Borrowing resource monitoring methods, devices, computer equipment and storage media

Abstract

This application relates to a method, apparatus, computer equipment, storage medium, and computer program product for monitoring borrowed resources, and belongs to the field of artificial intelligence technology. The method includes: determining the target resource use and resource usage location information of the borrowed resources corresponding to the resource borrowing information; determining a borrowed resource monitoring model corresponding to the target resource use, and deploying the borrowed resource monitoring model to a drone; generating an image acquisition task including resource usage location information and sending it to the drone, enabling the drone to acquire multiple images of the target location corresponding to the resource usage location information; performing target detection processing on the multiple images of the target location based on the borrowed resource monitoring model deployed in the drone, obtaining a detection result for each image; and determining the monitoring result for the borrowed resources corresponding to the resource borrowing information based on the detection results of each image returned by the drone. Using this method can improve the monitoring efficiency of the actual resource use of borrowed resources.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for monitoring borrowed resources. Background Technology

[0002] Borrowed resources refer to resources that the resource recipient obtains from the resource provider in advance for a specified target purpose; therefore, the resource provider needs to monitor the actual use of the borrowed resources during the resource recipient's use of the borrowed resources.

[0003] In related technologies, monitoring the actual use of borrowed resources usually relies on on-site inspections by staff, which consumes a lot of time and manpower, resulting in low efficiency in monitoring the actual use of borrowed resources. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for monitoring borrowed resources that can improve the monitoring efficiency of the actual use of borrowed resources in addressing the aforementioned technical problems.

[0005] Firstly, this application provides a method for monitoring seconded resources. The method includes:

[0006] Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information;

[0007] Determine the resource borrowing monitoring model corresponding to the target resource use, and deploy the resource borrowing monitoring model to the drone;

[0008] Generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location;

[0009] Based on the resource borrowing monitoring model deployed in the UAV, target detection processing is performed on multiple images of the target location to obtain detection results for each image; the detection results are used to characterize the association between the image and the purpose of the target resource.

[0010] Based on the detection results of each image returned by the drone, the monitoring results of the borrowed resources corresponding to the resource borrowing information are determined.

[0011] In one embodiment, the resource borrowing monitoring model corresponding to the target resource usage is trained in the following manner:

[0012] Obtain a sample image corresponding to the intended use of the target resource, and the actual activity area in the sample image where there is a related behavior of the intended use of the target resource;

[0013] By using the borrowing resource monitoring model to be trained corresponding to the target resource use, the region in the sample image where there is a related behavior of the target resource use is predicted to obtain the predicted activity region.

[0014] Based on the actual activity area, the predicted activity area, and a preset adjustment factor, the loss information between the actual activity area and the predicted activity area is determined;

[0015] Based on the loss information, the borrowing resource monitoring model to be trained is trained to obtain the borrowing resource monitoring model corresponding to the target resource use.

[0016] In one embodiment, determining the loss information between the real activity region and the predicted activity region based on the real activity region, the predicted activity region, and a preset adjustment factor includes:

[0017] Determine the distance information, area difference information, and size difference information between the actual activity area and the predicted activity area;

[0018] Based on a preset adjustment factor, the distance information, the area difference information, and the size difference information are fused to obtain loss information between the actual activity area and the predicted activity area.

[0019] In one embodiment, determining the distance information between the actual activity region and the predicted activity region includes:

[0020] Determine the enclosing regions corresponding to the actual activity region and the predicted activity region, and based on the enclosing regions, determine the diagonal distance between the actual activity region and the predicted activity region;

[0021] The distance between the first center point of the actual activity area and the second center point of the predicted activity area is determined as the center distance between the actual activity area and the predicted activity area.

[0022] The diagonal distance and the center distance are determined as the distance information between the actual activity area and the predicted activity area.

[0023] In one embodiment, the target detection processing is performed on multiple images of the target location based on the resource borrowing monitoring model deployed in the UAV to obtain detection results for each image, including:

[0024] For each of the multiple images of the target location, the borrowed resource monitoring model deployed in the UAV identifies the activity area in the image that is associated with the use of the target resource, and determines the confidence information of the activity area.

[0025] The detection result of the image is determined based on the confidence information of the active region.

[0026] In one embodiment, determining the monitoring result of the borrowed resources corresponding to the resource borrowing information based on the detection result of each image returned by the drone includes:

[0027] Based on the detection results of each image, target images that exhibit associated behaviors related to the use of the target resource are identified among the multiple images.

[0028] If the number of images in the target image meets the preset image quantity condition, the monitoring result of the borrowed resources corresponding to the resource borrowing information is confirmed as qualified.

[0029] In one embodiment, after determining the monitoring result of the borrowed resources corresponding to the resource borrowing information based on the detection result of each image returned by the drone, the method further includes:

[0030] Based on the resource borrowing information, determine the quantity of borrowed resources and the recipient of the resources corresponding to the resource borrowing information;

[0031] Based on the monitoring results of the borrowed resources corresponding to the resource borrowing information and the quantity of borrowed resources, the borrowing credit information of the resource recipient is updated; the borrowing credit information is used to evaluate the resource borrowing request of the resource recipient.

[0032] Secondly, this application also provides a resource borrowing monitoring device. The device includes:

[0033] The information determination module is used to determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information based on the resource borrowing information;

[0034] The model deployment module is used to determine the borrowed resource monitoring model corresponding to the target resource use and deploy the borrowed resource monitoring model to the UAV;

[0035] An image acquisition module is used to generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location;

[0036] The target detection module is used to perform target detection processing on multiple images of the target location based on the borrowed resource monitoring model deployed in the UAV, and obtain the detection result for each image; the detection result is used to characterize the association between the image and the purpose of the target resource;

[0037] The result determination module is used to determine the monitoring result of the borrowed resources corresponding to the resource borrowing information based on the detection result of each image returned by the UAV.

[0038] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0039] Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information;

[0040] Determine the resource borrowing monitoring model corresponding to the target resource use, and deploy the resource borrowing monitoring model to the drone;

[0041] Generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location;

[0042] Based on the resource borrowing monitoring model deployed in the UAV, target detection processing is performed on multiple images of the target location to obtain detection results for each image; the detection results are used to characterize the association between the image and the purpose of the target resource.

