A method for earthwork construction management

Abstract

The present invention provides a method for earthwork construction management, comprising the following steps: S1, collecting transport vehicle information and identifying the volume of the transport vehicle's compartment; S2, counting the number of bucket loading times during earthwork loading; S3, collecting the compartment image after loading is completed and judging the full load of the compartment; S4, calculating the actual loading amount according to the number of bucket loading times, comparing the actual loading amount with the full load, and issuing an abnormal report for early warning if the vehicle exceeds a set threshold. Beneficial effects of the present invention: abnormal loading conditions of transport vehicles can be discovered in a timely manner, and abnormal reports can be issued, thereby ensuring the accuracy of transportation fee calculation.

Description

Technical Field

[0001] The invention belongs to the technical field of building construction, and in particular relates to an earthwork construction management method. Background Art

[0002] With the rapid development of the national economy, a large amount of construction waste will inevitably be generated in the process of vigorously developing infrastructure. The transportation of construction waste has the characteristics of short construction period, difficult measurement, and high monitoring cost. A large number of earthwork project managers cannot calculate the cost at all during the project process. They will not know whether the entire project is profitable until the end of the project. If problems occur in the middle of the process, they often lose the opportunity to adjust.

[0003] During the transportation process, the participants are the drivers of the muck trucks and the people who fill the muck. Although management personnel can be set up, the problems caused by human management are also obvious. Especially in large projects, whether the muck trucks are fully loaded or not directly involves huge transportation costs, and the accuracy of the counting cannot be guaranteed, which will also cause inaccurate transportation costs. Summary of the invention

[0004] In view of this, the present invention aims to propose an earthwork construction management method in order to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A first aspect of the present invention provides an earthwork construction management method, comprising the following steps:

[0007] S1. Collecting transport vehicle information, identifying the state of the transport vehicle's compartment, and identifying the volume of the transport vehicle's compartment;

[0008] S2. Counting the number of bucket loadings and the amount of each bucket loading during the earthwork loading process;

[0009] S3, collecting the image of the carriage after loading, and judging the fullness of the carriage;

[0010] S4. Calculate the actual load according to the number of bucket loadings and the load, compare the actual load with the full load, and issue an abnormal report for early warning if the vehicle exceeds the set threshold.

[0011] Furthermore, the S1 comprises the following steps:

[0012] The transport vehicle is provided with an information tag, and the excavator is provided with an information collector. The information tag on the transport vehicle is collected by the information collector. The information tag corresponds to the vehicle information in the database. The vehicle information includes the model of the transport vehicle and the vehicle box volume corresponding to the model.

[0013] Furthermore, the S2 specifically includes the following steps:

[0014] S21, calculating the theoretical bucket loading times according to the vehicle box volume;

[0015] S22, collecting video data during the bucket loading process, and counting the bucket loading actions according to the video data to obtain the actual number of bucket loadings;

[0016] S23, comparing the theoretical bucket loading times and the actual bucket loading times;

[0017] If the error exceeds the set threshold, an exception report is issued.

[0018] Further, the S22 comprises the following steps:

[0019] S221, collecting video data during bucket loading;

[0020] S222, identifying the loaded bucket state picture, the unloaded bucket state picture, and the empty bucket state picture in the video data as a bucket loading action, and counting the number of bucket loading actions after visually observing that the loading is completed as the number of bucket loading times.

[0021] Furthermore, the training process of the vehicle state recognition model is as follows:

[0022] B1. Collect historical carriage photos and add a label to each historical carriage photo. The label is the loading rate. A loading rate within the first interval is considered full, a loading rate within the second interval is considered abnormal, and a loading rate within the third interval is considered abnormal;

[0023] The first interval is 90%-100%;

[0024] The second interval is 5%-90%;

[0025] The third interval is 0%-5%;

[0026] The unloaded label also includes vehicle information such as the vehicle model;

[0027] B2. Preprocessing of noise reduction and normalization of historical carriage photos;

[0028] B3, extract features of historical carriage photos;

[0029] B4. Use the features extracted in B3 to train the neural network model to generate a vehicle state recognition model.

[0030] Further, the S3 comprises the following steps:

[0031] S31, the vehicle box image after acquisition;

[0032] S32, inputting the completed carriage image into the carriage state recognition model;

[0033] S33, the carriage state recognition model outputs the carriage state of the current carriage, and the carriage state includes empty, fully loaded, and abnormal;

[0034] If the vehicle box is fully loaded, proceed to step S4;

[0035] If it is empty or abnormal, an abnormal report will be issued to alert the excavator operator.

[0036] Furthermore, in S4, the theoretical load capacity is calculated based on the full load rate and the compartment volume identified by the compartment state identification model, and the theoretical load capacity is compared with the actual load capacity. If the error exceeds a set threshold, an abnormality report is issued.

[0037] Furthermore, the identification of the vehicle box state of the transport vehicle in S1 is as follows: before the bucket is loaded, an image of the vehicle box before loading is collected, and whether there is any residual material in the vehicle box is identified based on the image of the vehicle box before loading, and if there is no residual material, the process proceeds to step S2.

[0038] Furthermore, the specific steps of identifying whether there is residual material in the carriage according to the carriage image before loading are as follows:

[0039] A1. Collect the image of the carriage before loading;

[0040] A2, inputting the image of the carriage before loading into the carriage state recognition model;

[0041] A3. The carriage state recognition model outputs the current carriage state, which includes empty, fully loaded, and abnormal.

[0042] If the carriage is empty, it is determined that there is no remaining material and the process proceeds to step S2 for loading;

[0043] If the box is fully loaded, an abnormal report will be issued to alert the excavator operator;

[0044] If the carriage state is abnormal, and the error in step S23 exceeds the set threshold, it is determined that the carriage is abnormal and the transport vehicle operator is reminded to check the carriage state;

[0045] The vehicle state recognition model outputs the actual vehicle information of the current vehicle. The actual vehicle information is compared with the vehicle information collected by S1. If there is any inconsistency, an abnormality report is issued.

[0046] A second aspect of the present invention provides an earthwork construction management system, comprising:

[0047] An information collection module configured to collect transport vehicle information through information tags and GPS;

[0048] A camera configured to collect bucket loading video and image information before and after the vehicle box is loaded;

[0049] A data processing module is configured to perform the following steps:

[0050] Identify the volume of the transport vehicle's box, count the number of bucket loadings during earthwork loading based on the bucket loading video, and determine the fullness of the box;

[0051] Calculate the actual load based on the number of bucket loading times, compare the actual load with the full load, and issue an abnormal report as an early warning if the vehicle exceeds the set threshold;

[0052] A display input module, which is configured to display the transport vehicle information and the processing results of the data processing module, and send an order opening instruction to the management platform;

[0053] The management platform is configured to record the transport vehicle list, excavator list, data collected by cameras, generate waybill data according to order opening instructions, and record historical waybill data.

