considered stable.

[0199] If the analysis shows that there are elements in the set of items with large intervals between them, then the items are considered unstable.

[0200] Get the number of elements in the item set that are considered stable, and define it as the stable quantity;

[0201] Get the number of elements in the item collection, and define it as the item quantity;

[0202] If the stable quantity is greater than or equal to 90% of the item quantity, the item stacking is considered normal and no abnormal alarm is triggered.

[0203] If the stable quantity is less than 90% of the item quantity, the item stacking is determined to be abnormal, and an abnormal alarm is triggered.

[0204] If the number of unsafe factors is determined to be small, a stacking analysis of the items will be performed, which also includes:

[0205] Acquire all items within the current monitoring area and form an item set;

[0206] Obtain the type data of each item in the item collection. The type data includes the material, size, and other data of the item, in order to determine whether the stacked items in the area are the same type of item.

[0207] If the type data of all items in the item collection are not of the same type, then each type data will be identified as the target type in turn;

[0208] Extract all items of the target type within the item set to form a decision set;

[0209] Randomly extract any item from the judgment set and designate it as the judgment item;

[0210] The set of decisions to remove the decision items is defined as the decision analysis set;

[0211] Each item in the decision analysis set is sequentially identified as a scoring item;

[0212] The number of items between the judgment item and the scoring item is defined as the judgment quantity;

[0213] If the number of judgments is 0, then the item being judged is judged to be a continuous item;

[0214] If the number of judgments is not equal to 0, then all items in the interval between the judgment item and the scoring item are extracted and defined as the interval items;

[0215] Obtain the type data for each interval item and define it as the analysis type;

[0216] If the analysis type equals the target type, then the items in between are determined to be the same item;

[0217] If the analysis type is not the same as the target type, then the items in between are determined to be different items.

[0218] If all the interval items are the same item, then the item being judged is considered a consecutive item;

[0219] If there are different items among all the interval items, the item being judged is determined to be a discontinuous item;

[0220] If all elements in the analysis set are consecutive items, then the item set is determined.

[0221] If there are elements in the analysis set that are discontinuous items, then the items are determined to be scattered.

[0222] If all elements in the item set are determined to be in a stacked state, then the item stacking is considered normal, and no abnormal alarm is triggered.

[0223] If the items in the collection are found to be scattered, then the items are considered to be stacked abnormally, and an alarm is triggered.

[0224] This embodiment also provides that if there are too many unsafe factors, an environmental analysis of the item will be performed:

[0225] B1. Obtain the target area, which is all areas monitored by the UAV at the construction site;

[0226] B2. Obtain all regional data for the target area. The regional data refers to the structural distribution within the area, such as the distribution data of structures like stairs and walls within the area.

[0227] B3. Obtain all items within the current monitoring area and form an item set;

[0228] B4. Obtain the spatial coordinates of each item in the item collection and define them as item coordinates;

[0229] B5. Extract the coordinates of any item and designate it as the target coordinates;

[0230] B6. Obtain the distance between the coordinates of each item and the target coordinates, and define it as the coordinate distance;

[0231] B7. Select the coordinates of the item with the smallest numerical distance and consider them as the selected coordinates;

[0232] B8. Connect the target coordinates and the selected coordinates to form a reference connection line;

[0233] B9. Remove the target coordinates from the item collection and update the item collection;

[0234] B10. Update the selected coordinates to the target coordinates;

[0235] B11. Repeat B6-B11 until there are no more elements in the updated item set, and then take the selection coordinates corresponding to the last formed reference connection line as the termination coordinates.

[0236] B12. Use the target coordinates in B5 as the initial coordinates;

[0237] B13. Connect the initial coordinates and the final coordinates to form an analysis region;

[0238] B14. Obtain regional data within the analysis area and designate it as the analysis region data;

[0239] The regional data includes ordinary regional data and special regional data. The ordinary regional data refers to common terrain such as the ground that is unlikely to cause danger to people, while the special regional data refers to special terrain such as staircases and elevator shafts that are likely to cause danger to people.

