An intelligent supervision method and system for building construction

Abstract

The present invention discloses an intelligent construction supervision method and system. Construction areas that do not need to be monitored are removed. The length of time each construction machine category is used is displayed as a punctuation point on a circle of different radius, and a category detection network is used for judgment. When the category detection network is trained, a similarity comparison image is extracted. During real-time monitoring, the length of time each construction machine category is used is distinguished from the length of time the construction machine category in the similarity comparison image is used, so as to determine whether it is similar to the usual use of the construction machine. The technical effect of more accurately monitoring whether a construction machine fails by monitoring the driving path and residence time of the construction machine is achieved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular, to an intelligent supervision method and system for building construction. Background Art

[0002] Currently, building intelligent supervision is often adopted in building construction and management. Generally, it refers to using advanced technologies such as the Internet of Things, big data analysis, and artificial intelligence to improve the management efficiency and sustainability of the building industry. By deploying a sensor network and connected devices, data on the operating status of the building can be collected, including energy consumption, environmental conditions, facility maintenance requirements, etc. However, since automated or semi-automated construction machines are often used during building construction, the working status of the construction machines needs to be supervised. And for the construction machines used by workers, it is necessary to monitor whether there are behaviors such as abuse or theft of the machines, so as to supervise the construction machines. Summary of the Invention

[0003] The purpose of the present invention is to provide an intelligent supervision method and system for building construction to solve the above problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides an intelligent supervision method for building construction, including:

[0005] Obtain a construction map, multiple construction machine categories, and the positions of construction machines at corresponding multiple time points; the position of the construction machine represents the position where the construction machine moves in the construction map; the construction map represents a map of the area including the construction site;

[0006] Based on the construction machine categories, through a category detection network, determine the usage status of the construction machines to obtain the machine categories to be detected; the machine categories to be detected represent the categories of construction machines that need to be subject to fault detection;

[0007] Based on the machine categories to be detected and the positions of construction machines at multiple time points, through a position path network, predict faults to obtain the numbers of faulty machines;

[0008] Send a signal to stop the construction machine corresponding to the number of the faulty machine.

[0009] Optionally, the step of determining the usage status of the construction machines through a category detection network based on the construction machine categories to obtain the machine categories to be detected includes:

[0010] Obtain the total working time of a construction machine category in a day as the working time length of the construction machine; multiple construction machine categories correspond to multiple working time lengths of construction machines;

[0011] Divide 360 by the construction machine category to obtain the difference angle;

[0012] Use the center point of the construction map as the discriminant origin point;

[0013] According to the difference angle, fill in the construction machine categories in turn at the positions in the construction map that are the construction machine time length away from the discriminant origin point, and obtain a construction correlation image; the construction correlation image represents the tightness of the correlation when the construction machine is building a building;

[0014] Based on the construction correlation image, through the category detection network, obtain the machine category to be detected.

[0015] Among them, input the construction correlation image and the similarity comparison image into the category detection network to obtain a predicted construction category similarity value. If the predicted construction category similarity value is greater than the similarity threshold, use the construction machine category corresponding to the construction correlation image as the machine category to be detected.

[0016] Optionally, based on the machine category to be detected and the positions of the construction machines at multiple time points, through the position path network, determine the usage status of the construction machines to obtain the faulty machine numbers, including:

[0017] Mark the positions of the construction machines at the multiple time points on the construction map. If there is a repetition, subtract 1 from the gray value through one position to obtain a construction machine position map;

[0018] Obtain the building position; the building position represents the position of the building under construction;

[0019] Mark the building position in the construction machine position map to obtain a marked building image;

[0020] Draw a circle with the building position in the marked building image as the center and a fixed distance as the radius to obtain a building area; the building area represents the area where the construction machine position can stay for a long time;

[0021] Based on the marked building image containing the building area, through the position path network, determine the usage status of the construction machines to obtain an error discrimination value; the error discrimination value of 1 indicates that the path of the construction machine does not conform to the path of the construction machine in normal work; the error discrimination value of 0 indicates that the path of the construction machine conforms to the path of the construction machine in normal work;

[0022] Use the numbers corresponding to the construction machines with the error discrimination value of 1 as the machine numbers to be detected.