[0054] Furthermore, the transport vehicle is provided with an information tag, and the excavator is provided with an information collector, and the information tag on the transport vehicle is collected by the information collector, and the information tag corresponds to the vehicle information in the database, and the vehicle information includes the model of the transport vehicle and the vehicle box volume corresponding to the model;

[0055] The information tag is an RFID tag or a Bluetooth tag.

[0056] The information collector is an RFID signal collector or a tablet or smart phone.

[0057] Furthermore, the display input module is a tablet. Clicking the start task transmits the order opening instruction to the background, and the database of the management platform creates the waybill data. The waybill data includes the vehicle information consisting of the waybill number and the vehicle number obtained through the camera;

[0058] After clicking Start Task, the camera starts collecting image information of the vehicle box;

[0059] After the task is completed, the camera is turned off and the image information of the vehicle compartment is stopped;

[0060] The exception report sent by the data processing module is displayed and sent to the management platform, which stores the exception report.

[0061] Furthermore, the instruction to end the task can be obtained by the following method:

[0062] Manually select the End Task option on the tablet;

[0063] Automatically identify the end of the task through the excavator action.

[0064] A third aspect of the present invention provides an electronic device, comprising a processor and a memory that is communicatively connected to the processor and is used to store instructions executable by the processor, wherein the processor is used to execute the method described in the first aspect above.

[0065] The fourth aspect of the present invention provides a server, comprising at least one processor, and a memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the processor so that the at least one processor executes the method described in the first aspect.

[0066] Compared with the prior art, the earthwork construction management method described in the present invention has the following beneficial effects:

[0067] (1) The earthwork construction management method described in the present invention can timely discover abnormal loading conditions of transport vehicles and issue abnormality reports, thereby ensuring the accuracy of transportation fee calculation.

[0068] (2) The earthwork construction management method described in the present invention includes setting an information tag on the transport vehicle and setting an information collector on the excavator to collect and identify the vehicle information of the vehicle. The management platform records the waybill information in the vehicle information, including the waybill number, construction site name, license plate, excavator model, loading time, etc., which is used to track the remaining material status, full load status, total number of buckets, number of times the bucket is not full, etc., so as to monitor the transportation efficiency. The vehicle box volume can also be identified for subsequent verification. Multiple detection means prevent transportation personnel from misusing vehicle information, thereby ensuring the accuracy of the calculation of earthwork transportation costs.

[0069] (3) In the earthwork construction management method described in the present invention, the vehicle loading capacity recognition model can quickly identify whether the bucket is full. Manual counting may cause misjudgment due to the empty bucket loading action. Compared with manual counting, the counting accuracy is improved.

[0070] (4) The earthwork construction management method described in the present invention can avoid inaccurate calculation of the loading volume caused by half-loaded vehicle or bottom sludge by comparing the theoretical bucket loading times with the actual bucket loading times, and can also avoid inaccurate loading volume caused by negligence of personnel, thereby further improving the accuracy of earthwork transportation cost calculation.

[0071] (5) The earthwork construction management method described in the present invention ultimately determines the loading rate of the vehicle box through a vehicle box loading capacity recognition model, thereby avoiding the subjective bias of manual judgment of the loading rate and further improving the accuracy of earthwork transportation cost calculation.

[0072] (6) The earthwork construction management method described in the present invention can automatically generate a large amount of data by generating a three-dimensional model, and has a wider range of applications. The vehicle model can be obtained by license plate recognition or card swiping, and then the corresponding three-dimensional model is matched, and the corresponding visual recognition model is called to estimate the load. Historical data requires a large amount of historical data, which improves the automation and accuracy of load recognition.

[0073] (7) The earthwork construction management method described in the present invention uses a tablet to turn on or off the image data collection of the camera, thereby avoiding the ineffective working time of the camera.

[0074] (8) The earthwork construction management method described in the present invention uses image recognition technology to identify the vehicle's license plate number, queries the license plate number in the database and compares it with the Bluetooth information, installs GPS positioning on the vehicle, uses a positioning device to locate the vehicle's position, and performs Bluetooth, video, and position triple verification, which can achieve a 99% accuracy rate in identifying engineering operation vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0076] Figure 1 A schematic diagram of the process flow of the earthwork construction management method according to an embodiment of the present invention;

[0077] Figure 2 This is a schematic diagram of the work flow of the earthwork construction management system according to an embodiment of the present invention;

[0078] Figure 3 This is a schematic diagram of the residual material identification process according to an embodiment of the present invention;

[0079] Figure 4 It is a schematic diagram of a flow chart of counting the number of bucket loading times in a process of earthwork loading according to an embodiment of the present invention;

[0080] Figure 5 It is a schematic diagram of a process of collecting a carriage image after loading to determine the full load status of the carriage according to an embodiment of the present invention;

[0081] Figure 6 A schematic diagram of the construction process of the visual recognition model according to an embodiment of the present invention;

[0082] Figure 7 A schematic diagram of the training process of the vehicle state recognition model according to an embodiment of the present invention;

[0083] Figure 8It is a schematic flow chart of the earthwork construction management method with surplus material identification process described in an embodiment of the present invention. DETAILED DESCRIPTION

[0084] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0085] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0086] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0087] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0088] Embodiment 1:

[0089] like Figures 1 to 8 As shown, a method for earthwork construction management comprises the following steps:

[0090] S1. Collecting transport vehicle information and identifying the volume of the transport vehicle's compartment;

[0091] S1 includes the following steps:

[0092] The transport vehicle is provided with an information tag, and the excavator is provided with an information collector. The information tag on the transport vehicle is collected by the information collector. The information tag corresponds to the vehicle information in the database. The vehicle information includes the model of the transport vehicle and the vehicle box volume corresponding to the model.

[0093] In some embodiments, the information tag is an RFID tag, and the information collector is an RFID signal collector;

[0094] In other embodiments, the information tag is a Bluetooth speaker tag, an information collector tablet, a smart phone or other independently developed equipment.

[0095] In this embodiment, a Bluetooth external tag is used. The unique device number of the Bluetooth external tag is bound to the vehicle. The tablet collects signals through the Bluetooth function, pulls the vehicle tag dictionary through server communication, and maps the tag to the license plate. The vehicle only needs to drive into the work area normally. The specific implementation process is: (1) The Bluetooth retrieval interface performs continuous signal retrieval. (2) Entry rules: When a new signal is retrieved, the license plate mapping is immediately fed back to the background and written into the record list. (3) Departure rules: A specified period is agreed upon, and all retrieved signals are included in the polling list. At the end of the period, the polling list is compared with the record list, and the overlap is retained. If it does not appear in the polling list, the record list is removed. The Bluetooth module sets a sensing and recognition range of 15-20M, which generally corresponds to at least one working vehicle and one waiting vehicle.