[0043] Based on the detection results of each image returned by the drone, the monitoring results of the borrowed resources corresponding to the resource borrowing information are determined.

[0044] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0045] Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information;

[0046] Determine the resource borrowing monitoring model corresponding to the target resource use, and deploy the resource borrowing monitoring model to the drone;

[0047] Generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location;

[0048] Based on the resource borrowing monitoring model deployed in the UAV, target detection processing is performed on multiple images of the target location to obtain detection results for each image; the detection results are used to characterize the association between the image and the purpose of the target resource.

[0049] Based on the detection results of each image returned by the drone, the monitoring results of the borrowed resources corresponding to the resource borrowing information are determined.

[0050] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0051] Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information;

[0052] Determine the resource borrowing monitoring model corresponding to the target resource use, and deploy the resource borrowing monitoring model to the drone;

[0053] Generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location;

[0054] Based on the resource borrowing monitoring model deployed in the UAV, target detection processing is performed on multiple images of the target location to obtain detection results for each image; the detection results are used to characterize the association between the image and the purpose of the target resource.

[0055] Based on the detection results of each image returned by the drone, the monitoring results of the borrowed resources corresponding to the resource borrowing information are determined.

[0056] The aforementioned resource borrowing monitoring method, apparatus, computer equipment, storage medium, and computer program product first determine the target resource use and resource usage location information of the borrowed resources based on the resource borrowing information; then, they determine the borrowed resource monitoring model corresponding to the target resource use and deploy the borrowed resource monitoring model to a drone; next, they generate an image acquisition task including resource usage location information, send the image acquisition task to the drone, causing the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location; then, based on the borrowed resource monitoring model deployed in the drone, they perform target detection processing on the multiple images of the target location to obtain a detection result for each image; the detection result is used to characterize the correlation between the image and the target resource use; finally, based on the detection result of each image returned by the drone, they determine the monitoring result for the borrowed resources corresponding to the resource borrowing information. In this way, by using information about the target resource use and location of the borrowed resources, a corresponding monitoring model for the borrowed resources can be determined, and a corresponding image acquisition task can be generated. Then, by executing the image acquisition task using a drone, multiple images of the target location corresponding to the resource usage location information can be remotely acquired. Based on the borrowed resource monitoring model deployed on the drone, it is possible to remotely detect whether there are any behaviors at the target location related to the target resource use, thereby achieving remote monitoring of the actual resource use of the borrowed resources. In the above-described borrowed resource monitoring method, the monitoring of the actual resource use of the borrowed resources does not rely on on-site inspections by staff, saving time and manpower, thus improving the efficiency of monitoring the actual resource use of the borrowed resources. Attached Figure Description

[0057] Figure 1 This is an application environment diagram of the resource borrowing monitoring method in one embodiment;

[0058] Figure 2 This is a flowchart illustrating a resource borrowing monitoring method in one embodiment;

[0059] Figure 3 This is a flowchart illustrating the steps of training a resource borrowing monitoring model corresponding to the intended use of a target resource in one embodiment.

[0060] Figure 4 This is a schematic diagram of the center distance and diagonal distance in one embodiment;

[0061] Figure 5(a) is a schematic diagram of the intersection area between the actual activity area and the predicted activity area in one embodiment;

[0062] Figure 5(b) is a schematic diagram of the union region between the actual activity region and the predicted activity region in one embodiment;

[0063] Figure 6This is a flowchart illustrating the resource borrowing monitoring method in another embodiment;

[0064] Figure 7 This is a structural block diagram of a resource borrowing monitoring device in one embodiment;

[0065] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0067] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0068] It should also be noted that the secondment resource monitoring method, device, computer equipment, storage medium and computer program products provided in this application can be applied to the financial technology field, such as monitoring the actual use of loans such as agricultural loans, livestock loans, self-built housing loans and post-disaster reconstruction loans; they can also be used in other related fields, such as in the field of computer technology, to monitor the actual use of electronic secondment resources in the real world through the secondment resource monitoring method provided in this application.

[0069] The resource borrowing monitoring method provided in this application embodiment can be applied to, for example... Figure 1 The application environment shown includes a server 102 and a drone 104; the server 102 communicates with the drone 104 via a network. The server 102 can be implemented as a standalone server or a server cluster consisting of multiple servers; the drone 104 is a commercially available drone device with image acquisition capabilities.

[0070] Specifically, server 102 first determines the target resource use and resource usage location information of the borrowed resource based on the resource borrowing information. Then, server 102 determines the borrowed resource monitoring model corresponding to the target resource use and deploys the borrowed resource monitoring model to drone 104. Next, server 102 generates an image acquisition task including resource usage location information and sends the image acquisition task to drone 104. After receiving the image acquisition task, drone 104 flies to the target location corresponding to the resource usage location information and acquires multiple images of the target location. Then, drone 104 performs target detection processing on the multiple images of the target location through the deployed borrowed resource monitoring model, obtains the detection result for each image, and returns the detection result of each image to server 102. The detection result is used to characterize the association between the image and the target resource use, such as whether there is a related behavior of the target resource use on the image. Finally, server 102 determines the monitoring result of the borrowed resource corresponding to the resource borrowing information based on the detection result of each image returned by drone 104, such as whether the actual resource use of the borrowed resource is consistent with the target resource use of the borrowed resource.

[0071] The following describes in detail the resource secondment monitoring method provided in this application, using a bank's monitoring scenario of the actual use of loans. In an exemplary embodiment, as follows... Figure 2 As shown, a method for monitoring borrowed resources is provided, which is then applied to... Figure 1 Taking the server in the example, the steps include:

[0072] Step S202: Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information.

[0073] Among them, borrowed resources refer to resources that the recipient obtains from the resource provider in advance, such as a loan obtained by a lender from a bank. It is understandable that borrowed resources need to be used for a specific purpose. For example, borrowed resources for self-built housing need to be used for building self-built housing, borrowed resources for agriculture need to be used for agricultural activities, and borrowed resources for post-disaster reconstruction need to be used for post-disaster reconstruction, etc.