[0023] Optionally, the training method of the category detection network includes:

[0024] Obtain multiple training construction correlation images; the training construction correlation images represent multiple construction machine categories and the corresponding construction machine time lengths;

[0025] Obtain labeled data; the labeled data being 1 indicates normal construction, and the labeled data being 0 indicates an error during construction;

[0026] Input two training construction-related images into a category detection network, determine the similarity of the images, and obtain a construction category similarity value; (n - 1)n / 2 construction category similarity values are obtained corresponding to n training construction-related images;

[0027] Based on the construction category similarity value, backpropagate to train the category detection network to obtain a trained category detection network.

[0028] Optionally, the step of backpropagating to train the category detection network based on the construction category similarity value to obtain a trained category detection network includes:

[0029] Use the training construction-related images corresponding to the construction category similarity values greater than the similarity threshold as a training set;

[0030] Label the training construction-related images in the training set as similar to obtain labeled data;

[0031] Calculate the average of the construction machine time lengths corresponding to the same angle of the training construction-related images in the training set to obtain a similarity comparison image; the similarity comparison image represents a benchmark for similarity judgment;

[0032] Input the training construction-related images and the similarity comparison image into the category detection network to obtain a training similarity value;

[0033] Calculate the loss between the training similarity value and the representation data, and backpropagate to train the category detection network to obtain a trained category detection network.

[0034] Optionally, the step of determining the usage status of the construction machine through a position path network based on the marked building image including the construction area to obtain an error determination value includes:

[0035] Obtain a road image; the road image represents an image that marks the road on which the construction machine can travel in the construction image;

[0036] Based on the marked building image including the construction area, remove the construction area part to obtain a driving path image;

[0037] Input the driving path image and the road image into the position path network to judge the driving path of the construction machine and obtain a path similarity value; the path similarity value represents whether the path traveled by the construction machine exits the position of the path in the road image;

[0038] Based on the driving path image, the length of the stay time is determined to obtain a stay time similarity value; the stay time similarity value indicates whether the stay time of the construction machine at different locations complies with regulations;

[0039] Based on the path similarity value and the stay time similarity value, an error determination value is obtained.

[0040] Optionally, judging the length of stay time based on the driving path image to obtain a stay time similarity value includes:

[0041] Obtaining a dwell time value; the dwell time value represents a fixed dwell time length;

[0042] Converting the grayscale value in the driving path image that is equal to the length of the stay time into the grayscale value of the length of the stay time minus 1 to obtain a long-term stay position for construction;

[0043] Obtaining the number of fixed positions; the number of fixed positions indicates limiting the number of times the construction machine stays for a long time;

[0044] The construction long-term stay position is divided by the number of fixed positions to obtain the stay time similarity value.

[0045] Optionally, obtaining an error discrimination value based on the path similarity value and the stay time similarity value includes:

[0046] If the path similarity value is less than or equal to the path threshold, and the stay time similarity value is less than or equal to the stay threshold, the error discrimination value is set to 1;

[0047] If the path similarity value is greater than the path threshold, or the stay time similarity value is greater than the stay threshold, the error judgment value is set to 0.

[0048] Optionally, the removing the building area portion based on the marked building image including the building area to obtain the driving path image includes:

[0049] Converting the labeled building image containing the building area into a binary image to obtain a binary construction machine path image;

[0050] The value of the building area in the binary construction machine path image is set to 255 to obtain a driving path image; the driving path image represents an image without the influence of the construction machine staying in the construction area.

[0051] In a second aspect, an embodiment of the present invention provides a construction intelligent supervision system, including:

[0052] An acquisition module for obtaining a construction map, multiple construction machine categories, and the positions of construction machines at corresponding multiple time points; the positions of the construction machines represent the positions where the construction machines move in the construction map; the construction map represents a map of an area containing construction.