[0096] Vehicle information includes:

[0097] Party A’s company information, earthwork boss information: company name, project name (multiple projects are carried out simultaneously), project location, administrator name.

[0098] Information of the transport company, the name of the transport company, the legal person, the company's business license, road transport license, qualification certificate, all information of the vehicle (such as volume), whether the vehicle's disposal certificate is compliant, whether the insurance is legal, driver's identity information, driver's operating qualification certificate, telephone number, etc.

[0099] Excavator company information, company name, legal person, business license, excavator factory information, excavator exit certificate, environmental protection certificate, insurance information, driver identity information, driver's operating qualification certificate, telephone number, etc.

[0100] The camera follows the excavator.

[0101] The license plate list includes Bluetooth tag information, license plate list, and vehicle details. The tag information and license plate information are bound and have a corresponding relationship. The process of having a corresponding relationship on the tablet: the tablet scans, obtains the vehicle's Bluetooth signal, matches it with the server background according to the mapping relationship, obtains the vehicle's entry data, informs the vehicle to enter the site, and updates the background information list. The management platform obtains the two vehicle lists before and after, compares them, and after comparison, it can determine the vehicle leaving the site.

[0102] Waybill information: waybill number, construction site name, license plate, excavator model-excavator name, remaining material status, full load status, total number of buckets, number of incomplete buckets, waybill status, pictures, loading time.

[0103] In addition, the vehicle's license plate number is identified through image recognition technology, the license plate number query database is compared with the Bluetooth information, the vehicle is installed with GPS positioning, and the vehicle's position is located using a positioning device. Bluetooth, video, and location triple verification can achieve a 99% accuracy rate in identifying engineering operation vehicles.

[0104] By setting information tags on transport vehicles and information collectors on excavators, the vehicle information of the vehicles is collected and identified. The management platform records the waybill information in the vehicle information, including waybill number, construction site name, license plate, excavator model, loading time, etc. It is used to track the remaining material status, full load status, total number of buckets, number of incomplete buckets, etc., to monitor transportation efficiency. It can also identify the vehicle box volume for subsequent verification. Multiple detection methods prevent transportation personnel from impersonating vehicle information, ensuring the accuracy of earthwork transportation cost calculation.

[0105] S2. Count the number of bucket loadings during earthwork loading; the bucket is the bucket of an excavator.

[0106] Before the bucket is loaded, the image of the vehicle box before loading is collected, and the vehicle box status and vehicle information are identified based on the image of the vehicle box before loading. The specific steps are as follows:

[0107] A1. Collect the image of the carriage before loading;

[0108] A2. Inputting the image of the vehicle box before loading into the vehicle box loading capacity recognition model;

[0109] A3. The vehicle loading capacity recognition model outputs the current vehicle status, which includes empty, fully loaded, and abnormal.

[0110] If the vehicle box is empty, proceed to step S2 for loading;

[0111] If the box is fully loaded, an abnormal report will be issued to alert the excavator operator;

[0112] If the carriage state is abnormal, and the error in step S23 exceeds the set threshold, it is determined that the carriage is abnormal and the transport vehicle operator is reminded to check the carriage state;

[0113] The vehicle loading capacity recognition model outputs the actual vehicle information of the current vehicle, and the actual vehicle information is compared with the vehicle information collected by S1. If there is any inconsistency, an abnormality report is issued. S2 specifically includes the following steps:

[0114] S21, calculating the theoretical bucket loading times according to the vehicle box volume;

[0115] S22, collecting video data during the bucket loading process, and counting the bucket loading actions according to the video data to obtain the actual number of bucket loadings;

[0116] S22 includes the following steps:

[0117] S221, collecting video data during bucket loading;

[0118] S222, identify the bucket loading state picture, unloading state picture, and empty bucket state picture in the video data as a bucket loading action, and count the number of bucket loading actions after visually observing the loading completion as the bucket loading times. The vehicle loading capacity recognition model can quickly identify whether the bucket is full, which improves the counting accuracy compared to manual counting (manual counting may cause misjudgment due to empty bucket loading actions).

[0119] S23, comparing the theoretical bucket loading times and the actual bucket loading times;

[0120] If the error exceeds the set threshold (the set threshold is 5%-10%, preferably 8%), an abnormality report is issued.

[0121] By comparing the theoretical bucket loading times with the actual bucket loading times, we can avoid not only the inaccurate loading calculation caused by half-loaded truck or sludge at the bottom, but also the inaccurate loading caused by human negligence, further improving the accuracy of earthwork transportation cost calculation.

[0122] S3, collecting the image of the carriage after loading, and judging the fullness of the carriage;

[0123] S3 includes the following steps:

[0124] S31, the vehicle box image after acquisition;

[0125] S32, inputting the completed vehicle box image into the vehicle box loading capacity recognition model;

[0126] S33, the vehicle loading capacity recognition model outputs the vehicle state of the current vehicle, and the vehicle state includes empty, fully loaded, and abnormal;

[0127] If the vehicle box is fully loaded, proceed to step S4;

[0128] If it is empty or abnormal, an abnormal report will be issued to alert the excavator operator;

[0129] S4. Calculate the actual load according to the number of bucket loading times, compare the actual load with the full load, and issue an abnormal report if it exceeds a set threshold (the set threshold is 5%-10%, preferably 8%).

[0130] Finally, the loading rate of the carriage is determined by the carriage loading capacity recognition model, avoiding the subjective bias in manual judgment of the loading rate and further improving the accuracy of earthwork transportation cost calculation.

[0131] In other embodiments, the bucket state recognition model is used to detect the loading amount of the bucket each time, and the actual loading amount is calculated based on the number of bucket loading times. The specific process is as follows:

[0132] The bucket state recognition model includes a bucket action recognition model and a bucket load recognition model. The bucket action recognition model is used to recognize and record the number of bucket loadings, and the bucket load recognition model is used to recognize and record the bucket load each time.

[0133] The bucket action recognition model is specifically to collect the relative position between the bucket and the small swing arm of the excavator in real time through a camera device, and determine the current action state of the bucket according to the collection result; wherein the action state includes the loading completion state and the unloading completion state;

[0134] The judgment process of the collection results is to preset a visual positioning line in the camera device, and the relative position of the visual positioning line and the small swing arm of the excavator is fixed. When the bucket is completely above the visual positioning line, it is judged that the bucket is in an unloading completion state at this time. Otherwise, it is judged that the bucket is in a loading completion state at this time.

