Target detection method, device and equipment for unmanned aerial vehicle swarm cooperation and medium

Abstract

This application discloses a method, apparatus, device, and medium for target detection using a swarm of unmanned aerial vehicles (UAVs). Applied to a first UAV, when a first image is acquired from a first sub-region of a target area within a preset detection period, the first image is added to an image detection queue. When the image detection queue contains several images to be detected, a collaborative decision is made regarding the current processing progress of the first image to be detected, combining the current initial remaining time within the preset detection period with a preset image detection time. When a decision requests other second UAVs to collaboratively compute the first image to be detected, a target second UAV with the largest communication link bandwidth is selected from among the multiple second UAVs. The first image to be detected is then sent to the target second UAV corresponding to the target communication link for collaborative computation. In this way, the UAV swarm can collaboratively perform recognition and computation on the acquired images, improving the target detection efficiency and real-time performance of the UAV swarm.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, device and medium for target detection in a swarm of unmanned aerial vehicles (UAVs). Background Technology

[0002] With the rapid development of unmanned aerial vehicle (UAV) technology, UAVs have been applied in various industries, such as security, fire protection, and surveying. However, a single UAV may not be able to meet the needs of some complex scenarios. For example, for complex operations such as surveying multiple regions or large-scale areas and target detection, multiple UAVs need to work together to improve detection efficiency.

[0003] The relevant technology uses multiple drones to collect images of a specific area and transmits the image data to a server. This avoids the need for a single drone to spend a lot of time collecting images of a specific area, saving collection time. Then, the server uses the images to perform target recognition to detect targets within the specific area.

[0004] However, while the related technology can save time by having multiple drones collect images of a specific area, it only serves to collect image information. Furthermore, since drones themselves lack computing power, they need to rely on cloud computing power to complete target detection, which results in time delays and a lack of real-time performance. Summary of the Invention

[0005] This application provides a target detection method, apparatus, device, and medium for UAV swarm collaboration, which can collaboratively perform recognition and calculation on acquired images by UAV swarms, thereby improving the target detection efficiency and real-time performance of UAV swarms.

[0006] In a first aspect, the present application provides a target detection method for a collaborative unmanned aerial vehicle (UAV) swarm, applied to a first UAV in a UAV swarm, wherein the UAV swarm further includes multiple second UAVs, including: According to a preset detection cycle, a first image is acquired for the first sub-region within the target range area, and the first image is added to the image detection queue. The multiple second UAVs are used to perform target detection on other sub-regions within the target range area other than the first sub-region. When it is detected that the image detection queue contains at least one image to be detected, the initial remaining detection time within the preset detection period is determined, and the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue is determined according to the preset image detection time and the initial remaining detection time. When the collaborative decision identifier is detected as an unload decision identifier, the communication link bandwidth between the local first UAV and each second UAV and the total communication bandwidth corresponding to the multiple communication links are obtained, and the target communication link is selected according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The first image to be detected is sent to the target second UAV corresponding to the target communication link, so that the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server.

[0007] Secondly, the present application provides a target detection device for drone swarm collaboration, applied to a first drone in a drone swarm, the drone swarm further including multiple second drones, including: The acquisition unit is used to acquire a first image for a first sub-region within the target range according to a preset detection cycle, and add the first image to the image detection queue. The plurality of second UAVs are used to perform target detection on other sub-regions within the target range except for the first sub-region. The determining unit is configured to, when detecting that the image detection queue contains at least one image to be detected, determine the current initial remaining detection time within the preset detection period, and determine the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue based on the preset image detection time and the initial remaining detection time; The communication scheduling unit is used to, when the cooperative decision identifier is detected as an unload decision identifier, obtain the communication link bandwidth of the communication link between the local first UAV and each second UAV and the total communication bandwidth corresponding to the multiple communication links, and select the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth; The collaborative request unit is used to send the first image to be detected to the target second UAV corresponding to the target communication link, so that the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server.

[0008] In some embodiments, the determining unit is further configured to: Based on the initial remaining detection time and the preset image detection time, the local processing capacity is estimated; Based on the local processing capacity, determine the local processing probability and decision scheduling probability of the first image to be detected in the image detection queue with the current processing progress. The collaborative decision identifier of the first image to be detected is randomly determined according to the local processing probability and the decision scheduling probability.

[0009] In some embodiments, the determining unit is further configured to: When the local processing capacity is greater than or equal to a preset processing threshold, the first preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue, and the decision scheduling probability is calculated based on the local processing probability, wherein the sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is greater than the decision scheduling probability. When the local processing capacity is less than the preset processing capacity threshold, the second preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue, and the decision scheduling probability is calculated based on the local processing probability, wherein the sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is less than the decision scheduling probability.

[0010] In some embodiments, the communication scheduling unit is further configured to: Based on the ratio between the bandwidth of each communication link and the total communication bandwidth, a communication scheduling score is determined for the communication link between the local first UAV and each second UAV. According to the communication scheduling score in descending order, the multiple communication links are sorted to obtain the link sorting relationship; According to the link sorting relationship, the target communication link that is ranked first among the multiple communication links is selected.

[0011] In some embodiments, the target detection device for UAV swarm collaboration further includes a loop execution unit for: Acquire the first data volume of the first image to be detected, and estimate the target transmission time of the first image to be detected from the local first UAV to the target second UAV based on the first data volume; Obtain the actual processing time of the local first UAV when processing the second image to be detected in the image detection queue; Obtain the time it takes for the local first drone to transmit the current detection results back to the target edge server; Select the target unit occupancy time from the target transmission time, the actual processing time, and the result return time; The target remaining detection time within the preset detection period is determined based on the initial remaining detection time and the target unit occupancy time. When the remaining detection time of the target is greater than or equal to the preset time threshold, return to the step of acquiring the first image for the first sub-region within the target range area.