[0053] A category detection module for determining the usage status of construction machines based on the construction machine categories through a category detection network to obtain the machine categories to be detected; the machine categories to be detected represent the categories of construction machines that need to be subjected to fault detection.

[0054] A fault detection module for predicting faults based on the machine categories to be detected and the positions of construction machines at multiple time points through a position path network to obtain the numbers of faulty machines.

[0055] A signal sending module for sending a signal to stop the construction machine corresponding to the number of the faulty machine.

[0056] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0057] The embodiments of the present invention also provide a method and a system for intelligent supervision of building construction.

[0058] In the present invention, since the construction machines stay in the building area, it is convenient for construction, or the construction machines are being used for construction, so the building area does not need to be monitored. The different lengths of time used by each construction machine category outside the building area indicate that the use of the construction machines is different from the usual efficient use method. The length of time used by each construction machine category is displayed by a punctuation mark on a circle with a different radius, and a category detection network is used for judgment. When the category detection network is trained, a similar comparison image is extracted. During real-time monitoring, the length of time used by each construction machine category is compared with the length of time used by the construction machine category in the similar comparison image to determine whether it is similar to the usual use method of the construction machines. The driving paths of the construction machines excluding the building area are used to judge whether they are different from the set driving paths. A two-dimensional grayscale image is used to jointly represent the staying positions of the construction machines and the corresponding lengths of staying time, and the one-dimensional data representing the lengths of staying time of the construction machines is converted into grayscale values on the two-dimensional image. The technical effect of more accurately monitoring whether the construction machines have faults by monitoring the driving paths and staying times of the construction machines is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flowchart of a method for intelligent supervision of building construction provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention will be described in detail below with reference to the accompanying drawings.

[0061] Example 1

[0062] As Figure 1 shown, an embodiment of the present invention provides an intelligent supervision method for building construction, and the method includes:

[0063] S101: Obtain a construction map, multiple construction machine categories, and the positions of construction machines at corresponding multiple time points; the position of the construction machine represents the position where the construction machine moves in the construction map; the construction map represents a map of the area containing the construction.

[0064] Among them, the position of the construction machine is recorded by a positioning device fixed on the construction machine.

[0065] Among them, the construction machine category represents the category to which the construction machine belongs. For example, excavators and mixers belong to two different construction machine categories.

[0066] Among them, the construction map is a grayscale image, and the initial grayscale value of the construction map is 255. The grayscale value of the grayscale image indicates that the grayscale image is initially white.

[0067] Among them, in this embodiment, the length and width of the construction map are the same, which is 512 * 512.

[0068] S102: Based on the construction machine category, through a category detection network, determine the usage status of the construction machine to obtain a machine category to be detected; the machine category to be detected represents the category of the construction machine that needs to be fault-detected;

[0069] S103: Based on the machine category to be detected and the positions of the construction machines at multiple time points, through a position path network, predict faults to obtain a faulty machine number;

[0070] S104: Send a signal to stop the construction machine corresponding to the faulty machine number.

[0071] Among them, if reuse is required, manual approval is needed.

[0072] Optionally, the predicting faults based on the construction machine category through a category detection network to obtain a machine category to be detected includes:

[0073] Obtain all the working hours of a construction machine category in a day as the construction machine time length; multiple construction machine categories correspondingly obtain multiple construction machine time lengths.

[0074] Among them, the unit of the construction machine time length is minutes.

[0075] Divide 360 by the construction machine category to obtain the difference angle.

[0076] Use the center point of the construction map as the discriminant origin point.

[0077] According to the difference angle, sequentially fill the construction machine categories into the positions on the construction map that are at a distance from the discriminant origin point equal to the construction machine time length, to obtain a construction correlation image; the construction correlation image represents the tightness of the correlation when the construction machine is building a building.