[0135] After obtaining the actual loading capacity, obtain the actual loading capacity of the vehicle box and compare it to determine whether the current loading situation is abnormal.

[0136] Compared with video capture and image recognition of bucket loading actions, the visual positioning line is used as a reference to determine the position status of the bucket. When the bucket is completely above the visual positioning line, the system determines that the unloading is completed. The judgment process is more convenient and improves work efficiency. When the bucket is not above the visual positioning line, the system determines that the loading is completed. The application of the bucket state recognition model not only improves the automation level of loading operations, but also helps to reduce human errors and improve work efficiency. It can also be used to monitor abnormal conditions during the loading process, such as overloading or underloading, to ensure operation safety and compliance. In this way, engineering projects can manage resources more effectively, reduce costs, and improve overall operation quality.

[0137] In S4, the theoretical load is calculated based on the full load rate and the volume of the vehicle compartment identified by the vehicle compartment load identification model, and the theoretical load is compared with the actual load. If the error exceeds the set threshold, an abnormality report is issued.

[0138] In some embodiments, the training process of the vehicle loading capacity recognition model is as follows:

[0139] B1. Collect historical carriage photos and add a label to each historical carriage photo. The label is the loading rate. A loading rate within the first interval is considered fully loaded, a loading rate within the second interval is considered abnormal, and a loading rate within the third interval is considered empty;

[0140] The first interval is 90%-100%;

[0141] The second interval is 5%-90%;

[0142] The third interval is 0%-5%;

[0143] The unloaded label also includes vehicle information such as the vehicle model;

[0144] B2. Preprocessing of noise reduction and normalization of historical carriage photos;

[0145] B3, extract features of historical carriage photos;

[0146] B4. Use the features extracted in B3 to train the neural network model to generate a vehicle loading capacity recognition model.

[0147] The training method of the bucket load recognition model is the same as the training process of the car body load recognition model and will not be repeated here. The bucket is marked as the loading rate, and the final theoretical loading capacity is obtained by accumulating multiple bucket loading calculations each time. Compared with the ordinary method of only counting the number of bucket loadings, the theoretical loading capacity can be accurately calculated by identifying the bucket loading rate, thereby improving the accuracy of the calculation results.

[0148] In other embodiments, the above-mentioned vehicle loading capacity recognition model and bucket loading capacity recognition model are obtained through adaptive adjustment of the visual recognition model in different application scenarios. The construction process of the visual recognition model is as follows:

[0149] C1. Obtain the specifications of the carriages of different types of transport vehicles and the specifications of the buckets of different types of excavators, create multiple three-dimensional models as container models, and at the same time, create filler blocks. The specifications of the filler blocks are set according to actual physical parameters and calculation needs;

[0150] A physics engine is introduced to add physical effects of filler movement and stacking in the above three-dimensional model, and parameters such as gravity, physical collision, and friction coefficient are adjusted to corresponding values.

[0151] C2, the generator network is responsible for simulating the process of adding fillers in the above three-dimensional model, randomly generating three-dimensional models with different filling ratios and filling methods, and extracting observation images from different angles;

[0152] C3. The discriminator network determines the filling ratio of the filler in the corresponding 3D model based on the observation image sent by the generator network, verifies the judgment result based on the filling ratio, and adjusts the discriminator based on the verification result until the judgment accuracy of the discriminator meets the business needs;

[0153] C4. Migrate the trained multiple generator networks to the convolutional neural network to obtain a visual recognition model.

[0154] When a vehicle enters the construction site, the vehicle model is obtained through license plates, card swiping, etc., and then the corresponding three-dimensional model is matched. The corresponding visual recognition model is called according to the three-dimensional model, so as to use the visual recognition model to realize the loading capacity recognition function of the vehicle box and bucket.

[0155] In some embodiments, the training process of the discriminator network is as follows:

[0156] When the judgment result of the discriminator network is incorrect, the loss function is calculated:

[0157] ;

[0158] in, is the actual fill ratio, is the padding ratio predicted by the discriminator, δ is a hyperparameter used to control the turning point from quadratic to linear loss;

[0159] By calculating the partial derivative of the loss function with respect to the predicted value, the output layer error is obtained:

[0160] ;

[0161] Starting from the output layer, the error of the output layer is passed forward to the hidden layer layer by layer, and the error of each layer is calculated. The calculation method of the error term of the hidden layer depends on the error of the output layer. Using the chain rule, the error term propagates to the hidden layer as follows:

[0162] ;

[0163] in, is the weight of the output layer, f′ is the derivative of the activation function, is the net input value of hidden layer node j;

[0164] ;in, is the weight from the previous layer node i to the hidden layer node j, is the output (activation value) of node i in the previous layer, is the bias of hidden layer node j.

[0165] After full propagation, the hyperparameters are updated, the loss function is updated, and the discriminator network makes judgments again. The above process is repeated until the discriminator meets the requirements.

[0166] Compared with the historical data training model, a large amount of data can be automatically generated by generating a 3D model, which has a wider range of applications. The vehicle model can be obtained through license plate recognition or card swiping, and then the corresponding 3D model is matched, and the corresponding visual recognition model is called to estimate the load. Historical data requires the use of a large amount of historical data, which improves the automation and accuracy of load recognition.

[0167] The above method can timely discover abnormal loading conditions of transport vehicles and issue abnormal reports, thus ensuring the accuracy of transportation fee calculation.

[0168] An earthwork construction management system, comprising:

[0169] An information collection module configured to collect transport vehicle information through information tags and GPS;

[0170] A camera configured to collect bucket loading video and image information before and after the vehicle box is loaded;

[0171] A data processing module is configured to perform the following steps:

[0172] Identify the volume of the transport vehicle's box, count the number of bucket loadings during earthwork loading based on the bucket loading video, and determine the fullness of the box;

[0173] Calculate the actual load based on the number of bucket loading times, compare the actual load with the full load, and issue an abnormal report as an early warning if the vehicle exceeds the set threshold;

[0174] A display input module, which is configured to display the transport vehicle information and the processing results of the data processing module, and send an order opening instruction to the management platform;

[0175] The management platform is configured to record the transport vehicle list, excavator list, data collected by cameras, generate waybill data according to order opening instructions, and record historical waybill data.

[0176] The transport vehicle is provided with an information tag, and the excavator is provided with an information collector. The information tag on the transport vehicle is collected by the information collector. The information tag corresponds to the vehicle information in the database. The vehicle information includes the model of the transport vehicle and the volume of the vehicle box corresponding to the model.