[0074] Among them, resource borrowing information refers to the resource borrowing contract signed between the resource recipient and the resource provider when the resource recipient obtains borrowed resources from the resource provider, such as the loan contract signed between the lender and the bank.

[0075] The target resource use refers to the designated purpose of the borrowed resources. For example, for agricultural borrowed resources, the target resource use is agricultural activities; for livestock borrowed resources, the target resource use is livestock activities.

[0076] Among them, the resource usage location information is the location information for using borrowed resources that the resource recipient promises to the resource provider.

[0077] Specifically, the server first obtains the resource borrowing information, and then, through text detection, identifies the target resource purpose and resource usage location information of the borrowed resource from a specified position in the resource borrowing information.

[0078] For example, suppose the server uses a text detection algorithm to identify, from a specified text location in the resource loan agreement, that the type of loaned resource recorded in the agreement is agricultural loaned resource. Then, agricultural activity is the target resource use of the agricultural loaned resource corresponding to the resource loan agreement. At the same time, the server can also use a text detection algorithm to identify, from a specified text location in the resource loan agreement, the location information of the location where the resource recipient promises to use the loaned resource, such as the latitude and longitude of the farmland and the size of the farmland. This location information is the resource use location information of the agricultural loaned resource corresponding to the resource loan agreement.

[0079] Step S204: Determine the resource borrowing monitoring model corresponding to the target resource use, and deploy the resource borrowing monitoring model to the UAV.

[0080] Among them, the borrowed resource monitoring model corresponding to the target resource use is a target monitoring model trained based on sample images of the target resource use, such as the YOLOv5 model (You Only Look Once version 5, a target detection model for identifying and locating objects in videos or images); the borrowed resource monitoring model is used to detect whether there are related behaviors of the target resource use at the target location corresponding to the resource use location information, such as detecting whether there are agricultural behaviors related to agricultural activities on the farmland of the resource recipient.

[0081] Among them, drones are commonly used drone devices on the market that have the function of acquiring images.

[0082] Specifically, the server determines the borrowed resource monitoring model corresponding to the target resource use based on the target resource use, and deploys the borrowed resource monitoring model to the drone.

[0083] For example, assuming the target resource is used for agricultural activities, the server will use a borrowed resource monitoring model to monitor agricultural activities, which will be determined as the borrowed resource monitoring model corresponding to the target resource use; assuming the target resource is used for aquaculture activities, the server will use a borrowed resource monitoring model to monitor aquaculture activities, which will be determined as the borrowed resource monitoring model corresponding to the target resource use.

[0084] Step S206: Generate an image acquisition task including resource usage location information, send the image acquisition task to the UAV, and make the UAV fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location.

[0085] The image acquisition task is used to trigger the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location.

[0086] Specifically, the server generates an image acquisition task that includes the resource usage location information based on the resource usage location information, and sends the image acquisition task to the drone; after receiving the image acquisition task, the drone flies to the target location corresponding to the resource usage location information and acquires multiple images of the target location.

[0087] For example, assuming the target resource is used for agricultural activities, and the resource usage location information includes the latitude and longitude of the farmland and its size range, the server generates an image acquisition task based on the latitude and longitude of the farmland and its size range, and sends the image acquisition task to the drone; the drone flies to the farmland based on the latitude and longitude of the farmland and its size range, and collects multiple images of the farmland.

[0088] Step S208: Based on the resource borrowing monitoring model deployed in the UAV, target detection processing is performed on multiple images of the target location to obtain the detection results for each image.

[0089] The detection results are used to characterize the association between the image and the target resource use; the association between the image and the target resource use refers to whether there is any associated behavior of the target resource use in the image.

[0090] Specifically, the server uses a borrowing resource monitoring model deployed in the drone that corresponds to the target resource usage of the borrowed resources to perform target detection processing on multiple images of the target location, detects whether there are related behaviors of the target resource usage on each image, and thus obtains the detection result of each image.

[0091] For example, the server uses a resource borrowing monitoring model corresponding to agricultural activities to perform target detection on multiple images of farmland, and determines whether there are agricultural activities related to agricultural activities in each image. If there are, the image is marked as "agricultural activities exist"; if not, the image is marked as "agricultural activities do not exist".

[0092] Step S210: Based on the detection results of each image returned by the UAV, determine the monitoring results of the borrowed resources corresponding to the resource borrowing information.

[0093] Among them, the monitoring results of the borrowed resources corresponding to the resource borrowing information are used to characterize whether the actual resource use of the borrowed resources corresponding to the resource borrowing information is consistent with the target resource use of the borrowed resources.

[0094] Specifically, based on the distribution of detection results for each image returned by the drone, the server determines whether the recipient of the borrowed resources corresponding to the resource borrowing information uses the borrowed resources for the target purpose, and uses the determination result as the monitoring result for the actual use of the borrowed resources corresponding to the resource borrowing information.

[0095] For example, if, among multiple images of farmland, the number of images marked "agricultural activity exists" is greater than a preset image number threshold, or if the ratio between the number of images marked "agricultural activity exists" and the total number of images is greater than a preset ratio, then the server can confirm that the recipient of the agricultural borrowed resources corresponding to the resource borrowing information will use the agricultural borrowed resources for agricultural activities.

[0096] In the resource borrowing monitoring method provided in the above embodiments, the server first determines the target resource use and resource usage location information of the borrowed resource based on the resource borrowing information; then, it determines the borrowing resource monitoring model corresponding to the target resource use and deploys the borrowing resource monitoring model to the drone; next, it generates an image acquisition task including resource usage location information, sends the image acquisition task to the drone, and causes the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location; then, based on the borrowing resource monitoring model deployed in the drone, it performs target detection processing on the multiple images of the target location to obtain a detection result for each image; the detection result is used to characterize the association between the image and the target resource use; finally, based on the detection result of each image returned by the drone, it determines the monitoring result for the borrowed resource corresponding to the resource borrowing information. In this way, by using the target resource usage and location information of the borrowed resources, the server can determine the corresponding borrowed resource monitoring model and generate the corresponding image acquisition task. Then, by executing the image acquisition task through a drone, the server can remotely acquire multiple images of the target location corresponding to the resource usage location information. Based on the borrowed resource monitoring model deployed on the drone, the server can remotely detect whether there are any related behaviors at the target location that are associated with the target resource usage, thereby realizing remote monitoring of the actual resource usage of the borrowed resources. In the above borrowed resource monitoring method, the monitoring of the actual resource usage of the borrowed resources does not rely on on-site inspections by staff, saving time and manpower, thereby improving the monitoring efficiency of the actual resource usage of the borrowed resources.