[0012] In some embodiments, the target detection device for UAV swarm collaboration further includes a collaborative decision-making unit, used for: Determine the first data volume of the first image to be detected, and obtain the target transmission bandwidth between the local first UAV and the target second UAV; Based on the first data volume and the target transmission bandwidth, the estimated transmission time of the first image to be detected to the target second UAV is estimated. The collaborative request unit is further configured to send the first image to be detected to the target second UAV corresponding to the target communication link when the estimated transmission duration is detected to be greater than or equal to the initial remaining detection duration.

[0013] In some implementations, the drone swarm cooperative target detection model further includes a cooperative computing unit for: Receive the target image to be detected sent by any second UAV; The target image to be detected is input into a pre-trained target detection model to obtain the corresponding collaborative target detection results; The collaborative target detection results are sent to the target edge server, so that the target edge server sorts the collaborative target detection results with at least one historical target detection result corresponding to the second sub-region to obtain a target detection result sequence.

[0014] Thirdly, this application also provides a target detection system for a swarm of unmanned aerial vehicles (UAVs) working together, the target detection system comprising a first UAV and multiple second UAVs, including: The first UAV acquires a first image of a first sub-region within the target range according to a preset detection cycle, and adds the first image to an image detection queue. The plurality of second UAVs are used to perform target detection on other sub-regions within the target range besides the first sub-region. When the first UAV detects that the image detection queue contains at least one image to be detected, it determines the current initial remaining detection time within the preset detection period, and determines the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue based on the preset image detection time and the initial remaining detection time. When the first UAV detects that the collaborative decision identifier is an unload decision identifier, it obtains the communication link bandwidth of the communication link between the local first UAV and each second UAV, as well as the total communication bandwidth corresponding to the multiple communication links, and selects the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The first UAV sends the first image to be detected to the target second UAV corresponding to the target communication link; The target second UAV performs target detection on the first image to be detected using a pre-trained target detection model, obtains the target detection result, and sends the target detection result to the target edge server; The target edge server receives the target detection result and sorts the target detection result with at least one historical target detection result corresponding to the first sub-region to obtain a target detection result sequence. The historical target detection result is the detection result sent by the first UAV and / or any one of the second UAVs for a historical first image within the first sub-region.

[0015] Furthermore, this application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned target detection method for UAV swarm collaboration.

[0016] Furthermore, embodiments of this application also provide a computer-readable storage medium storing multiple instructions adapted for loading by a processor to execute the aforementioned UAV swarm cooperative target detection method.

[0017] The target detection method for drone swarm collaboration provided in this application embodiment is applied to a first drone in a drone swarm, which also includes multiple second drones. Each drone has computing power. Each drone in the drone swarm can collect corresponding images in different sub-regions within the target range area. Taking the first drone as an example, the first drone can collect a first image corresponding to a first sub-region according to a preset detection cycle and add the first image to the image detection queue. When the image detection queue contains multiple images to be detected, the initial remaining time within the preset detection cycle is determined. Based on the preset image detection time and the initial remaining detection time, the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue is determined. When the collaborative decision identifier is an unloading decision identifier, the communication link bandwidth between the local first drone and each second drone and the total communication bandwidth corresponding to multiple communication links are obtained. According to the ratio between the bandwidth of each communication link and the total communication bandwidth, a target communication link is selected from multiple communication links. Finally, the first image to be detected is sent to the target second drone corresponding to the target communication link, so that the target second drone sends the target detection result corresponding to the first image to be detected to the target edge server.

[0018] Therefore, when the first UAV of this application acquires a first image of a first sub-region of the target area within a preset detection period, it adds the first image to the image detection queue. When the image detection queue contains a number of images to be detected, it makes a collaborative decision on the first image to be detected with the current processing progress by combining the current initial remaining time within the preset detection period and the preset image detection time. When the decision requests other second UAVs to perform collaborative calculation on the first image to be detected, it selects the target second UAV with the largest communication link bandwidth from among the multiple second UAVs, and sends the first image to be detected to the target second UAV corresponding to the target communication link for collaborative calculation. In this way, the individual UAVs in the UAV swarm can share computing resources, and the UAV swarm can collaboratively perform recognition calculation on the acquired images, thereby improving the target detection efficiency and real-time performance of the UAV swarm. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a target detection system for drone swarm collaboration provided in an embodiment of this application. Figure 2 A flowchart illustrating the steps of the target detection method for UAV swarm collaboration provided in this application embodiment; Figure 3 Example diagram of task division and trajectory planning scenario for target area provided in the embodiments of this application; Figure 4 This is an example diagram of the target detection system architecture for UAV swarm collaboration provided in an embodiment of this application; Figure 5 An example diagram illustrating a target detection processing scenario provided in an embodiment of this application for drone swarm collaboration; Figure 6 This is a schematic diagram of the structure of the target detection device for UAV swarm collaboration provided in the embodiments of this application; Figure 7 This is a schematic diagram of the terminal structure provided in the embodiments of this application; Figure 8 This is a schematic diagram of the server structure provided in an embodiment of this application. Detailed Implementation

[0021] To enable those skilled in the art to better understand the solutions of this application, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] It is understood that in the specific implementation of this application, data such as preset detection cycle, first image, initial remaining detection time, first image to be detected, collaborative decision identifier, and communication link bandwidth are involved. When the above embodiments of this application are applied to specific products or technologies, permission or consent from the target is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0023] Furthermore, when this application embodiment needs to obtain relevant data, it will obtain separate permission or separate consent for data such as the preset detection period, the first image, the initial remaining detection time, the first image to be detected, the collaborative decision identifier, and the communication link bandwidth through pop-up windows or redirection to a confirmation page. After clearly obtaining separate permission or separate consent for the relevant data such as the preset detection period, the first image, the initial remaining detection time, the first image to be detected, the collaborative decision identifier, and the communication link bandwidth, it will then obtain the necessary data for enabling the application embodiment to operate normally.