[0078] Among them, divide the construction machine time length by (24 * 60) and multiply by the width of the grayscale image to obtain the construction machine radius.

[0079] Among them, use the lower left corner of the grayscale image as the origin, use the length of the grayscale image as the abscissa, and use the width of the grayscale image as the ordinate.

[0080] Among them, as in this embodiment, the construction machine category is 36, 360 / 36 = 10. Take the direction parallel to the abscissa and with the abscissa value increasing from small to large as 0 degrees. At the 0-degree direction of one construction machine category, find the position where the clustering discriminant center is the construction machine radius and mark it. Rotate 10 degrees clockwise, find the position where the clustering discriminant center is the construction machine radius corresponding to another construction machine category and mark it.

[0081] Among them, the construction machine category corresponding to each angle is fixed.

[0082] Based on the construction correlation image, through the category detection network, obtain the machine category to be detected.

[0083] Optionally, based on the machine category to be detected and the construction machine positions at multiple time points, through the position path network, determine the usage status of the construction machine to obtain the faulty machine number, including:

[0084] Mark the construction machine positions at the multiple time points on the construction map. If a position is repeated 5 times, subtract 1 from the grayscale value to obtain the construction machine position map.

[0085] Among them, if a position is repeated 5 times, subtract 1 from the grayscale value until the grayscale value equals 0 and then stop.

[0086] Among them, because (256 * 5) > (24 * 60), add 1 to the grayscale value only after 5 times.

[0087] Obtain the building position; the building position represents the position of the building under construction.

[0088] Mark the building position on the construction machine position map to obtain the marked building image.

[0089] Taking the building position in the marked building image as the center of a circle and drawing a circle with a fixed distance as the radius to obtain a building area; the building area represents the area where the construction machine position can stay for a long time.

[0090] Among them, in this embodiment, the fixed distance is 5, with the unit of pixel value, corresponding to 512*512 of the construction map.

[0091] Based on the marked building image containing the building area, through the position path network, the usage status of the construction machine is discriminated to obtain an error discrimination value; the error discrimination value of 1 indicates that the path of the construction machine does not conform to the path of the construction machine in normal work; the error discrimination value of 0 indicates that the path of the construction machine conforms to the path of the construction machine in normal work.

[0092] Taking the number corresponding to the construction machine with the error discrimination value of 1 as the machine number to be detected.

[0093] Optionally, the training method of the category detection network includes:

[0094] Obtaining a plurality of training construction-related images; the training construction-related images represent multiple construction machine categories and the corresponding construction machine time lengths;

[0095] Obtaining annotation data; the annotation data of 1 indicates normal construction, and the annotation data of 0 indicates an error in construction.

[0096] Among them, normal construction means constructing according to the construction machine of a fixed used construction machine category, and an error in construction means randomly using a construction machine.

[0097] Inputting two training construction-related images into the category detection network to judge the similarity of the images and obtain a construction category similarity value; (n - 1)n / 2 construction category similarity values are obtained corresponding to n training construction-related images.

[0098] Among them, n is a natural number greater than 0.

[0099] Among them, the similarity recognition network is the discriminator (Discriminator) of the generative adversarial network (GAN).

[0100] Based on the construction category similarity value, the category detection network is trained by backpropagation to obtain a trained category detection network.

[0101] Optionally, the training the category detection network by backpropagation based on the construction category similarity value to obtain a trained category detection network includes:

[0102] Taking the training construction-related images corresponding to the construction category similarity values greater than the similarity threshold as the training set.

[0103] Among them, in this embodiment, the similarity threshold is 0.8.

[0104] Label the training construction-related images in the training set as similar, and label the training construction-related images that are not in the training set as dissimilar to obtain labeled data.

[0105] Among them, in this embodiment, the labeled data labeled as similar is 1, and the labeled data labeled as dissimilar is 0.

[0106] Calculate the average of the construction machine time lengths corresponding to the same angle of the training construction-related images in the training set to obtain a similarity comparison image; the similarity comparison image represents a benchmark for similarity judgment.