[0097] like Figure 3As shown in an exemplary embodiment, the resource borrowing monitoring model corresponding to the target resource usage in the above embodiments is trained in the following manner:

[0098] Step S302: Obtain a sample image corresponding to the target resource use, and the real activity area in the sample image where there is a related behavior of the target resource use.

[0099] Step S304: Using the borrowing resource monitoring model to be trained corresponding to the target resource use, predict the regions in the sample image that contain associated behaviors of the target resource use to obtain the predicted activity regions.

[0100] Step S306: Determine the loss information between the actual activity area and the predicted activity area based on the actual activity area, the predicted activity area, and the preset adjustment factor.

[0101] Step S308: Based on the loss information, train the borrowing resource monitoring model to be trained to obtain the borrowing resource monitoring model corresponding to the target resource usage.

[0102] It is understandable that in the field of object detection, the real active region is the region where the target to be detected actually exists, which can be represented by a bounding box; the predicted active region is the region where the target to be detected exists as identified by the object detection model, which can be represented by a predicted bounding box.

[0103] Specifically, for different target resource uses, the server obtains multiple sample images corresponding to the target resource use from the Internet, and performs the following operations on each sample image: Through manual labeling, regions in the sample images that exhibit related behaviors for the target resource use are identified as the true activity regions of the sample images; then, the server uses the borrowing resource monitoring model to be trained for the target resource use to predict the regions in the sample images that exhibit related behaviors for the target resource use, which are then identified as the predicted activity regions of the sample images; next, the server constructs a loss function between the true activity regions and the predicted activity regions based on the true activity regions, the sample activity regions, and a preset adjustment factor, and determines the loss value between the true activity regions and the predicted activity regions based on the loss function, which is then used as the loss information between the true activity regions and the predicted activity regions; finally, if the loss value is greater than or equal to a preset loss threshold, the server retrains the borrowing resource monitoring model to be trained until the corresponding loss value is less than the preset loss threshold, thus obtaining the borrowing resource monitoring model corresponding to the target resource use.

[0104] For example, for agricultural activities, the server can use crop-enhanced images and agricultural behavior-enhanced images as sample images; for self-built housing activities, the server can use self-built housing construction images as sample images.

[0105] In this embodiment, the server can train the borrowed resource monitoring model corresponding to the target resource use by using sample images. The repeated training of the borrowed resource monitoring model based on the loss information between the real activity area and the predicted activity area of ​​the sample images can improve the target detection accuracy of the borrowed resource monitoring model, thereby improving the monitoring accuracy of the actual resource use of the borrowed resources.

[0106] In an exemplary embodiment, step S306 above, which determines the loss information between the real active region and the predicted active region based on the real active region, the predicted active region, and a preset adjustment factor, specifically includes the following: determining the distance information, area difference information, and size difference information between the real active region and the predicted active region; and performing fusion processing on the distance information, area difference information, and size difference information based on the preset adjustment factor to obtain the loss information between the real active region and the predicted active region.

[0107] The distance information between the actual activity region and the predicted activity region includes at least the center distance between the first center point of the actual activity region and the second center point of the predicted activity region, as well as the diagonal distance between the enclosing regions of the actual and predicted activity regions; wherein the enclosing region of the actual and predicted activity regions is the smallest region capable of encompassing both the actual and predicted activity regions. Figure 4 The diagram shows the center distance and diagonal distance.

[0108] The area difference information between the actual activity area and the predicted activity area is used to characterize the difference between the intersection and union regions of the actual and predicted activity areas. This area difference information can be represented by the ratio between the area of ​​the intersection region and the area of ​​the union region. Figure 5(a) shows a schematic diagram of the intersection region between the actual and predicted activity areas; Figure 5(b) shows a schematic diagram of the union region between the actual and predicted activity areas.

[0109] Among them, the size difference information between the actual active region and the predicted active region is used to characterize the difference between the width-to-length ratio of the actual active region and the width-to-length ratio of the predicted active region.

[0110] Specifically, based on the actual activity area and the predicted activity area, the server first determines the first center point of the actual activity area and the second center point of the predicted activity area, thereby determining the center distance between the first center point and the second center point. Next, the server determines the smallest area that can enclose the actual activity area and the predicted activity area as the corresponding enclosing area, and uses the distance of the diagonal of the enclosing area as the diagonal distance between the actual activity area and the predicted activity area. Then, the server identifies the intersection and union areas between the actual activity area and the predicted activity area, and determines the area difference information between the actual activity area and the predicted activity area based on the ratio between the area of ​​the intersection area and the area of ​​the union area. Next, the server determines the size difference information between the actual activity area and the predicted activity area based on the aspect ratio of the actual activity area and the aspect ratio of the predicted activity area. Finally, the server fuses the distance information, area difference information, and size difference information according to a preset adjustment factor to obtain the loss information between the actual activity area and the predicted activity area.

[0111] For example, as shown in Formula 1, the loss function is used to determine the loss information between the actual activity area and the predicted activity area:

[0112]

[0113] Where L represents the loss information; A represents the actual activity region, a represents the first center point of the actual activity region; B represents the predicted activity region, b represents the second center point of the predicted activity region; IoU(A,B) represents the area difference information between the actual activity region and the predicted activity region; d(a,b) represents the center distance between the first center point and the second center point; ρ(A,B) represents the diagonal distance between the enclosing regions corresponding to the actual activity region and the predicted activity region; (βν) represents the size difference information between the actual activity region and the predicted activity region; α represents the preset adjustment factor, i.e., the adjustment power, and the optimal value of α is 3 after multiple experiments.