[0024] It should be noted that while some processes described in the specification, claims, and accompanying drawings contain multiple steps that appear in a specific order, it should be clearly understood that these steps may not be performed in the order they appear herein, or may be performed in parallel. The step numbers are merely used to distinguish different steps and do not represent any particular order of execution. Furthermore, descriptions such as "first," "second," or "objective" in this document are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] This application provides a method, apparatus, device, and medium for target detection in a swarm-based manner. Specifically, the swarm-based target detection method of this application can be implemented in a computer device, which can be a user terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The user terminal device can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, smart home appliance, vehicle terminal, smart voice interaction device, aircraft, drone, etc., but is not limited to these.

[0027] For ease of understanding, this application will describe the implementation process of the UAV swarm cooperative target detection method through multiple embodiments, as follows: This application provides a method for collaborative target detection in a drone swarm. The method is applied to a first drone in a swarm, which also includes multiple second drones. Each drone has computing power. Each drone in the swarm can collect corresponding images within different sub-regions of a target area. Taking the first drone as an example, the first drone can collect a first image corresponding to a first sub-region according to a preset detection cycle and add the first image to an image detection queue. When the image detection queue contains multiple images to be detected, the initial remaining time within the preset detection cycle is determined. Based on the preset image detection time and the initial remaining detection time, a collaborative decision identifier for the first image to be detected with the current processing progress in the image detection queue is determined. When the collaborative decision identifier is an unloading decision identifier, the communication link bandwidth between the local first drone and each second drone, as well as the total communication bandwidth corresponding to multiple communication links, is obtained. A target communication link is selected from the multiple communication links according to the ratio between the bandwidth of each communication link and the total communication bandwidth. Finally, the first image to be detected is sent to the target second drone corresponding to the target communication link, so that the target second drone sends the target detection result corresponding to the first image to be detected to a target edge server. Please refer to the following specific embodiments for details.

[0028] It should be noted that this drone swarm collaborative target detection method can be executed by the terminal alone, or by the terminal and the server together.

[0029] For example, taking the target detection method of UAV swarm collaboration jointly executed by the terminal and the server as an example, the UAV swarm collaboration target detection system may include terminal 110 and server 120.

[0030] Taking terminal 110 as the first drone in a drone swarm as an example, the drone swarm also includes multiple other second drones. When the first drone executes the drone swarm collaborative target detection method, it can collect a first image for a first sub-region within the target range according to a preset detection cycle and add the first image to the image detection queue. Multiple second drones are used to perform target detection on other sub-regions within the target range, excluding the first sub-region. When at least one image to be detected is detected in the image detection queue, the initial remaining detection time within the preset detection cycle is determined, and the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue is determined based on the preset image detection time and the initial remaining detection time. When the collaborative decision identifier is detected as an unloading decision identifier, the communication link bandwidth between the local first drone and each second drone and the total communication bandwidth corresponding to multiple communication links are obtained. The target communication link is selected according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The first image to be detected is sent to the target second drone corresponding to the target communication link, so that the target second drone sends the target detection result corresponding to the first image to be detected to the target edge server.

[0031] The server 120 can be a single service node, a distributed system composed of multiple service nodes, or a service node within a distributed system. For example, taking the server 120 as a target edge server, the target edge server receives the target detection results and sorts them with at least one historical target detection result corresponding to the first sub-region to obtain a target detection result sequence. The historical target detection results are the detection results sent by the first UAV and / or any second UAV to the historical first image within the first sub-region.

[0032] Furthermore, to better understand the above embodiments, the following description uses a target detection system based on a swarm of unmanned aerial vehicles (UAVs) as an example. This target detection system includes a first UAV and multiple second UAVs, including: The first UAV collects a first image of the first sub-region within the target range according to a preset detection cycle, and adds the first image to the image detection queue. Multiple second UAVs are used to perform target detection on other sub-regions within the target range, excluding the first sub-region. When the first UAV detects that the image detection queue contains at least one image to be detected, it determines the current initial remaining detection time within the preset detection period, and determines the collaborative decision identifier of the first image to be detected in the image detection queue with the current processing progress based on the preset image detection time and the initial remaining detection time. When the first UAV detects that the collaborative decision identifier is the unload decision identifier, it obtains the communication link bandwidth between the local first UAV and each second UAV, as well as the total communication bandwidth corresponding to multiple communication links, and selects the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The first UAV sends the first image to be detected to the target second UAV corresponding to the target communication link; The second UAV uses a pre-trained target detection model to perform target detection on the first image to be detected, obtains the target detection result, and sends the target detection result to the target edge server; The target edge server receives the target detection results and sorts the target detection results with at least one historical target detection result corresponding to the first sub-region to obtain a target detection result sequence. The historical target detection results are the detection results sent by the first UAV and / or any second UAV to the historical first image in the first sub-region.

[0033] Therefore, when the first UAV of this application acquires a first image of a first sub-region of the target area within a preset detection period, it adds the first image to the image detection queue. When the image detection queue contains a number of images to be detected, it makes a collaborative decision on the first image to be detected with the current processing progress by combining the current initial remaining time within the preset detection period with the preset image detection time. When the decision requests other second UAVs to perform collaborative calculation on the first image to be detected, it selects the target second UAV with the largest communication link bandwidth from among the multiple second UAVs, and sends the first image to be detected to the target second UAV corresponding to the target communication link for collaborative calculation. In this way, the individual UAVs in the UAV swarm can share computing resources and perform recognition calculation on the acquired images collaboratively through the UAV swarm, thereby improving the target detection efficiency and real-time performance of the UAV swarm.

[0034] For ease of understanding, the steps of the UAV swarm cooperative target detection method will be described in detail below. It should be noted that the order of the following embodiments is not intended to limit the preferred order of the embodiments.