[0177] The information tag is an RFID tag or a Bluetooth tag.

[0178] The information collector is an RFID signal collector or a tablet or smart phone.

[0179] The display input module is a tablet. Click Start Task to pass the order opening instruction to the backend, and the database of the management platform creates the waybill data. The waybill data includes the vehicle information consisting of the waybill number and vehicle number obtained through the camera;

[0180] After clicking Start Task, the camera starts collecting image information of the vehicle box;

[0181] After the task is completed, the camera is turned off and the image information of the vehicle compartment is stopped;

[0182] The exception report sent by the data processing module is displayed and sent to the management platform, which stores the exception report.

[0183] The command to end the task can be obtained by:

[0184] Manually select the End Task option on the tablet;

[0185] Automatically identify the end of the task through the excavator action.

[0186] It also includes construction waste soil loading equipment, which includes an engineering operation fleet consisting of excavators and waste soil transport vehicles);

[0187] The loading process monitoring device includes an information collection module, which includes but is not limited to a video collection device installed at the central support point of the excavator arm, a vehicle identification device installed on the muck transport vehicle, and a sensing device installed in the excavator cab and coordinated with the vehicle identification device. The vehicle identification device can be an RFID tag, and the sensing device can be an RFID signal collector. The vehicle identification device can also use a Bluetooth external tag, and the sensing device can also use a tablet, a smart phone or other independently developed equipment.

[0188] The video acquisition device includes a camera and a router. The camera is installed at the central support point of the excavator's digging arm. This position is the center of gravity of the excavator, the hoisting position of the excavator, and the installation position of the spotlight. It can ensure the stability of the camera and ensure that the camera has enough light to carry out daytime and nighttime operations. The camera is bound to the excavator, and the binding relationship table is stored in the management platform. After the camera is installed, it is connected to the router intranet. The camera is equipped with an algorithm module. The camera communication performs algorithm recognition and realizes tablet interaction through the router intranet communication. The core functions of the camera are: residual material recognition, full load recognition, and bucket counting.

[0189] Camera: video acquisition module, processing module (algorithm module), communication module. The processing module is the algorithm module: Under normal circumstances, the video acquisition module is used to collect images or video data and transmit it to the background for calculation, but the working field of the present invention is uncertain, and it is generally an undeveloped area with a poor network environment, and there are often serious delays. Therefore, the edge computing method is adopted, that is, the camera has a built-in algorithm module to collect data while processing data. The video acquisition module is an edge computing box that collects videos and pictures on the scene and provides each frame of the picture, which can be a full frame or a frame extraction, segmentation, recognition, and comparison with the picture we set to achieve the corresponding judgment structure. The key frame can be a set full bucket, empty bucket, half bucket, special action. After obtaining the picture, compare it with the model, match the degree, and roughly how much proportion matches and matches. After training, the model is obtained, developed on the PC side, converted, and transplanted to the camera side to adapt to the mobile environment. After the model is converted, it is also necessary to tune, such as determining full frame acquisition or frame extraction acquisition, and program adaptation. The model is related to the amount of collection. The processing module (algorithm module) includes communication and calculation functions. It receives commands from other devices, and the external device tells the camera to take action. The command receiving module receives commands from external devices. The command sending module sends the results to the tablet or other matching devices after receiving the command.

[0190] After the camera is turned on, the business processing center is automatically opened. The video acquisition module includes: image capture module and recognition module. The image capture module takes frames (full frame, or frame extraction) from the camera video stream, recognizes and matches each frame, recognizes the image through the trained model, and performs confidence matching. The empty and full load status of the bucket and the car box are obtained through this process. Task process: After receiving the connection command, the external tablet is connected to the camera, obtained through the communication module, and the acquisition module is activated. The subsequent task is started, the residual material recognition (the command issued by the external device), the driver determines the operation vehicle at the same time. Full load recognition, the driver determines the end. Start the task command (connection of the external device), it is time to work, feedback to the tablet, OK. Connect the residual material recognition, tell the camera to recognize the silt, and after recognition, communicate and feedback to the external tablet. After the external device is connected, the bucket technology algorithm enters the full load recognition and silt recognition. Until the external device gives the end command, the bucket count recognition ends. After the external device is connected: start 2 modules. Silt, full load, get pictures at intervals, input the model, and get the data of the corresponding calibration state. Counting buckets is also a functional module. It is passively started, and is told by the outside what operation to perform at what time. The camera is not a separate video channel, but two video channels, primary and secondary. One video channel is for viewing, and the other is for calculation.

[0191] The management platform includes a data processing module, a judgment module, a video acquisition device, a vehicle identification device, a sensing device cooperating with the vehicle identification device and the management platform, which are connected via wireless signals.

[0192] The tablet is a terminal and client, and is installed with an APP to provide interactive functions. The camera is responsible for collecting video streams, obtaining pictures, and performing calculations locally on the camera through the algorithm module. Then, the router interacts with the tablet, and the tablet interacts with the management platform. The vehicle can be an RFID device, but the external RFID on the tablet requires technical externalization, which is relatively expensive. Therefore, the signal induction is modified, and the Bluetooth of the tablet itself is used. The Bluetooth external tag is used on the muck truck, which can effectively reduce the overall cost. However, RFID strength and distance are not the only influencing factors. They will be affected by other factors, and the correlation is not strong. There is a problem that the distance and signal strength are not comparable, and the results are not accurate. Of course, this problem can also be solved through special settings, such as increasing the power of the RFID tag, only judging the entry, not judging the distance, sensing the signal, even if it is entered, not using the distance as the entry mark, uninterrupted sensing, if the tag is not detected, it will be notified to leave.

[0193] The flat panel has two modules, the task module - opening orders, ending, receiving the return confirmation of the start of the task, telling the equipment that it has been started and the work can proceed normally. Receive the residual material identification results, full load identification results and bucket identification results. One bucket, two buckets, three buckets, and the number of buckets are sent and received in real time. Silt and full load are sent in real time. The flat panel receives the silt framing and full load status, and transmits it to the background system in real time as a transfer station. The flat panel needs to compare the vehicle presence information table with the background. Bluetooth only obtains the Bluetooth device number, and the Bluetooth device number and detailed information of the vehicle are obtained by interacting with the background database.

[0194] After pressing the start button on the tablet, the order opening instruction is first sent to the backend, and the database creates an order (the backend database has an order opening module). The backend provides the following information: the waybill number and the vehicle number. This is the backend record. Except for the license plate and the construction site name, the camera algorithm results are all transferred by the tablet. Then the start instruction is sent to the camera, which only sends the start signal and the power-on signal. The camera does not need to perform recognition.