[0114] The area difference information IoU(A,B) in Formula 1 is obtained through Formula 2 as follows:

[0115]

[0116] Among them, S A∩B S represents the area of ​​the intersection region between the actual activity region and the predicted activity region. A∪B This represents the area of ​​the union region between the actual activity region and the predicted activity region.

[0117] The dimensional difference information (βν) in Formula 1 is obtained through Formula 3 as follows:

[0118]

[0119] Among them, w A and h A These represent the width and length of the actual activity area, respectively; w B and h B These represent the width and length of the predicted activity area, respectively.

[0120] In this embodiment, the server can obtain distance information, area difference information, and size difference information between the real activity area and the predicted activity area. Based on a preset adjustment factor, the server fuses the above-mentioned distance information, area difference information, and size difference information to obtain loss information between the real activity area and the predicted activity area. Based on the loss function shown in Formula 1, the server can improve the target detection accuracy of the borrowed resource monitoring model, thereby improving the monitoring accuracy of the actual resource usage of the borrowed resources.

[0121] It should be noted that, through verification, it was found that the borrowing resource monitoring model trained based on the loss function shown in Equation 1 has significantly improved precision (the proportion of images that actually contain the associated behavior of the target resource among the images whose detection result is "there is an associated behavior of the target resource") and data processing speed compared with the borrowing resource monitoring model trained based on the loss function determined by the existing GIoU_LOSS (Generalized Intersection Over Union-LOSS) algorithm. Therefore, the borrowing resource monitoring model trained based on the loss function shown in Equation 1 can perform target detection with higher accuracy and faster processing speed.

[0122] In an exemplary embodiment, determining the distance information between the real active region and the predicted active region specifically includes the following: determining the enclosing regions corresponding to the real active region and the predicted active region; determining the diagonal distance between the real active region and the predicted active region based on the enclosing regions; determining the distance between the first center point of the real active region and the second center point of the predicted active region as the center distance between the real active region and the predicted active region; and determining the diagonal distance and the center distance as the distance information between the real active region and the predicted active region.

[0123] Specifically, the server first determines the smallest region that can enclose the actual and predicted activity regions based on the actual and predicted activity regions, and identifies this smallest region as the enclosing region corresponding to the actual and predicted activity regions. The diagonal distance of this enclosing region is then identified as the diagonal distance between the actual and predicted activity regions. Next, the server determines the first center coordinates of the first center point of the actual activity region and the second center coordinates of the second center point of the predicted activity region. Based on the first and second center coordinates, the server determines the distance between the first and second center points as the center distance between the actual and predicted activity regions. Finally, the server identifies the diagonal distance and the center distance as the distance information between the actual and predicted activity regions.

[0124] In this embodiment, the center distance and diagonal distance between the actual activity area and the predicted activity area of ​​the server are used to measure the distance between the actual activity area and the predicted activity area from multiple dimensions, thereby measuring the difference between the actual activity area and the predicted activity area. This is helpful for determining loss information and training the borrowed resource monitoring model to be trained.

[0125] In an exemplary embodiment, step S208 above, based on the borrowed resource monitoring model deployed in the UAV, performs target detection processing on multiple images of the target location to obtain a detection result for each image. Specifically, it includes the following: for each of the multiple images of the target location, the borrowed resource monitoring model deployed in the UAV identifies an active area in the image that is associated with the use of the target resource, and determines the confidence information of the active area; based on the confidence information of the active area, the detection result of the image is determined.

[0126] Among them, the confidence information of the activity area is used to characterize the confidence level of the existence of associated behaviors for the purpose of the target resource in the activity area.

[0127] Specifically, for each of the multiple images of the target location, the server uses a resource borrowing monitoring model deployed in the drone to identify the activity area in the image that is associated with the use of the target resource, and determines the confidence level of the activity area that is associated with the use of the target resource. Then, if the corresponding confidence level is greater than a preset confidence threshold, the server confirms that the activity area is associated with the use of the target resource, and thus determines that the image contains the associated behavior of the use of the target resource.

[0128] For example, if the target resource is used for agricultural activities, and assuming a pre-set confidence threshold of 50%, then if the confidence of the active area in the image is greater than 50%, the server will mark the image as "having agricultural activity," and if the confidence of the active area in the image is less than or equal to 50%, the server will mark the image as "not having agricultural activity."

[0129] In this embodiment, by comparing the confidence information of the active area with the preset confidence threshold, the server can further confirm whether there is any associated behavior of the target resource use in the active area, reduce the impact of detection error on the monitoring results, and thus further improve the monitoring accuracy of the actual resource use of borrowed resources.

[0130] In an exemplary embodiment, step S210, which determines the monitoring result of the borrowed resources corresponding to the resource borrowing information based on the detection result of each image returned by the UAV, specifically includes the following: based on the detection result of each image, identifying the target image with associated behavior of the target resource use among multiple images; and confirming the monitoring result of the borrowed resources corresponding to the resource borrowing information as qualified if the number of target images meets the preset image number condition.

[0131] The preset image quantity condition can be either a preset target image quantity threshold or a preset target image quantity ratio threshold; the target image quantity ratio is the ratio between the number of target images and the total number of images.

[0132] Specifically, based on the detection results of each image, the server identifies images among multiple images that exhibit related behaviors for the purpose of the target resource, and designates them as target images. The server then counts the number of target images. If the number of target images exceeds a preset image count threshold, or if the ratio between the number of target images and the total number of images exceeds a preset ratio, the server confirms that there is a related behavior for the purpose of the target resource at the target location. In other words, the actual use of the borrowed resource by the resource recipient is consistent with the target resource use (i.e., the designated use) of the borrowed resource. Therefore, the monitoring result of the borrowed resource corresponding to the resource borrowing information is confirmed as qualified.

[0133] In this embodiment, the server can further confirm whether there is any associated behavior of target resource use at the target location by the number of target images, thereby reducing the impact of detection errors on the monitoring results and further improving the monitoring accuracy of the actual resource use of borrowed resources.