[0035] See Figure 2 , Figure 2 This is a flowchart illustrating the steps of a drone swarm collaborative target detection method provided in an embodiment of this application. In this embodiment, the drone swarm collaborative target detection method can be executed by a computer device, such as a server or terminal. For example, taking a computer device as a terminal device (such as a first drone) as an example, the specific process of executing the drone swarm collaborative target detection method is as follows: 101. According to the preset detection cycle, acquire the first image for the first sub-region within the target range area and add the first image to the image detection queue.

[0036] With the rapid development of unmanned aerial vehicle (UAV) technology, UAVs have been applied in various industries, such as security, fire protection, and surveying. However, a single UAV may not be able to meet the needs of some complex scenarios, such as UAV swarm performances and multi-regional surveys. For example, complex operations such as multi-regional or large-scale surveys and target detection require multiple UAVs to work together to improve detection efficiency.

[0037] However, related technologies utilize multiple drones to collaboratively collect images of a specific area and transmit the image data to a server. This avoids the significant time a single drone would need to spend collecting images of that area, saving time. The server then performs target recognition on the images to detect targets within the specific area. However, while this method saves time, it only serves the purpose of image information acquisition. Furthermore, because drones lack their own computing power and rely on cloud computing for target detection, target detection is delayed and lacks real-time capability.

[0038] To address the above issues, this application embodiment employs a drone swarm collaborative target detection method, enabling multiple drones to share computing resources. Specifically, taking a first drone as an example, when the first drone acquires a first image of a first sub-region of the target area within a preset detection period, it adds the first image to an image detection queue. Upon detecting that the image detection queue contains several images to be detected, it makes a collaborative decision regarding the current processing progress of the first image to be detected, combining the current initial remaining time within the preset detection period with the preset image detection time. When requesting other second drones to collaboratively compute the first image to be detected, it selects the target second drone with the largest communication link bandwidth from among the multiple second drones, and sends the first image to be detected to the target second drone corresponding to the target communication link for collaborative computation. This allows individual drones in the drone swarm to share computing resources, improving the target detection efficiency and real-time performance of the drone swarm by collaboratively performing recognition and computation on the acquired images.

[0039] In this embodiment, the drone swarm comprises multiple drones. A local terminal is designated as the first drone in the swarm, and the other drones in the swarm besides the first drone are designated as second drones. The first drone and the other multiple second drones jointly perform target detection on a target area. Specifically, the target area can be divided into multiple sub-regions according to the number of drones in the swarm. Each drone performs target detection on one sub-region. For example, the first drone performs target detection on the first sub-region, and the other second drones also perform target detection on one of their respective sub-regions. In other words, multiple second drones are used to perform target detection on the sub-regions within the target area other than the first sub-region. For example, taking the terminal as the first drone, the first drone can collect information from a first sub-region within the target area within a preset detection period to obtain a first image. This first image mainly includes image content of a certain location point or small location area within the first sub-region. Specifically, it can be a top-down view of a certain location point or small location area within the first sub-region. Furthermore, the collected first image is added to the image detection queue to wait for the first drone to perform target detection operations. In this way, through the division of labor and cooperation among multiple drones in the drone swarm (such as the first drone and other second drones), image information is collected from each sub-region within the target area, so as to carry out subsequent collaborative target detection, thereby improving the target detection efficiency of the drone swarm for the target area.

[0040] Multiple second drones are used to detect targets in sub-regions other than the first sub-region within the target area. This target area can be the total detection range of the drone swarm, specifically including coastal defense lines, beaches, forests, deserts, residential areas, industrial areas, etc., without limitation. Sub-regions refer to areas within the target area. The target area can be divided into a corresponding number of sub-regions based on the number of drones in the swarm, with each drone responsible for image acquisition and detection in one sub-region.

[0041] The preset detection cycle can be a pre-set unit detection cycle for the first UAV during the detection process of the first sub-region. Specifically, multiple unit detection cycles can be used to perform target detection on the first sub-region. For example, to perform full-range target detection on the first sub-region, the first sub-region can be divided into multiple row regions according to the unit shooting area of ​​the UAV's camera components, using a horizontal flight and row-by-row traversal method. Each row region consists of multiple unit shooting areas, and the position corresponding to each unit shooting area is regarded as a hovering point. Each row region corresponds to one unit detection cycle, which can be determined based on the sum of the total flight time of the UAV and the hovering shooting time of multiple unit shooting areas in a row region. Therefore, taking the first UAV as an example, when the first UAV performs target detection on the first sub-region, it needs to go through multiple preset detection cycles, specifically depending on the actual number of rows in the first sub-region.

[0042] The image detection queue can be a queue for storing the first images acquired by the first UAV. This queue can store multiple first images, specifically ordered according to their acquisition time sequence. Subsequently, target detection can be performed on the first-acquired first image using a first-in, first-out (FIFO) approach. It should be noted that target detection can be performed locally on the first UAV or by sending the first image to any other target second UAV to request collaborative target detection, demonstrating resource sharing among multiple UAVs in a swarm. It should also be noted that other second UAVs also have their own image detection queues, as described above.

[0043] In this way, the first local UAV can collect information on the first sub-region within the target range within a preset detection period and add the collected first image to the image detection queue, waiting for the first UAV to perform target detection operations. In this way, multiple UAVs in the UAV swarm (such as the first UAV and other second UAVs) cooperate to collect image information on each sub-region within the target range, so as to carry out subsequent collaborative target detection, thereby improving the target detection efficiency of the UAV swarm for the target range area.

[0044] 102. When it is detected that the image detection queue contains at least one image to be detected, determine the current initial remaining detection time within the preset detection period, and determine the collaborative decision identifier of the first image to be detected in the image detection queue with the current processing progress based on the preset image detection time and the initial remaining detection time.