[0195] Communication module: Bluetooth module is used to assist the task module, retrieve vehicles, issue orders and push vehicles on site. The video display module displays the area where the excavator is working in real time to supplement the blind spots of the vision under the pit. The tablet will also display: abnormal, normal, which is the calculation result of the camera data. When the tablet receives an abnormality, it can directly display the abnormality. It will also display the license plate and the retrieved nearby Bluetooth vehicles. The tablet retrieves the list of vehicles by itself and feeds it back to the background. The background stores all the vehicle tables, both those present and those not present. Entry rules: Scan and tell the background. Different tablets collect different vehicles and store them in the background. The tablet only displays the list of vehicles it has scanned.

[0196] Before the loading operation begins, the remaining material is identified first: after receiving the judgment instruction, the algorithm module immediately performs the carriage identification. If there is an identification result in the carriage, the identification result is immediately returned. If there is no identification result within the specified time, the abnormal identification result is returned, and the system defaults to the presence of remaining material in the carriage.

[0197] For residual material recognition, the camera is set to obtain images of the front third of the car body, and photos with silt are marked to train the system to identify the presence of silt. If there is silt, the excavator is notified to loosen the silt so that it can be removed when the vehicle is unloaded, so that the next loading can be normal. When silt hangs on the walls, the existing camera is single-channel and has no depth information, so it is very likely to make a misjudgment. In this case, the plane data needs to be converted into three-dimensional data, and human intervention is required to train the system to recognize it by comparing the collected photos with the set photos.

[0198] Full load identification is performed after the loading operation is completed: after receiving the judgment instruction, the algorithm immediately performs car compartment identification. If there is an identification result for the car compartment, the identification result is returned immediately. If there is no identification result for a specified period of time, an abnormal identification result is returned. The system defaults to the car compartment being not fully loaded.

[0199] In theory, a full compartment can be directly seen from the picture, but if there is a shadow, the system will prompt that there is a problem, so you can mark it, collect photos, inform the system that this situation is full, and train the system. The logic of prompting an error can be exchanged with the number of full buckets, because the volume data of the vehicle compartment can be obtained, and the number of buckets that theoretically fill the compartment can also be obtained, such as: 15 buckets are full, and the number of buckets can be used to verify whether the compartment is full. That is, when the picture has a shadow, etc., it prompts the system that it is not full, but the number of buckets is full, the picture can be marked, and the system can be trained to recognize such pictures. When judging whether the compartment is full, it is preferred to use the vehicle photo as the basis to judge the full load of the vehicle.

[0200] There are two types of full load: flat and piled. The volume of the excavator bucket is fixed, and full load can be achieved by pushing it down. The picture is a standard, and the volume of the bucket is also set according to different working conditions, such as flat loading, pile loading and other scenes, and the full load situation is set according to different projects. The picture is a flat car box. Inward dumping means large capacity loading, full load, and pile loading can also be set as full load. Off the road: flat loading. If the error is within a certain degree, the conditions are met and no warning is given. The picture is the standard, and the amount of the bucket is auxiliary. Get 1.2, 1.3.1.4. buckets. Through Bluetooth data, the volume of the transport vehicle is obtained, and the number of buckets after filling is calculated to verify the workload of the bucket and the volume of the vehicle. After obtaining more data, the problem of quantity can be solved. There are inaccuracies in the simple picture: different soil types lead to different loading processes. Different shadows have errors. The actual loading volume is calculated by the loading volume of each bucket image and the vehicle volume has data, which can verify the accuracy of the road.

[0201] Count the number of bucket operations during loading: After receiving the task start command, the algorithm continues to identify the bucket and the car body. The identification is carried out throughout the entire task process. The specific counting method is to combine the identification with the set logical conditions as the bucket number calculation standard. The bucket shape recognition is divided into three states: (1) full bucket, (2) empty bucket mark, (3) half bucket, with material but not full, and the three states need to be labeled. The bucket action recognition is (1) bucket dumping action mark, and the car body recognition is the car body mark. The agreed judgment cycle is used, and the empty bucket is used as the cycle start mark. When the bucket dumping action mark is encountered or the full bucket mark is changed to the empty bucket mark and there is a car body mark, a bucket is recorded and the cycle ends. Full bucket, (2) empty bucket mark, (3) half bucket, with material but not full, three states need to be labeled. After a certain amount of data, a model is built, and similar graphics are compared. Adding 1.2, 1.3, and 1.4 buckets can achieve accurate matching of various situations. The implementation technology is the existing technology, and this implementation method is the innovation of this case. When identifying residual materials, the image of the front third of the car is collected. When the silt hangs on the four walls, it is very likely to be misjudged. The reason why it is difficult to identify the silt is that the volume of the car is fixed, length, width, and height. The camera is single-channel and has no depth information. Depth information needs to be loaded. To change the plane into three-dimensional, human data intervention is required. The system is trained to identify by comparing the photo with the set photo.

[0202] The vehicle identification process is as follows:

[0203] First, recognize the vehicle license plate number by identifying the video signal;

[0204] Second, identify other markers, such as the top of the vehicle, such as the front, rear, left, and right sides, and set unique identification marks, such as the unique mark of a certain company.

[0205] The unique identification mark can be the first type: the light box on the roof, the region of ownership, such as the Zen in Chancheng, and the ID card issued by the Garbage Association. The second identifier: the license plate, the color or pattern sprayed on the sides and tail plate. The third identifier: the door panel of the driver's building will have the identification of the company to which it belongs, all entered into the system, and can be an identification marker. Scanning and obtaining image information of any area can be used as an information source or a verification information source.

[0206] Vehicle identification adds positioning identification. The management platform can bind numbers, install GPS equipment in the fleet, or obtain the GPS signal of the fleet. Positioning data can be added to the tag, and Bluetooth identification tags can be added to the GPS to increase the GPS positioning of the vehicle. After the platform obtains the signal, there is high-precision positioning on the excavator and the vehicle, which can be transmitted to the background. After the background judgment, it is reflected on the tablet, the electronic fence, locked out, and after entering, it is regarded as admission. Auxiliary Bluetooth and camera, Bluetooth can be installed on the camera, and GPS positioning can also be installed for integration.

[0207] When using Bluetooth tag recognition, at least two vehicles must be identified, one is the working vehicle and the other is the queue vehicle. Under normal circumstances, the site conditions are particularly good (large and square), and there can be one working vehicle and multiple queue vehicles. The license plate is identified by the camera, and the license plate data is displayed on the tablet of the excavator. The information on the tablet comes from Bluetooth interaction (with license plate information), camera recognition, to verify the license plate, and finally displayed on the tablet. There are errors in the scanning, and a bunch of vehicles may appear. Therefore, identification and GPS assistance are used for verification.