[0134] In an exemplary embodiment, after determining the monitoring results of the borrowed resources corresponding to the resource borrowing information based on the detection results of each image returned by the UAV in step S210, the method further includes the following: determining the number of borrowed resources and the resource recipient corresponding to the resource borrowing information based on the resource borrowing information; and updating the borrowing credit information of the resource recipient based on the monitoring results and the number of borrowed resources corresponding to the resource borrowing information.

[0135] The quantity of borrowed resources refers to the numerical value of the borrowed resources, such as the loan amount.

[0136] Among them, the borrowing credit information is used to evaluate the resource borrowing request of the resource recipient and characterize the credit of the resource recipient, such as the credit information of the lender.

[0137] Specifically, after receiving the monitoring results of the borrowed resources corresponding to the resource borrowing information, the server first determines the number of borrowed resources and the resource recipient based on the resource borrowing information; then, based on the monitoring results and the number of borrowed resources, it updates the borrowing credit information of the resource recipient, so that the resource recipient can be evaluated when applying for resource borrowing in the future.

[0138] For example, assuming the monitoring result is satisfactory, meaning the resource recipient uses the borrowed resources for the intended purpose (specified purpose), the server can increase the resource recipient's borrowing credit score and increase the number of resources that can be borrowed to the resource provider in the future. Conversely, assuming the monitoring result is unsatisfactory, meaning the resource recipient does not use the borrowed resources for the intended purpose (specified purpose), the server can decrease the resource recipient's borrowing credit score and decrease the number of resources that can be borrowed to the resource provider in the future.

[0139] In this embodiment, the server updates the borrowing credit information of the resource recipient based on the monitoring results of the borrowed resources. This can effectively assess the creditworthiness and repayment ability of the resource recipient, thereby facilitating the determination of whether to continue providing borrowed resources to the resource recipient and determining the amount of borrowed resources that can be provided to the resource recipient.

[0140] In one exemplary embodiment, such as Figure 6 As shown, another method for monitoring borrowed resources is provided, which can be applied to applications such as... Figure 1 The following steps are used as an example of the server shown:

[0141] Step S601: Obtain a sample image corresponding to the target resource use, and the actual activity area in the sample image that contains the associated behavior of the target resource use.

[0142] Step S602: Using the borrowing resource monitoring model to be trained corresponding to the target resource use, predict the regions in the sample image that have associated behaviors of the target resource use to obtain the predicted activity regions.

[0143] Step S603: Determine the distance information, area difference information, and size difference information between the actual activity area and the predicted activity area.

[0144] Step S604: Based on a preset adjustment factor, the distance information, area difference information, and size difference information are fused to obtain the loss information between the actual activity area and the predicted activity area.

[0145] Step S605: Based on the loss information, train the borrowing resource monitoring model to be trained to obtain the borrowing resource monitoring model corresponding to the target resource usage.

[0146] Step S606: Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information.

[0147] Step S607: Determine the borrowed resource monitoring model corresponding to the target resource use, and deploy the borrowed resource monitoring model to the drone.

[0148] Step S608: Generate an image acquisition task including resource usage location information, send the image acquisition task to the UAV, and make the UAV fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location.

[0149] Step S609: For each of the multiple images of the target location, the borrowed resource monitoring model deployed in the UAV identifies the activity area in the image that is associated with the use of the target resource, determines the confidence information of the activity area, and determines the detection result of the image based on the confidence information of the activity area.

[0150] Step S610: Based on the detection results of each image, identify the target image with associated behavior of the target resource use among multiple images. If the number of target images meets the preset image number condition, confirm the monitoring result of the borrowed resource corresponding to the resource borrowing information as qualified.

[0151] Step S611: Based on the resource borrowing information, determine the number of borrowed resources and the recipient of the borrowed resources corresponding to the resource borrowing information.

[0152] Step S612: Update the borrowing credit information of the resource recipient based on the monitoring results and the number of borrowed resources corresponding to the resource borrowing information.

[0153] In this embodiment, firstly, the server, through the execution of image acquisition tasks by a drone, can remotely acquire multiple images of the target location corresponding to the resource usage location information. Based on the borrowed resource monitoring model deployed on the drone, it can remotely detect whether there are any behaviors associated with the target resource usage at the target location, thereby achieving remote monitoring of the actual resource usage of the borrowed resources. This avoids the time and manpower consumed by on-site inspections by staff and improves the monitoring efficiency of the actual resource usage of the borrowed resources. Secondly, based on a preset adjustment factor, the server fuses the distance information, area difference information, and size difference information between the actual activity area and the predicted activity area. The resulting loss information between the actual activity area and the predicted activity area can improve the target detection accuracy of the borrowed resource monitoring model. This facilitates subsequent training of the borrowed resource monitoring model corresponding to the target resource usage using sample images, thereby improving the monitoring accuracy of the actual resource usage of the borrowed resources. Furthermore, based on the monitoring results of borrowed resources, the server updates the borrowing credit information of the resource recipient, which can effectively assess the resource recipient's creditworthiness and repayment ability, thereby facilitating the determination of whether to continue providing borrowed resources to the resource recipient and determining the amount of borrowed resources that can be provided to the resource recipient.

[0154] To more clearly illustrate the resource borrowing monitoring method provided in the embodiments of this application, a specific embodiment is given below for detailed description of the resource borrowing monitoring method. However, it should be understood that the embodiments of this application are not limited thereto. In an exemplary embodiment, this application also provides a method for applying a YOLOv5 model based on an improved α-IoU (Intersection over Union) loss function algorithm in the field of UAV monitoring, specifically including the following steps:

[0155] 1. Select a suitable YOLOv5 model.

[0156] Considering the program capacity of small drones, the YOLOV5-s model, which has lower hardware requirements, was selected as the target monitoring model for drone monitoring.

[0157] Understandably, when the recognition rate is not high, the YOLOV5-m model can be considered as a target monitoring model for UAV monitoring.

[0158] 2. Improve the YOLOv5 model.

[0159] The commonly used loss function algorithm in the YOLOv5-s model is the GIoU_LOSS algorithm shown in Equation 4:

[0160]

[0161] Where C is the enclosing region corresponding to the actual activity region and the predicted activity region.