[0045] In this embodiment, after adding the acquired first image to the image detection queue, the local first UAV can detect the number of images to be detected in the image detection queue. The image to be detected refers to the first image in the image detection queue. When it is detected that the image detection queue contains at least one image to be detected, the initial remaining detection time in the prediction detection period is first determined. The preset image detection time and the initial remaining detection time are used as decision factors to determine the collaborative decision identifier of the first image to be detected in the image detection queue with the current processing progress. Thus, for each first image in the local image detection queue, a collaborative decision identifier is generated by combining the preset image detection time and the remaining time in the image detection period. This determines whether to request other second UAVs in the UAV swarm to perform target detection on the first image to be detected, thereby realizing the sharing of computing resources in the UAV swarm and improving the target detection efficiency of the UAV swarm for the target range area.

[0046] The image to be detected refers to the first image added to the image detection queue by the local first drone after being collected at a historical time, which can be understood as the first historical image.

[0047] The initial remaining detection time can be the remaining time within a preset detection period. For example, if a preset detection period is 10 seconds and 6 seconds have elapsed in the current preset detection period, then the initial remaining detection time is 4 seconds. Or, if 1 second has elapsed in the current preset detection period, then the initial remaining detection time is 9 seconds. It should be noted that the initial remaining detection time is calculated in real time.

[0048] The preset image detection time can be the average recognition time for a single image, i.e., the average time spent by the local first UAV on target detection of the acquired image, or it can be the average time spent by any UAV in the UAV swarm on target detection of the acquired image. This preset image detection time can be determined based on the historical average target detection time. It should be noted that this preset image detection time can be the average detection time of the local first UAV on the image, or it can be the average detection time of multiple UAVs in the UAV swarm on the image. That is, the preset image detection time of the first UAV may differ from the preset detection times of other second UAVs, or it may be the same as the preset image detection times of the first UAV and other second UAVs. It should also be noted that this preset image detection time can be equal to the hovering and shooting time of the UAV at each hovering point in the row area, or it can be the sum of the hovering and shooting time at each hovering point in the row area and the flight time to the next hovering point.

[0049] The first image to be detected can be the first image to be detected currently being processed in the image detection queue of the first UAV, that is, the first image to be detected that needs to be processed according to the acquisition sequence of the first image.

[0050] The collaborative decision identifier can be a decision identifier for the corresponding first image to be detected. This identifier indicates whether the image recognition end or the image detection end of the first image to be detected is performing target detection locally on the first UAV, or whether it requests other target second UAVs to collaborate in target detection. For example, 1 and 1 can represent the two types of collaborative decisions. When the collaborative decision identifier is 0, it means that the first image to be detected is performed by the local first UAV. When the collaborative decision identifier is 1, it means that the first image to be detected is collaboratively performed by other second UAVs, realizing the sharing of computing power among multiple UAVs in the UAV swarm.

[0051] In some implementations, the local processing capacity of the first UAV can be assessed by combining the initial remaining detection time and the preset image detection time, and a collaborative decision identifier for the first image to be detected can be generated based on the local processing capacity. For example, the step "determining the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue based on the preset image detection time and the initial remaining detection time" may include: (102.1) Estimate the local processing capacity based on the initial remaining detection time and the preset image detection time; (102.2) Based on the local processing capacity, determine the local processing probability and decision scheduling probability of the first image to be detected in the image detection queue with the current processing progress; (102.3) The collaborative decision identifier of the first image to be detected is randomly determined according to the local processing probability and the decision scheduling probability.

[0052] The local processing capacity can be an indicator of the number of images that can be processed within the current remaining time of a preset detection period. Specifically, it can be determined by combining the real-time remaining detection time within the preset detection period with the preset average processing time per image. For example, the ratio between the real-time initial remaining detection time and the preset image detection time within the preset detection period can be determined. This ratio can be used as the local processing capacity, representing the estimated number of images that can be processed within the current initial remaining detection time of the preset detection period based on the preset image detection time (the time taken to detect a target in a single image).

[0053] The local processing probability can be the probability that the corresponding first image to be detected will remain in the local first UAV for processing (object detection), while the decision scheduling probability refers to the probability that the corresponding first image to be detected will be collaboratively processed by any other second UAV in the UAV swarm. It should be noted that for the same "first image to be detected," the corresponding local processing probability and decision scheduling probability are generated, and the sum of the local processing probability and the decision scheduling probability is 1. For example, if the local processing probability is 0.65, then the decision scheduling probability is 0.35.

[0054] It should be noted that the local processing probability and decision scheduling probability refer to the probability values ​​of generating the final collaborative decision identifier. For example, if the local processing probability is 0.65, it means the probability value of generating a collaborative decision identifier of 0, while the decision scheduling probability is 0.35, which means the probability value of generating a collaborative decision identifier of 1.

[0055] It should be noted that after determining the initial remaining detection time within the preset detection period, for the first image to be detected in the local image detection queue at the current processing progress, a collaborative decision identifier can be generated by combining the initial remaining detection time and the preset image detection time. Specifically, firstly, the ratio between the initial remaining detection time and the preset image detection time can be calculated, and this ratio is determined as the local processing capacity. Then, based on the local processing capacity, the local processing probability and decision scheduling probability of the first image to be detected at the current processing progress in the image detection queue are determined. For example, the local processing capacity can be compared with a preset processing capacity threshold (assumed to be 2). If the local processing capacity is greater than or equal to the preset processing capacity threshold, the local processing probability is set according to the preset probability value, while the decision scheduling probability is equal to 1 minus the local processing probability. Finally, a collaborative decision identifier for the current first image to be detected is randomly generated based on the local processing probability and the decision scheduling probability. Here, the randomness of probability is utilized; regardless of whether the local processing probability is greater than the decision scheduling probability, the collaborative decision identifier may be 0 or 1, determined randomly. In this way, the processing end of the first image to be detected can be decided according to the collaborative decision identifier, thereby determining whether to request other second drones in the drone swarm to perform target detection on the first image to be detected, so as to realize the sharing of computing resources of the drone swarm.