[0208] In addition, a camera is installed in the entrance area and a camera is also installed in the factory area. After the excavator completes loading, it can send a completion signal, and the import camera and the export camera will verify it. Only when the three-party verification is consistent can it be determined that the loading is qualified. This approach effectively prevents the transport vehicle from unloading the earth to a certain area in the yard. If the transport vehicle unloads the earth to a certain area in the yard, the camera data of the import camera and the export camera will be lost.

[0209] Specific implementation process: Fully automatic: human-assisted system. The whole process involves equipment judgment, identification of the loading process, and determination of the operating vehicle. Semi-automatic: The system assists manual work. The vehicle comes over, the flatbed identifies it, enters the list, the driver selects it, and determines the operating vehicle. The sludge identification module is started simultaneously, the vehicle is empty, normal, the system does not warn, and it operates normally. When there is sludge, the remaining material is abnormal on the flatbed, and the driver digs it loose. Abnormal, the system pushes it to the official account, and all relevant personnel (including the first responsible person, the actual person in charge, the project manager, the on-site manager, the transport vehicle driver, and the transport company manager) push information. A certain vehicle, a certain construction site, those abnormalities, with identification photos, and the approximate amount of sludge. When to prompt the driver to press the end button, after the flattening action is completed, the driver presses the end button, and at the same time starts the judgment of the full load degree. Real-time judgment of whether it is full load, the progress bar is displayed based on the number of buckets, the progress bar is: simple number of buckets, comparison of bucket volume and vehicle volume, auxiliary verification. The system detects the number of buckets, the full load status of the car box, and several data to determine whether it is fully loaded. System implementation: After the driver selects, the photo at that time is recorded as the full load photo, and finally as the end mark. Add an action mark of the car pressing action (pressing the car and leaving) as the end mark. After the driver confirms, take a full load photo. Add the car pressing action mark and the photos collected by the system as a mutual verification.

[0210] Content displayed in the background: data, all stored in the background database, waybill information, etc. Real-time video module, work process and surrounding work vehicles. Vehicle presence information table, data sensed by the platform, transmitted to this module, and entry and exit data are also interacted in this module. The task status of the operating excavator is displayed: waybill number, vehicle license plate, driver, how many vehicles were transported today, how many were abnormal, and how many were full-loaded abnormal. Progress bar for excavation. Module of historical waybills, waybills, related information of transport vehicles, full vehicle load, abnormal information of residual materials, etc. Basic equipment maintenance information: camera equipment, tablet equipment, bound to the excavator. Excavator module: record of excavator information.

[0211] The present invention can realize the full automatic operation of the system, the relationship between signal strength and distance, and inform the system of the distance of the equipment. Only when it is within the set distance can the order be issued. The system starts by itself. When the signal of the excavator is automatically sensed within the range of 15M and displayed, after the camera recognizes the basic action, the full bucket + car box is regarded as the start (automatic order creation), silt, bucket counting, and full load start. When the signal leaves, it leaves. Load a bucket, rotate to the side of the car, and start automatically. The number of buckets + bucket + volume, the bucket volume data of the bucket meets the relevant matching standards, which is equal to full load. The camera will misjudge, the delay is serious, and the data is inaccurate. Manual start, manual end, and accuracy of capture. The Bluetooth system can provide a positive correlation between strength and distance, and control the vehicle category by distance. For those vehicles that can be issued, the vehicle list can provide all the vehicles that can be issued.

[0212] The information of the camera, the car box, the license plate and the Bluetooth information are matched, and then the order can be issued. Only when there is a misjudgment, the personnel intervene. If CPS is added, there will be no problem. It can be purely automated, and the manual work is only auxiliary. The image assists in judging the working vehicle. According to the distance, determine which of the 1, 2, 3, and 4 vehicles is the closest. After loading the soil, it starts working and measuring. After 1, it is loaded to 2, and it shows that both vehicles are loading. The system will judge that the second vehicle is loading. At this time, the personnel need to intervene. The number of buckets and the volume of the vehicle are wirelessly close to 100%. It cannot be confirmed whether it is full. After the signal leaves, it ends automatically. When the vehicle is 80% full, it leaves. If the full load is abnormal, it will not end and an abnormal push is required. In the case of full automation, the driver quickly loads the first bucket, which will affect the accuracy of the residual material recognition. Only the full load situation can be recognized, and the full automatic method can be realized. The residual material recognition is not easy to adapt to the full automation. So semi-automatic is added to determine the start of the residual material. When the full bucket and the empty car box overlap, it is the first bucket. Before each order is opened, the first bucket is dug, the second bucket, the full bucket and the empty car box overlap + the bucket dumping action can also be the first bucket.

[0213] There are two ways to issue orders. Both methods are acceptable. Loading two cars at the same time and issuing two orders is also possible. Three cars are also possible. There is no problem with the signal and recognition. Automatic ordering and manual ordering are to solve the problem of which car the bucket is going to. They all overlap with the identifier. The camera obtains the license plate and Bluetooth assists to identify the vehicle. License plate recognition is independent each time. The bucket has no features. The only feature is the license plate. Adding an identifier can identify two cars at the same time. Automatic ordering, demarcate a certain area in the video to achieve automatic ordering. Prompt the driver to wait a few seconds and handle the residual mud.

[0214] The present invention also includes a push module, and the main device includes: a server and a background management system. The background management system and the database are deployed to the cloud server. The administrator creates accounts for personnel at all levels, pushes the accounts into groups, sets push rules for the groups, performs group binding in the push type, and then provides the binding QR code corresponding to the account provided by the system to the user who needs to receive the push. The user only needs to follow the designated public account and scan the account provided by the administrator for binding. When the waybill is completed, the corresponding type, group, and rule will be judged and pushed. Core functions: abnormal push, summary push. Core logic: After the user scans the QR code, the public account can obtain a unique identifier, and the public account can retrieve the concerned user through the unique identifier and push targeted content to it. The abnormal state is recorded in the task flow, and after the task is completed, the push rule is matched according to the abnormal information and the task order information. At twelve o'clock in the morning, the server timed service is used to perform targeted push. Create an account first, create the basic information, log in to the account to bind WeChat, and follow the public account before pushing information. Abnormal push, summary push.