[0107] Input the training construction-related image and the similarity comparison image into the class detection network to obtain a training similarity value.

[0108] Among them, the class detection network is used twice. The first time is equivalent to clustering, and the second time is to obtain a network that can discriminate the similarity comparison image after clustering.

[0109] Calculate the loss between the training similarity value and the representation data, and backpropagate to train the class detection network to obtain a trained class detection network.

[0110] Optionally, based on the marked building image including the building area, use the position path network to discriminate the usage status of the construction machine to obtain an error discrimination value, including:

[0111] Obtain a road image; the road image represents an image that marks the road on which the construction machine in the construction image can travel.

[0112] Based on the marked building image including the building area, remove the building area part to obtain a driving path image.

[0113] Among them, removing the building area part means that the duration of the construction machine staying in the building area is not restricted.

[0114] Input the driving path image and the road image into the position path network to judge the driving path of the construction machine to obtain a path similarity value; the path similarity value represents whether the path of the construction machine driving out of the path position in the road image.

[0115] Among them, in this embodiment, the position path network is the structure of the discriminator (Discriminator) of the generative adversarial network (GAN).

[0116] Based on the driving path image, determine the length of the stay time to obtain a stay time similarity value; the stay time similarity value indicates whether the stay time of the construction machine at different positions complies with the regulations.

[0117] Based on the path similarity value and the stay time similarity value, obtain an error discrimination value.

[0118] Through the above method, if the driving path is different from the set path, and the stay position and time length are different from usual, it indicates that someone has stolen and misused the driving machine.

[0119] Optionally, the determining the length of the stay time based on the driving path image to obtain a stay time similarity value includes:

[0120] Obtain a stay time value; the stay time value represents a fixed length of the stay time.

[0121] Among them, in this embodiment, the time stay value is set to 2.

[0122] Convert the gray value equal to the stay time length value in the gray values of the driving path image into a gray value of the stay time length minus 1 to obtain the long-time stay position of the construction;

[0123] Obtain the number of fixed positions; the number of fixed positions represents the number of times the construction machine is restricted from staying for a long time;

[0124] Divide the long-time stay position of the construction by the number of fixed positions to obtain a stay time similarity value.

[0125] Optionally, the obtaining an error discrimination value based on the path similarity value and the stay time similarity value includes:

[0126] If the path similarity value is less than or equal to the path threshold, and the stay time similarity value is less than or equal to the stay threshold, set the error discrimination value to 1.

[0127] Among them, the stay threshold is 0.85.

[0128] If the path similarity value is greater than the path threshold, or the stay time similarity value is greater than the stay threshold, set the error discrimination value to 0.

[0129] Among them, the path threshold is 0.9.

[0130] Optionally, the removing the building area part from the marked building image including the building area to obtain a driving path image includes:

[0131] Convert the marked building image including the building area into a binary image to obtain a binary construction machine path image;

[0132] Set the value of the building area in the binary construction machine path image to 255 to obtain a driving path image; the driving path image represents an image that removes the influence of the construction machine staying in the building area.

[0133] Embodiment 2

[0134] Based on the above building construction intelligent supervision method, an embodiment of the present invention further provides a building construction intelligent supervision system, which includes an acquisition module, a category detection module, a fault detection module, and a signal sending module.

[0135] The acquisition module is used to obtain a construction map, multiple construction machine categories, and the positions of construction machines at corresponding multiple time points; the position of the construction machine represents the position where the construction machine moves in the construction map; the construction map represents a map including the area under construction.

[0136] The category detection module is used to determine the usage status of the construction machine based on the construction machine category through a category detection network to obtain a machine category to be detected; the machine category to be detected represents the category of the construction machine that needs to be subjected to fault detection.

[0137] The fault detection module is used to predict a fault based on the machine category to be detected and the positions of the construction machines at multiple time points through a position path network to obtain a faulty machine number.

[0138] The signal sending module is used to send a signal to stop the construction machine corresponding to the faulty machine number.