[0056] For example, the central processing unit (CPU) of the first local drone retrieves images from the CPU image queue. Retrieved images from the image detection queue (i.e., the first image to be detected at the current processing progress), let For the task unloading decision (collaborative decision identifier), it indicates the first... The first drone Is the target detection task in Zhang Tu performed by the local first drone? If it is performed by the local first drone (collaborative decision-making identifier) Then the CPU will display the image. Stored in the GPU memory queue In the middle, if it is necessary to uninstall (collaborative decision identifier) Then the CPU will Store in image transmission queue In the meantime. If the computation time of the local GPU image queue... ,in This is the initial remaining detection time. Indicates the preset image detection duration, then execute locally ( The local processing probability of ) is Otherwise, the unloading probability (decision scheduling probability) is 1- Subsequently, a collaborative decision identifier for the first image to be detected is randomly generated based on the local processing probability and the decision scheduling probability; it may be 1 or 0.

[0057] In some implementations, the local processing probability and decision scheduling probability of the first image to be detected at the current processing progress in the image detection queue are determined based on the comparison between the local processing capacity and a preset processing threshold. For example, step (102.2) may include: when the local processing capacity is greater than or equal to the preset processing threshold, a first preset scalar is determined as the local processing probability of the first image to be detected at the current processing progress in the image detection queue, and a decision scheduling probability is calculated based on the local processing probability, wherein the sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is greater than the decision scheduling probability; when the local processing capacity is less than the preset processing threshold, a second preset scalar is determined as the local processing probability of the first image to be detected at the current processing progress in the image detection queue, and a decision scheduling probability is calculated based on the local processing probability, wherein the sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is less than the decision scheduling probability.

[0058] The preset processing capacity threshold can be a judgment value representing the capacity of the local first UAV to handle target detection tasks in the acquired images. When the local processing capacity is greater than or equal to the preset processing capacity threshold, it indicates that the local first UAV can accommodate a sufficient number of target detection tasks within the initial remaining detection time, and each "image to be detected" is considered as one target detection task, meaning that a large number of "images to be detected" can be processed. When the local processing capacity is less than the preset processing capacity threshold, it indicates that the local first UAV can accommodate a relatively small number of target detection tasks within the initial remaining detection time, meaning that a smaller number of "images to be detected" can be processed.

[0059] The first preset scalar can be a pre-set value, such as 0.65, which is not limited here and depends on the actual situation. The second preset scalar is also a pre-set value, such as 0.35, which depends on the actual situation.

[0060] Specifically, after obtaining the local processing capacity, it can be compared with a preset processing threshold to obtain the comparison result. On the one hand, if the comparison result is that the local processing capacity is greater than or equal to the preset processing threshold, the first preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue. That is, the local processing probability is assigned a value by the first preset scalar. Then, the decision scheduling probability is calculated based on the local processing probability. Since the sum of the local processing probability and the decision scheduling probability is 1, the decision scheduling probability is equal to 1 minus the local processing probability. It should be noted that the local processing probability is greater than the decision scheduling probability at this time.

[0061] On the other hand, if the comparison result shows that the local processing capacity is less than the preset processing threshold, the second preset scalar is determined as the local processing probability of the first image to be detected at the current progress in the image detection queue. That is, the local processing probability is assigned a value using the second preset scalar. Then, the decision scheduling probability is calculated based on the local processing probability. Since the sum of the local processing probability and the decision scheduling probability is 1, the decision scheduling probability is equal to 1 minus the local processing probability. It should be noted that the local processing probability is less than the decision scheduling probability at this time. In this way, the local processing probability and the decision scheduling probability of the first image to be detected at the current processing progress in the image detection queue can be determined, so as to randomly generate a collaborative decision label for the first image to be detected by combining the local processing probability and the decision scheduling probability.

[0062] The above method can detect the number of images to be detected in the image detection queue. The image to be detected refers to the first image in the image detection queue. When it is detected that the image detection queue contains at least one image to be detected, the initial remaining detection time in the prediction detection period is first determined. Then, the preset image detection time and the initial remaining detection time are used as decision factors to determine the collaborative decision identifier of the first image to be detected in the image detection queue with the current processing progress. Thus, for each first image in the local image detection queue, a collaborative decision identifier is generated by combining the preset image detection time and the remaining time in the image detection period. This determines whether to request other second drones in the drone swarm to perform target detection on the first image to be detected, so as to realize the sharing of computing resources in the drone swarm and improve the target detection efficiency of the drone swarm for the target range area.

[0063] 103. When the collaborative decision identifier is detected as the unload decision identifier, obtain the communication link bandwidth between the local first UAV and each second UAV, as well as the total communication bandwidth corresponding to multiple communication links, and select the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth.

[0064] In this embodiment, after obtaining the collaborative decision identifier of the first image to be detected with the current processing progress, the type of the collaborative decision identifier can be identified. If the type of the collaborative decision identifier is an unloading decision identifier, the communication link bandwidth between the local first UAV and each second UAV, as well as the total communication bandwidth corresponding to multiple communication links, are obtained. According to the ratio between the bandwidth of each communication link and the total communication bandwidth, a target communication link is selected. In this way, when it is necessary to send the first image to be detected to other second UAVs in the UAV swarm for target detection, the target communication link with the maximum communication bandwidth can be selected according to the size of the communication bandwidth between the first UAV and other second UAVs, so as to reduce the transmission time of the first image to be detected and improve the target detection efficiency of the UAV swarm for the target range area during the sharing of computing resources in the UAV swarm.

[0065] The unloading decision identifier indicates that the first image to be detected will be sent to another second UAV for collaborative target detection, and can be represented by 1. It should be noted that the local processing identifier indicates that the first UAV is performing target detection on the first image to be detected locally, and is represented by 0. The collaborative decision identifier can be either the unloading decision identifier or the local processing identifier.