[0162] To better measure the difference between the real activity region and the predicted activity region, the server introduces a preset adjustment factor, as well as distance and size difference information between the real activity region and the predicted activity region, based on Equation 4, to obtain the improved loss function as shown in Equation 1.

[0163] 3. The YOLOV5-s model is trained based on the improved loss function.

[0164] The server first acquires sample images mainly depicting agricultural activities through the National Agricultural Science Data Center. Then, it labels the sample images as either "agricultural activities present" or "agricultural activities absent." Next, the server randomly selects 90% of the labeled sample images as the training set and the remaining 10% as the validation set. Finally, the server trains the YOLOV5-s model based on the training set and the improved loss function, and loads the trained YOLOV5-s model into a programmable drone.

[0165] After training the YOLOV5-s model, the parameters of the trained YOLOV5-s model can be verified using a validation set, such as the precision (the proportion of sample images predicted as "agricultural activity exists" among all sample images labeled "agricultural activity exists") and recall (the proportion of sample images labeled as "agricultural activity exists" among all sample images predicted as "agricultural activity exists").

[0166] 4. Generate drone operations.

[0167] The server generates drone operations based on the target resource use (agricultural activities) and resource usage location information recorded in the agricultural resource borrowing information, and distributes the drone operations to drones equipped with trained YOLOv5-s models.

[0168] 5. Perform drone operations.

[0169] After receiving the drone operation request, the drone flies to the target location corresponding to the resource usage location information, collects multiple images of the target location, and performs target detection processing on the collected multiple images using the loaded YOLOv5-s model that has been trained.

[0170] When the trained YOLOV5-s model detects suspected agricultural activities in an image, it uses bounding boxes and confidence scores to label the areas where agricultural activities are suspected in the image. If the confidence score is greater than 50%, the image is labeled as "agricultural activities exist"; otherwise, it is labeled as "no agricultural activities exist". Then, the drone returns the image with the identification result of "agricultural activities exist", the corresponding confidence score, and the identification result to the server.

[0171] 6. Monitor the actual use of agricultural borrowed resources.

[0172] The server determines whether the actual use of agricultural borrowed resources is agricultural activity by identifying the proportion of images with the result of "agricultural activity exists" among all images. That is, whether it is consistent with the target resource use specified for agricultural borrowed resources. When the proportion of images with the result of "agricultural activity exists" among all images exceeds 50%, it is determined that the actual resource use of the agricultural borrowed resources is consistent with the target resource use specified for agricultural borrowed resources.

[0173] 7. Update the secondment credit information of the resource recipient.

[0174] Based on the monitoring results from step 6, the server updates the credit information of the resource recipient, completing the monitoring loop.

[0175] In this embodiment, by validating the trained YOLOv5-s model using a validation set, the server can determine that, based on the improved loss function, the trained YOLOv5-s model improves precision by approximately 20% and data processing speed by 8% while sacrificing 2.8% recall. Assuming that before the loss function improvement, the model could process 100 frames per second of real agricultural activity images, with a recall of 71% and correctly identifying 71 images, the improved model can process 108 frames per second of real agricultural activity images, with a recall of 69% and correctly identifying 75 images. Furthermore, the server uses a drone equipped with the improved loss function object detection model. Through automated drone operation, the resource provider's investment in monitoring the actual use of borrowed resources can be significantly reduced, and the monitoring results can be used to update borrowing credit information, helping the resource provider assess the creditworthiness of the resource recipient.

[0176] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0177] Based on the same inventive concept, this application also provides a resource transfer monitoring device for implementing the resource transfer monitoring method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more resource transfer monitoring device embodiments provided below can be found in the limitations of the resource transfer monitoring method described above, and will not be repeated here.

[0178] In one exemplary embodiment, such as Figure 7 As shown, a resource borrowing monitoring device is provided, comprising: an information determination module 702, a model deployment module 704, an image acquisition module 706, a target detection module 708, and a result determination module 710, wherein:

[0179] The information determination module 702 is used to determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information based on the resource borrowing information.

[0180] The model deployment module 704 is used to determine the borrowed resource monitoring model corresponding to the target resource use and deploy the borrowed resource monitoring model to the UAV.

[0181] The image acquisition module 706 is used to generate an image acquisition task that includes resource usage location information, send the image acquisition task to the UAV, and enable the UAV to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location.

[0182] The target detection module 708 is used to perform target detection processing on multiple images of the target location based on the borrowed resource monitoring model deployed in the UAV, and obtain the detection result for each image; the detection result is used to characterize the correlation between the image and the use of the target resource.

[0183] The result determination module 710 is used to determine the monitoring results of the borrowed resources corresponding to the resource borrowing information based on the detection results of each image returned by the UAV.

[0184] In an exemplary embodiment, the borrowing resource monitoring device further includes a model training module, used to acquire sample images corresponding to the target resource usage and real activity regions in the sample images that contain related behaviors of the target resource usage; predict the regions in the sample images that contain related behaviors of the target resource usage using a borrowing resource monitoring model to be trained corresponding to the target resource usage, to obtain predicted activity regions; determine loss information between the real activity regions and the predicted activity regions based on the real activity regions, the predicted activity regions, and a preset adjustment factor; and train the borrowing resource monitoring model to be trained based on the loss information to obtain a borrowing resource monitoring model corresponding to the target resource usage.

[0185] In an exemplary embodiment, the model training module is further configured to determine distance information, area difference information, and size difference information between the real active region and the predicted active region; and to fuse the distance information, area difference information, and size difference information according to a preset adjustment factor to obtain loss information between the real active region and the predicted active region.

[0186] In an exemplary embodiment, the model training module is further configured to determine the enclosing regions corresponding to the real active region and the predicted active region, and based on the enclosing regions, determine the diagonal distance between the real active region and the predicted active region; determine the distance between the first center point of the real active region and the second center point of the predicted active region as the center distance between the real active region and the predicted active region; and determine the diagonal distance and the center distance as distance information between the real active region and the predicted active region.