[0240] B15. If the analyzed area data is ordinary area data, it is determined that there is no prompt for item stacking and an abnormal alarm is triggered;

[0241] If the data analyzed is from a special region, then the stacking of items is considered a warning sign. The analyzed area is marked as an abnormal area for focused monitoring. This means that the area contains terrain such as elevator shafts that could easily cause people to fall. Stacking items in this area to form a closed area may be to prevent workers from falling due to not noticing the terrain during transportation and work. The location is marked, and the monitoring interval for this location is shortened. Real-time monitoring should be carried out as much as possible when conditions permit. This is to avoid situations where some people do not promptly notify other people after removing the stacked items, leading to dangerous situations due to untimely information transmission and lack of attention to the environmental conditions.

[0242] Among these measures, key monitoring will be conducted on abnormal areas, including:

[0243] The monitoring interval duration of the target device is obtained. The monitoring interval duration is the duration between two monitoring sessions for each monitoring area, which is manually controlled according to the size and complexity of the target area.

[0244] Acquire equipment condition data for the target area and define it as target condition data. The equipment condition data refers to whether there are sufficient drones to monitor the construction site.

[0245] The equipment condition data includes existing conditions and non-existent conditions. The existing condition is that there are more than two drones monitoring the target area of ​​the construction site, indicating that there are enough drones monitoring the construction site. The non-existent condition is that there are no more than two drones monitoring the target area of ​​the construction site, indicating that there are not enough drones monitoring the construction site.

[0246] If the target condition data is an existing condition, then the monitoring interval duration is cancelled;

[0247] Select any target device to monitor the abnormal area;

[0248] If the target condition data is non-existent, then calculate the adjustment interval duration: adjustment interval duration = monitoring interval duration × 20%;

[0249] Adjust the monitoring interval of the target device to monitor the abnormal area.

[0250] When items are removed from an abnormal area, an alarm is triggered, prompting staff to promptly set up warning signs and other facilities. This prevents situations where some staff members fail to notify others in time after removing stacked items, leading to dangerous situations due to delayed information transmission and lack of awareness of the environment.

[0251] Example 4: This example also discloses a system for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs). See [link to relevant documentation]. Figure 2 ,include:

[0252] Data Acquisition Module: Collects relevant data on the stacking of items within the target area using drones;

[0253] Analysis module: Analyzes areas within the target area where items are stacked to determine whether the stacking is normal;

[0254] Judgment module: Analyzes abnormal item stacking situations to determine whether the item stacking serves as a prompt.

[0255] Alarm module: It will promptly issue alarms for abnormal stacking of items without any prompts, as well as for the removal of items from abnormal areas, to prevent accidents.

[0256] This embodiment addresses the challenges of accurately monitoring and promptly identifying non-standard operations and potential hazards during construction, while simultaneously pinpointing the responsible construction personnel and providing direct, effective, and timely early warnings and interventions. Leveraging machine vision-based construction site safety monitoring technology, it proposes using cameras, sensors, and other equipment to acquire data from the construction site. This aims to overcome the limitations of traditional construction site monitors, which lack intelligent identification of hazardous behaviors and proactive AI intervention for early warning. By collecting a large amount of data, a database of hazardous behavior image recognition will be established to improve construction safety technology research. A hazardous behavior sample library will be created, and machine learning through big data will enhance the accuracy of recognizing and judging different construction hazards. Based on this, the data acquired through machine vision will be processed and analyzed to solve the problems of real-time intelligent monitoring of construction site safety status and prediction and early warning intervention of hazardous behaviors.