[0066] It should be noted that when the type of the collaborative decision identifier is an unloading decision identifier, in order to improve the speed of transmitting the first image to be detected to other target second drones in the drone swarm, it is necessary to obtain the communication link bandwidth between the local first drone and each second drone. Then, combining the communication link bandwidth of each communication link, the total communication bandwidth corresponding to multiple communication links between the local first drone and multiple second drones is calculated, that is, the communication link bandwidths of each communication link are added together to obtain the total communication bandwidth. Subsequently, by dividing the communication link bandwidth of each communication link by the total communication bandwidth, the bandwidth ratio between the communication link bandwidth of each communication link and the total communication bandwidth is determined. The transmission performance of each communication link is evaluated by the bandwidth ratio of each communication link, thereby identifying the target communication link with high transmission speed, so as to improve the efficiency of collaborative target detection of the first image to be detected across drone ends (other target second drones).

[0067] In some implementations, a communication scheduling score for each communication link can be determined based on the ratio between the bandwidth of each communication link and the total communication bandwidth, and a target communication link with the highest communication scheduling score can be selected from multiple communication links. For example, step 103, "selecting a target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth," may include: (103.1) Determine the communication scheduling score of the communication link between the local first UAV and each second UAV based on the ratio between the bandwidth of each communication link and the total communication bandwidth; (103.2) Sort multiple communication links according to the communication scheduling score from largest to smallest to obtain the link sorting relationship; (103.3) Select the target communication link that is ranked first from multiple communication links according to the link ranking relationship.

[0068] The communication scheduling score can be a decision score representing the bandwidth of the corresponding communication link. It is used to evaluate the proportion of bandwidth of the corresponding communication link among multiple communication links. The larger the communication scheduling score, the larger the proportion of the bandwidth of the communication link in the total bandwidth of multiple communication links, and the higher the data transmission rate. Conversely, if the communication scheduling score is smaller, the smaller the proportion of the bandwidth of the communication link in the total bandwidth of multiple communication links, and the lower the data transmission rate.

[0069] Specifically, after obtaining the communication link bandwidth of the communication links between the local first UAV and each of the other second UAVs, as well as the total communication bandwidth corresponding to multiple communication links, the bandwidth of each communication link is divided by the total communication bandwidth to determine the bandwidth ratio between each communication link and the total communication bandwidth. This bandwidth ratio is then used as the communication scheduling score for the communication link between the local first UAV and its corresponding second UAV. Next, the multiple communication links are sorted in descending order of their communication scheduling scores; the higher the score, the higher the ranking of the corresponding communication link, thus obtaining a link ranking relationship. Finally, based on the ranking order of the multiple communication links in the link ranking relationship, the communication link ranked first is selected as the target communication link, for example, the first-ranked communication link is selected as the target communication link. In this way, the target communication link with the largest bandwidth can be selected from multiple communication links so that the first image to be detected can be sent to the target second UAV corresponding to the target communication link, improving the transmission rate of the first image to be detected. This improves the target detection efficiency of the UAV swarm for the target area during the sharing of computing resources within the UAV swarm.

[0070] For example, from the image transmission queue Image extracted from That is, the first image to be detected, let For mission communication decisions, the sequence number of the sending drone is indicated, and then the corresponding first drone is used to establish a direct communication link (communication link bandwidth) to other second drones (D2D). )send To the corresponding UAV's GPU image queue In the middle. Task offloading communication decision (i.e., communication scheduling score). The offloading link (target communication link) is selected based on the communication scheduling score, specifying the drone to be offloaded for the designated task. It should be noted that the time consumed for sending each image (the first image to be detected) is... ,in, This indicates the amount of data for the first image to be detected. The transmission will stop when the remaining detection time is insufficient to send an image.

[0071] In some implementations, when the collaborative decision identifier is detected as a local processing identifier, target detection is performed on the first image to be detected using a locally pre-trained target detection model to obtain target detection results. These results are then sent to a target edge server, which sorts the target detection results against at least one historical target detection result corresponding to the first sub-region to obtain a target detection result sequence. The historical target detection results are the detection results sent by the first UAV and / or any second UAV for historical first images within the first sub-region. Thus, if the collaborative decision identifier is of the local processing identifier type, target detection is performed on the first image to be detected using a local target detection model to obtain target detection results.

[0072] By using the above method, when the collaborative decision identifier type is unload decision identifier, the communication link bandwidth between the local first UAV and each second UAV, as well as the total communication bandwidth corresponding to multiple communication links, can be obtained. According to the ratio between the bandwidth of each communication link and the total communication bandwidth, the target communication link is selected. In this way, when it is necessary to send the first image to be detected to other second UAVs in the UAV swarm for target detection, the target communication link with the largest communication bandwidth can be selected according to the size of the communication bandwidth between the first UAV and other second UAVs, so as to reduce the transmission time of the first image to be detected and improve the target detection efficiency of the UAV swarm for the target range area during the sharing of computing resources in the UAV swarm.

[0073] 104. Send the first image to be detected to the target second UAV corresponding to the target communication link.

[0074] In this embodiment, after selecting a target communication link from multiple communication links between the first UAV and multiple second UAVs, the first image to be detected can be sent to the target second UAV corresponding to the target communication link. In this way, the first image to be detected is transmitted to the corresponding target second UAV through the target communication link with the largest bandwidth, which improves the transmission rate of the first image to be detected and saves the transmission time of the first image to be detected. Furthermore, the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server. In this way, the sharing of computing resources in the UAV swarm is realized, and the target detection efficiency of the UAV swarm for the target range area is improved.