[0187] In an exemplary embodiment, the target detection module 708 is further configured to, for each of multiple images of the target location, identify an active area in the image where there is associated behavior of target resource use through a borrowed resource monitoring model deployed in the UAV, and determine confidence information of the active area; and determine the detection result of the image based on the confidence information of the active area.

[0188] In an exemplary embodiment, the result determination module 710 is further configured to, based on the detection results of each image, determine, among multiple images, a target image that exhibits associated behavior related to the use of the target resource; and, if the number of target images meets a preset image number condition, confirm the monitoring result of the borrowed resource corresponding to the resource borrowing information as qualified.

[0189] In an exemplary embodiment, the resource borrowing monitoring device further includes an information update module, which is used to determine the number of borrowed resources and the resource recipient corresponding to the resource borrowing information based on the resource borrowing information; and to update the borrowing credit information of the resource recipient according to the monitoring results and the number of borrowed resources corresponding to the resource borrowing information; the borrowing credit information is used to evaluate the resource recipient's resource borrowing request.

[0190] Each module in the aforementioned resource borrowing monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0191] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores data such as resource borrowing information and the borrowing credit information of resource recipients. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a resource borrowing monitoring method.

[0192] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0193] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0194] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.

[0195] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0196] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0197] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0198] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method of monitoring borrowed resources, characterized by, The method includes: Based on the resource borrowing information, determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information; Determine the resource borrowing monitoring model corresponding to the target resource use, and deploy the resource borrowing monitoring model to the drone; Generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location; Based on the resource borrowing monitoring model deployed in the UAV, target detection processing is performed on multiple images of the target location to obtain detection results for each image; the detection results are used to characterize the association between the image and the purpose of the target resource. Based on the detection results of each image returned by the drone, the monitoring results of the borrowed resources corresponding to the resource borrowing information are determined; The resource borrowing monitoring model corresponding to the target resource usage is trained in the following manner: Obtain a sample image corresponding to the intended use of the target resource, and the actual activity area in the sample image where there is a related behavior of the intended use of the target resource; By using the borrowing resource monitoring model to be trained corresponding to the target resource use, the region in the sample image where there is a related behavior of the target resource use is predicted to obtain the predicted activity region. Based on the actual activity area, the predicted activity area, and a preset adjustment factor, the loss information between the actual activity area and the predicted activity area is determined; Based on the loss information, the borrowing resource monitoring model to be trained is trained to obtain the borrowing resource monitoring model corresponding to the target resource use.

2. The method of claim 1, wherein, The step of determining the loss information between the actual activity region and the predicted activity region based on the actual activity region, the predicted activity region, and a preset adjustment factor includes: Determine the distance information, area difference information, and size difference information between the actual activity area and the predicted activity area; Based on a preset adjustment factor, the distance information, the area difference information, and the size difference information are fused to obtain loss information between the actual activity area and the predicted activity area.

3. The method of claim 2, wherein, Determining the distance information between the actual activity area and the predicted activity area includes: Determine the enclosing regions corresponding to the actual activity region and the predicted activity region, and based on the enclosing regions, determine the diagonal distance between the actual activity region and the predicted activity region; The distance between the first center point of the actual activity area and the second center point of the predicted activity area is determined as the center distance between the actual activity area and the predicted activity area. The diagonal distance and the center distance are determined as the distance information between the actual activity area and the predicted activity area.

4. The method of claim 1, wherein, The resource borrowing monitoring model deployed in the UAV performs target detection processing on multiple images of the target location to obtain detection results for each image, including: For each of the multiple images of the target location, the borrowed resource monitoring model deployed in the UAV identifies the activity area in the image that is associated with the use of the target resource, and determines the confidence information of the activity area. The detection result of the image is determined based on the confidence information of the active region.

5. The method of claim 1, wherein, The step of determining the monitoring results of the borrowed resources corresponding to the resource borrowing information based on the detection results of each image returned by the drone includes: Based on the detection results of each image, target images that exhibit associated behaviors related to the use of the target resource are identified among the multiple images. If the number of images in the target image meets the preset image quantity condition, the monitoring result of the borrowed resources corresponding to the resource borrowing information is confirmed as qualified.

6. The method according to any one of claims 1 to 5, characterized in that, After determining the monitoring results of the borrowed resources corresponding to the resource borrowing information based on the detection results of each image returned by the drone, the method further includes: Based on the resource borrowing information, determine the quantity of borrowed resources and the recipient of the resources corresponding to the resource borrowing information; Based on the monitoring results of the borrowed resources corresponding to the resource borrowing information and the quantity of borrowed resources, the borrowing credit information of the resource recipient is updated; the borrowing credit information is used to evaluate the resource borrowing request of the resource recipient.

7. A borrowed resource monitoring apparatus characterized by comprising: The device includes: The information determination module is used to determine the target resource purpose and resource usage location information of the borrowed resources corresponding to the resource borrowing information based on the resource borrowing information; The model deployment module is used to determine the borrowed resource monitoring model corresponding to the target resource use and deploy the borrowed resource monitoring model to the UAV; An image acquisition module is used to generate an image acquisition task including the resource usage location information, send the image acquisition task to the drone, and cause the drone to fly to the target location corresponding to the resource usage location information and acquire multiple images of the target location; The target detection module is used to perform target detection processing on multiple images of the target location based on the borrowed resource monitoring model deployed in the UAV, and obtain the detection result for each image; the detection result is used to characterize the association between the image and the purpose of the target resource; The result determination module is used to determine the monitoring result of the borrowed resources corresponding to the resource borrowing information based on the detection result of each image returned by the UAV. The device further includes a model training module, used to acquire sample images corresponding to the target resource use and real activity regions in the sample images that contain related behaviors of the target resource use; predict the regions in the sample images that contain related behaviors of the target resource use using a borrowing resource monitoring model to be trained corresponding to the target resource use, to obtain predicted activity regions; determine the loss information between the real activity regions and the predicted activity regions based on the real activity regions, the predicted activity regions, and a preset adjustment factor; and train the borrowing resource monitoring model to be trained based on the loss information to obtain the borrowing resource monitoring model corresponding to the target resource use.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images