Claims

1. A method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs), characterized in that: include: Obtain the current monitoring area of ​​the target device; Acquire the height data of various locations within the current monitoring area and define them as the monitoring height; Obtain all monitoring heights within the current monitoring area and form a height set; Obtain reference height data; If the monitored height equals the baseline height data, then it is determined that no object exists. Move the target device, update the current monitoring area, and repeat the above steps. If the monitored height is not equal to the reference height data, it is determined that an object exists, and all the sampling edge points of the object are obtained to form a set of sampling points; Obtain the analysis edge point corresponding to each element in the collection point set to form the analysis point set; If the number of elements in the analysis point set is not equal to the number of elements in the collection point set, then the target device is determined to have error data. If the number of elements in the analysis point set equals the number of elements in the collection point set, then the target device is determined to have no error data. Obtain the judgment region; If all elements in the set of analysis points are within the decision region, then there are fewer unsafe factors. If there are elements in the set of analysis points that are not within the decision region, then there are many unsafe factors in the decision. If the number of unsafe factors is determined to be small, then a stacking analysis of the items is performed: Acquire all items within the current monitoring area and form an item set; Retrieve the type data for each item in the item collection; If all items in the item set have the same type data, then any item extracted from the item set is identified as the target item. Obtain the center point of the target item and set it as the target center point; Retrieve all items in the item set that are adjacent to the target item, forming an analysis item set; Each element in the set of items to be analyzed is identified as an item to be analyzed in turn; Obtain all surfaces of the target object to form a first set of surfaces; Extract the surface from the first set of surfaces that is closest to the object being analyzed, and define it as the first plane. Obtain all surfaces of the analyzed object to form a second set of surfaces; Extract the surface from the second set that is closest to the target object and define it as the second plane; Several data analysis points are set on the first plane; Obtain the vertical distance between each data analysis point and the second plane, and define it as the analysis distance; Obtain all analyzed distances and form a distance set; Set the judgment distance; If the analysis distance is less than or equal to the judgment distance × (1 + 50%), then the judgment distance interval is small. If the analysis distance is greater than the judgment distance × (1 + 50%), then the judgment distance interval is large; If the distance intervals of all elements in the set are considered small, then the item intervals are considered small. If the distance between the elements in the set is large, then the distance between the items is large. If the analysis of all elements in the set of items determines that the intervals between items are small, then the items are considered stable. If the analysis shows that there are elements in the set of items with large intervals between them, then the items are considered unstable. Get the number of elements in the item set that are considered stable, and define it as the stable quantity; Get the number of elements in the item collection, and define it as the item quantity; If the stable quantity is greater than or equal to 90% of the item quantity, the item stacking is considered normal and no abnormal alarm is triggered. If the stable quantity is less than 90% of the item quantity, the item stacking is determined to be abnormal, and an abnormal alarm is triggered. If the number of unsafe factors is determined to be small, a stacking analysis of the items is performed, which also includes: Acquire all items within the current monitoring area and form an item set; Retrieve the type data for each item in the item collection; If the type data of all items in the item collection are not of the same type, then each type data will be identified as the target type in turn; Extract all items of the target type within the item set to form a decision set; Randomly extract any item from the judgment set and designate it as the judgment item; The set of decisions to remove the decision items is defined as the decision analysis set; Each item in the decision analysis set is sequentially identified as a scoring item; The number of items between the judgment item and the scoring item is defined as the judgment quantity; If the number of judgments is 0, then the item being judged is judged to be a continuous item; If the number of judgments is not equal to 0, then all items in the interval between the judgment item and the scoring item are extracted and defined as the interval items; Obtain the type data for each interval item and define it as the analysis type; If the analysis type equals the target type, then the items in between are determined to be the same item; If the analysis type is not the same as the target type, then the items in between are determined to be different items. If all the interval items are the same item, then the item being judged is considered a consecutive item; If there are different items among all the interval items, the item being judged is determined to be a discontinuous item; If all elements in the analysis set are consecutive items, then the item set is determined. If there are elements in the analysis set that are discontinuous items, then the items are determined to be scattered. If all elements in the item set are determined to be in a stacked state, then the item stacking is considered normal, and no abnormal alarm is triggered. If the items in the collection are found to be scattered, then the items are determined to be stacked abnormally, and an abnormal alarm is triggered. If there are too many unsafe factors, an environmental analysis of the items should be conducted: B1. Obtain the target area; B2. Obtain data for all regions within the target area; B3. Obtain all items within the current monitoring area and form an item set; B4. Obtain the spatial coordinates of each item in the item collection and define them as item coordinates; B5. Extract the coordinates of any item and designate it as the target coordinates; B6. Obtain the distance between the coordinates of each item and the target coordinates, and define it as the coordinate distance; B7. Select the coordinates of the item with the smallest numerical distance and consider them as the selected coordinates; B8. Connect the target coordinates and the selected coordinates to form a reference connection line; B9. Remove the target coordinates from the item collection and update the item collection; B10. Update the selected coordinates to the target coordinates; B11. Repeat B6-B11 until there are no more elements in the updated item set, and then take the selection coordinates corresponding to the last formed reference connection line as the termination coordinates. B12. Use the target coordinates in B5 as the initial coordinates; B13. Connect the initial coordinates and the final coordinates to form an analysis region; B14. Obtain regional data within the analysis area and designate it as the analysis region data; The regional data includes general regional data and special regional data; B15. If the analyzed area data is ordinary area data, it is determined that there is no prompt for item stacking and an abnormal alarm is triggered; If the data analyzed is from a special region, an alert will be issued indicating that there is an accumulation of items, and the analyzed area will be marked as an abnormal area for focused monitoring.