Claims

1. A method for intelligent supervision of building construction, characterized in that: include: A construction map, a plurality of construction machine categories and corresponding construction machine positions at a plurality of time points are obtained; the construction machine positions represent positions where the construction machines move in the construction map; the construction map represents a map of an area including construction; Based on the construction machine category, the use status of the construction machine is determined through a category detection network to obtain the category of the machine to be detected; the category of the machine to be detected represents the category of the construction machine that needs to be fault detected; based on the category of the machine to be detected and the location of the construction machine at multiple time points, the fault is predicted through a location path network to obtain the fault machine number; Sending a signal to stop the construction machine corresponding to the faulty machine number; The method of predicting a fault based on the type of the machine to be detected and the location of the construction machine at multiple time points and obtaining the faulty machine number through a location path network includes: Marking the positions of the construction machines at the multiple time points on the construction map, and if a position is repeatedly passed, subtracting 1 from the grayscale value corresponding to the position in the construction map to obtain a construction machine position map; Acquire a building location; the building location indicates a location of a building under construction; Marking the building location in the construction machine location map to obtain a marked building image; Taking the building position in the marked building image as the center of the circle and drawing a circle with a fixed distance as the radius, a building area is obtained; the building area indicates an area where the construction machine can stay for a long time; Acquire a road image; the road image represents an image marking a road in the construction image on which the construction machine can travel; Based on the marked building image including the building area, the building area portion is removed to obtain a driving path image; the driving path image and the road image are input into the position path network to determine the driving path of the construction machine and obtain a path similarity value; the path similarity value indicates whether the driving path of the construction machine has gone out of the position of the path in the road image; Subtract 1 from the grayscale value corresponding to the position where the construction machine stays at each position in the driving path image for a time equal to the preset fixed stay time length value, to obtain the construction long-term stay position, and then count the number of the construction long-term stay positions; Obtain the number of fixed positions; the number of fixed positions indicates the number of times the construction machine is restricted to stay for a long time; divide the number of construction long-term stay positions by the number of fixed positions to obtain a stay time similarity value; the stay time similarity value indicates whether the stay time of the construction machine at different positions complies with regulations; based on the path similarity value and the stay time similarity value, obtain an error judgment value, the error judgment value of 1 indicates that the path of the construction machine does not comply with the path of the construction machine at normal work; the error judgment value of 0 indicates that the path of the construction machine complies with the path of the construction machine at normal work; The number corresponding to the construction machine with the error judgment value of 1 is used as the number of the machine to be detected.

2. The intelligent construction supervision method according to claim 1 is characterized in that: Based on the category of the construction machine, the use status of the construction machine is determined through a category detection network to obtain the category of the machine to be detected, including: The total working time of a construction machine category in a day is obtained as the construction machine time length; multiple construction machine categories are correspondingly obtained with multiple construction machine time lengths; Divide 360 ​​by the number of construction machine categories to get the gap angle; The center point of the construction map is used as the judgment point; According to the gap angle, the construction machine categories are sequentially filled into the positions of the distance determination dots in the construction map as the time length of the construction machines, and a construction association image is obtained; the construction association image represents the closeness of the association of the construction machines when constructing the building; Based on the construction-related image, the category of the machine to be detected is obtained through a category detection network.

3. The intelligent construction supervision method according to claim 2 is characterized in that: The training method of the category detection network comprises: Acquire a plurality of training construction-related images; the training construction-related images represent a plurality of construction machine categories and corresponding construction machine time lengths; Acquire annotation data; if the annotation data is 1, it indicates that the construction is normal, and if the annotation data is 0, it indicates that an error occurs during the construction; Input two training construction-related images into the category detection network, judge the similarity of the images, and obtain the construction category similarity value; n training construction-related images correspond to (n-1)n / 2 construction category similarity values; Based on the construction category similarity values, the category detection network is trained by back propagation to obtain a trained category detection network.