[0215] The present invention also includes a soil identification function. After collecting video images, the soil is identified. The density of the soil and the volume of the bucket can be used to obtain the weight, distinguish the purpose of the soil, and classify it into: planting soil, organic soil, loose soil, and miscellaneous soil with stones. According to different soil qualities, the destination and function are prompted. If there is soil, it can be used for paving. If the sand content is high, it can have other uses. There are more than 10 kinds of timely decisions. The destination after digging out is originally randomly piled up. If it is classified, it can be directly reused. This function is also realized by this camera to realize the function of identifying soil.

[0216] Embodiment 2:

[0217] An electronic device includes a processor and a memory connected to the processor for storing instructions executable by the processor, wherein the processor is used to execute the method of the first embodiment.

[0218] Embodiment three:

[0219] A server includes at least one processor and a memory connected to the processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor so that the at least one processor executes the method of the first embodiment.

[0220] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

[0221] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for earthwork construction management, characterized in that: The following steps are involved: S1. Collecting transport vehicle information, identifying the state of the transport vehicle's compartment, and identifying the volume of the transport vehicle's compartment; S2. Counting the number of bucket loadings and the amount of each bucket loading during the earthwork loading process; S3, collecting the image of the carriage after loading, and judging the fullness of the carriage; S4. Calculate the actual loading amount according to the number of bucket loadings and the loading amount, calculate the theoretical loading amount according to the full load rate and the volume of the vehicle box identified by the vehicle box state identification model, compare the theoretical loading amount with the actual loading amount, and issue an abnormality report if the error exceeds the set threshold; The identification of the vehicle box state of the transport vehicle in S1 is as follows: before the bucket is loaded, an image of the vehicle box before loading is collected, and whether there is residual material in the vehicle box is identified based on the image of the vehicle box before loading, and if there is no residual material, the process proceeds to step S2; The S2 specifically includes the following steps: S21, calculating the theoretical bucket loading times according to the vehicle box volume; S22, collecting video data during the bucket loading process, and counting the bucket loading actions according to the video data to obtain the actual number of bucket loadings; S23, comparing the theoretical bucket loading times and the actual bucket loading times; If the error exceeds the set threshold, an exception report is issued.

2. The earthwork construction management method according to claim 1, characterized in that: The S1 comprises the following steps: The transport vehicle is provided with an information tag, and the excavator is provided with an information collector. The information tag on the transport vehicle is collected by the information collector. The information tag corresponds to the vehicle information in the database. The vehicle information includes the model of the transport vehicle and the vehicle box volume corresponding to the model.

3. The earthwork construction management method according to claim 1, characterized in that: The S22 comprises the following steps: S221, collecting video data during bucket loading; S222, identifying the loaded bucket state picture, the unloaded bucket state picture, and the empty bucket state picture in the video data as a bucket loading action, and counting the number of bucket loading actions after visually observing that the loading is completed as the number of bucket loading times.

4. The earthwork construction management method according to claim 1, characterized in that: The training process of the vehicle box state recognition model is as follows: B1. Collect historical carriage photos and add a label to each historical carriage photo. The label is the loading rate. A loading rate within the first interval is considered fully loaded, a loading rate within the second interval is considered abnormal, and a loading rate within the third interval is considered empty; The first interval is 90%-100%; The second interval is 5%-90%; The third interval is 0%-5%; The unloaded label also includes vehicle information such as the vehicle model; B2. Preprocessing of noise reduction and normalization of historical carriage photos; B3, extract features of historical carriage photos; B4. Use the features extracted in B3 to train the neural network model to generate a vehicle state recognition model.

5. The earthwork construction management method according to claim 1, characterized in that: The S3 comprises the following steps: S31, the vehicle box image after acquisition; S32, inputting the completed carriage image into the carriage state recognition model; S33, the carriage state recognition model outputs the carriage state of the current carriage, and the carriage state includes empty, fully loaded, and abnormal; If the vehicle box is fully loaded, proceed to step S4; If it is empty or abnormal, an abnormal report will be issued to alert the excavator operator.

6. The earthwork construction management method according to claim 1, characterized in that: The specific steps for identifying whether there is residual material in the carriage based on the carriage image before loading are as follows: A1. Collect the image of the carriage before loading; A2, inputting the image of the carriage before loading into the carriage state recognition model; A3. The carriage state recognition model outputs the current carriage state, which includes empty, fully loaded, and abnormal. If the carriage is empty, it is determined that there is no remaining material; The vehicle state recognition model outputs the actual vehicle information of the current vehicle, and the actual vehicle information is compared with the vehicle information collected by S1. If there is any inconsistency, an abnormality report is issued.

7. An earthwork construction management system, characterized in that: It is used to implement the method described in any one of claims 1 to 6, comprising: An information collection module, which is configured to collect transport vehicle information through information tags and GPS; A camera configured to collect bucket loading video and image information before and after the vehicle box is loaded; A data processing module is configured to perform the following steps: Identify the volume of the transport vehicle's box, count the number of bucket loadings during earthwork loading based on the bucket loading video, and determine the fullness of the box; Calculate the actual load based on the number of bucket loading times, compare the actual load with the full load, and issue an abnormal report for early warning if the error exceeds the set threshold; A display input module, which is configured to display the transport vehicle information and the processing results of the data processing module, and send an order opening instruction to the management platform; The management platform is configured to record the transport vehicle list, excavator list, data collected by cameras, generate waybill data according to order opening instructions, and record historical waybill data.

8. The earthwork construction management system according to claim 7, characterized in that: The transport vehicle is provided with an information tag, and the excavator is provided with an information collector. The information tag on the transport vehicle is collected by the information collector. The information tag corresponds to the vehicle information in the database. The vehicle information includes the model of the transport vehicle and the volume of the vehicle box corresponding to the model. The information tag is an RFID tag or a Bluetooth tag. The information collector is an RFID signal collector or a tablet or a smart phone.

9. The earthwork construction management system according to claim 7, characterized in that: The display input module is a tablet. Click Start Task to pass the order opening instruction to the backend, and the database of the management platform creates the waybill data. The waybill data includes the vehicle information consisting of the waybill number and vehicle number obtained through the camera; After clicking Start Task, the camera starts collecting image information of the vehicle box; After the task is completed, the camera is turned off and the image information of the vehicle compartment is stopped; The exception report sent by the data processing module is displayed and sent to the management platform, which stores the exception report.

10. The earthwork construction management system according to claim 9, characterized in that: The command to end the task is obtained by the following method: Manually select the End Task option on the tablet; Automatically identify the end of the task through the excavator action.

11. An electronic device, comprising a processor and a memory connected to the processor for storing instructions executable by the processor, characterized in that: The processor is used to execute the method described in any one of claims 1 to 6.

12. A server, characterized in that: It includes at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor so that the at least one processor executes the method according to any one of claims 1-6.

13. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images