2. The method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The acquisition of all edge points of the item is specifically as follows: A1. Obtain all initial images of the item; A2. Capture all the edge lines of the object and form a set of edge lines; A3. Identify each element in the edge line set as the target edge line in turn; A4. Obtain all data points on the target edge line to form a data point set; A5. Obtain the coordinates of each element in the data point set and define them as data coordinates; A6. Match each data coordinate with each element in the data point set to form a coordinate set; A7. Match each set of coordinates with each element in the set of edge lines to form an edge dataset; A8. Extract each element from the edge dataset sequentially and identify it as the analysis element; A9. Define the edge dataset after removing the analysis elements as the analysis dataset; A10. Compare the element to be analyzed with each element in the dataset in turn; If there is no element in the dataset that is the same as the element being analyzed, then it is determined that the element being analyzed does not have any overlapping elements. If the analysis dataset contains an element that is identical to the element being analyzed, then the analysis element is determined to have overlapping elements. A11. Analyzed elements that are determined to have overlapping elements are identified as edge points for data collection.

3. The method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs) according to claim 2, characterized in that: The process of acquiring the analysis edge point corresponding to each element in the collection point set specifically involves: Get all initial images of the item; Acquire the device data when the target device acquires each initial image, and define it as the target device data; Obtain the location of the item within the current monitoring area and designate it as the comparison location; Obtain the restriction data of the target device; Based on the constraint data, comparison location, and data from each target device, adjust the images captured by the target devices for the items, and designate them as comparison images; Based on the comparison location and target device data, adjust the image captured by the target device on the item and set it as the comparison image; Execute A2-A10; Analytical elements that are found to have overlapping elements are identified as analytical edge points.

4. The method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: If the target device is determined to have erroneous data, then an error analysis of the data collection is performed on the item: Get all comparison images of the item; Extract the number of obstacles in the comparison images; Acquire the footprint data for each obstacle, including volume data and area data; Calculate the volume impact data in the comparison image. The volume impact data = the sum of the volume data of each obstacle under the number of obstacles. Calculate the area impact data in the comparison image. The area impact data = the sum of the area data of each obstacle under the number of obstacles. If the volume impact data is greater than or equal to the area impact data × 15%, then the placement of the item is considered reliable, and there are relatively few unsafe factors. If the volume impact data is less than 15% of the area impact data, the item placement is deemed unreliable, and an alarm is triggered.

5. The method for monitoring unsafe behaviors at construction sites using unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: Focus on monitoring abnormal areas, including: Obtain the monitoring interval duration of the target device; Obtain the device condition data for the target area and designate it as the target condition data; The device condition data includes both presence and absence conditions; If the target condition data is an existing condition, then the monitoring interval duration is cancelled; Select any target device to monitor the abnormal area; If the target condition data is non-existent, then calculate the adjustment interval duration: Adjustment interval duration = Monitoring interval duration × 20%; Adjust the monitoring interval of the target device to monitor the abnormal area. An alarm will be triggered if an item is removed from the abnormal area.

6. A system for implementing the unmanned aerial vehicle (UAV) monitoring method for unsafe behaviors at construction sites as described in claim 1, characterized in that: include: Data Acquisition Module: Collects relevant data on the stacking of items within the target area using drones; Analysis module: Analyzes areas within the target area where items are stacked to determine whether the stacking is normal; Judgment module: Analyzes abnormal item stacking situations to determine whether the item stacking serves as a prompt. Alarm module: It will promptly issue alarms for abnormal stacking of items without any prompts, as well as for the removal of items from abnormal areas, to prevent accidents.

Images