4. The intelligent construction supervision method according to claim 3 is characterized in that: The back-propagation training category detection network based on the construction category similarity value to obtain the trained category detection network includes: taking the training construction-related images corresponding to the construction category similarity values ​​greater than the similarity threshold as the training set; Annotating the training construction-related images in the training set as similar to obtain annotated data; The construction machine time lengths corresponding to the same angle of the corresponding training construction-related images in the training set are averaged to obtain a similarity comparison image; the similarity comparison image represents a benchmark for similarity judgment; the training construction-related image and the similarity comparison image are input into a category detection network to obtain a training similarity value; the training similarity value and the labeled data are used to calculate the loss, and the training category detection network is back-propagated to obtain a trained category detection network.

5. The intelligent construction supervision method according to claim 1 is characterized in that: The step of obtaining an error discrimination value based on the path similarity value and the residence time similarity value includes: If the path similarity value is less than or equal to the path threshold, and the stay time similarity value is less than or equal to the stay threshold, the error discrimination value is set to 1; If the path similarity value is greater than the path threshold, or the stay time similarity value is greater than the stay threshold, the error judgment value is set to 0.

6. The intelligent construction supervision method according to claim 1 is characterized in that: The method of removing the building area portion based on the marked building image including the building area to obtain a driving path image includes: converting the marked building image including the building area into a binary image to obtain a binary construction machine path image; setting the value of the building area in the binary construction machine path image to 255 to obtain a driving path image; the driving path image represents an image without the impact of the construction machine staying in the building area.

7. An intelligent construction supervision system, characterized in that: include: An acquisition module, used to obtain a construction map, a plurality of construction machine categories and corresponding construction machine positions at a plurality of time points; the construction machine positions represent positions where the construction machines move in the construction map; the construction map represents a map of an area including construction; A category detection module, used to determine the use status of the construction machine based on the construction machine category through a category detection network to obtain the category of the machine to be detected; the category of the machine to be detected represents the category of the construction machine that needs to be fault detected; A fault detection module is used to predict faults and obtain fault machine numbers through a location path network based on the type of machine to be detected and the location of the construction machine at multiple time points; A signal sending module, used for sending a signal to stop the construction machine corresponding to the faulty machine number; The fault detection module comprises: Marking the positions of the construction machines at the multiple time points on the construction map, and if a position is repeatedly passed, subtracting 1 from the grayscale value corresponding to the position in the construction map to obtain a construction machine position map; Acquire a building location; the building location indicates a location of a building under construction; Marking the building location in the construction machine location map to obtain a marked building image; using the building location in the marked building image as the center of a circle and a fixed distance as the radius to draw a circle to obtain a building area; the building area represents an area where the construction machine can stay for a long time; Acquire a road image; the road image represents an image marking a road in the construction image on which the construction machine can travel; Based on the marked building image including the building area, the building area portion is removed to obtain a driving path image; the driving path image and the road image are input into the position path network to determine the driving path of the construction machine and obtain a path similarity value; the path similarity value indicates whether the driving path of the construction machine has gone out of the position of the path in the road image; Subtract 1 from the grayscale value corresponding to the position where the construction machine stays at each position in the driving path image for a time equal to the preset fixed stay time length value, to obtain the construction long-term stay position, and then count the number of the construction long-term stay positions; Obtain the number of fixed positions; the number of fixed positions indicates the number of times the construction machine is restricted to stay for a long time; divide the number of construction long-term stay positions by the number of fixed positions to obtain a stay time similarity value; the stay time similarity value indicates whether the stay time of the construction machine at different positions complies with regulations; Based on the path similarity value and the stay time similarity value, an error discrimination value is obtained, wherein the error discrimination value of 1 indicates that the path of the construction machine does not conform to the path of the construction machine in normal operation; and the error discrimination value of 0 indicates that the path of the construction machine conforms to the path of the construction machine in normal operation; The number corresponding to the construction machine with the error judgment value of 1 is used as the number of the machine to be detected.

Images