[0075] In some implementations, the remaining detection time of the target within a preset detection period can be calculated in real time, and when the real-time remaining detection time of the target is greater than or equal to a preset time threshold, the process returns to steps 101 to 104. For example, after step 104, the process may include: obtaining the first data volume of the first image to be detected, and estimating the target transmission time of the first image to be detected from the local first UAV to the target second UAV based on the first data volume; obtaining the actual processing time of the local first UAV when processing the second image to be detected in the image detection queue; obtaining the result return time of the local first UAV when transmitting the current detection result to the target edge server; selecting the target unit occupancy time from the target transmission time, the actual processing time, and the result return time; determining the remaining detection time of the target within the preset detection period based on the initial remaining detection time and the target unit occupancy time; and when the remaining detection time of the target is greater than or equal to the preset time threshold, returning to the step of acquiring the first image for the first sub-region within the target range area.

[0076] The first data volume can be a numerical value representing the size of the first image to be detected, such as the size of the first image to be detected being in kilobytes, megabytes, or gigabytes.

[0077] The target transmission duration can be the estimated transmission time from the first image to be detected to the target second UAV.

[0078] The actual processing time refers to the time spent performing target detection on a single image to be detected. It should be noted that the second image to be detected refers to the image following the first image in the time sequence. If the first image to be detected is used for collaborative target detection on the second target UAV, the processing time of the first image to be detected can be directly obtained from the second target UAV as the actual processing time.

[0079] The result transmission time can refer to the time taken to transmit the detection results to the target edge server after performing target detection on the corresponding image to be detected.

[0080] The target unit occupancy time can be the maximum value among the target transmission time, actual processing time, and result return time. It should be noted that the transmission process of the first image to be detected, the target detection process of the local first UAV on the corresponding second image to be detected, and the process of the local first UAV transmitting the real-time detection results after target detection to the target edge server are performed in parallel; there is no temporal order among these three processes. Therefore, the target unit occupancy time can be the maximum value selected from the target transmission time, actual processing time, and result return time.

[0081] Among them, the target remaining detection time refers to the current real-time remaining time within the preset detection period.

[0082] Specifically, after sending the first image to be detected to the target second UAV corresponding to the target communication link, the first data volume of the first image to be detected can be obtained, and the target transmission time of the first image to be detected from the local first UAV to the target second UAV can be estimated based on the first data volume. For example, the calculation process of the target transmission time is as follows: The target transmission time consumed for sending one image is: ,in, This indicates the data volume of the first image to be detected. This indicates the bandwidth of the target communication link.

[0083] Then, obtain the actual processing time of the first local drone when processing the second image to be detected in the image detection queue. The actual processing time can be... .

[0084] Next, the transmission time of the local first UAV to the target edge server is obtained. Specifically, this can be achieved by directly obtaining the transmission time of the target detection result corresponding to the first image to be detected from the target second UAV to the target edge server; alternatively, the second data volume of the real-time detection result corresponding to historical target images to be detected can be obtained, and the transmission time is determined based on the second data volume and the transmission bandwidth between the local first UAV and the target edge server. For example, the second data volume is represented as... The transmission bandwidth between the local first drone and the target edge server is expressed as The result return time The calculation process is as follows: .

[0085] Then, the longest duration among the target transmission duration, actual processing duration, and result return duration is selected as the target unit occupancy duration.

[0086] Finally, based on the initial remaining detection time and the target unit's occupied time, the remaining detection time within the preset detection period is determined. Specifically, the remaining detection time within the preset detection period can be obtained by subtracting the target unit's occupied time from the initial remaining detection time. Then, the remaining detection time is compared with a preset time threshold. This preset time threshold can be used to determine whether the remaining detection time is sufficient. It can be set to 0 or nearly equal to 0. On one hand, if the remaining detection time is greater than or equal to the preset time threshold, it indicates that the preset detection period has not yet ended, and the process returns to step 101 to perform the step of "acquiring the first image for the first sub-region within the target range area," and continues with steps 101 to 104. On the other hand, if the remaining detection time is less than the preset time threshold, it indicates that the preset detection period has ended, and step 101 is paused until the local first drone switches to the next row of areas within the first sub-region, and then steps 101 to 104 are repeated until the target detection for the first sub-region is completed. This enables target detection by traversing the first sub-region, accurately identifying targets within the first sub-region and improving target detection accuracy.

[0087] In some implementations, the estimated transmission time of the first image to be detected to the target second UAV can be obtained, and when the estimated transmission time is detected to be greater than or equal to the initial remaining detection time, the first image to be detected is sent to the target second UAV corresponding to the target communication link. For example, before step 104, the method may further include: determining the first data volume of the first image to be detected and obtaining the target transmission bandwidth between the local first UAV and the target second UAV; estimating the estimated transmission time of the first image to be detected to the target second UAV based on the first data volume and the target transmission bandwidth; then step 104 includes: when the estimated transmission time is detected to be greater than or equal to the initial remaining detection time, sending the first image to be detected to the target second UAV corresponding to the target communication link.

[0088] Specifically, the system can obtain the first data volume of the first image to be detected, and the target transmission bandwidth between the local first UAV and the target second UAV. Then, based on the first data volume and the target transmission bandwidth, it can estimate the estimated transmission time of the first image to be detected to the target second UAV. For example, the first data volume of the first image to be detected can be expressed as... The target transmission bandwidth of the target communication link between the local first UAV and the target second UAV is The calculation process for the estimated transmission time is as follows: .

[0089] Furthermore, the estimated transmission time is compared to the initial remaining detection time. On one hand, if the estimated transmission time is greater than or equal to the initial remaining detection time, it indicates that the remaining detection time within the preset detection period is sufficient to transmit the first image to be detected to the target second UAV for collaborative target detection. Therefore, step 104 is executed, sending the first image to be detected to the target second UAV corresponding to the target communication link. On the other hand, if the estimated transmission time is less than the initial remaining detection time, it indicates that the remaining detection time within the preset detection period is insufficient to transmit the first image to be detected to the target second UAV for collaborative target detection. In this case, step 104 is executed, allowing the local first UAV to directly perform target detection on the first image to be detected, thereby improving the target detection efficiency of the first image to be detected.

[0090] In this embodiment, the first UAV can also receive target images sent by other second UAVs to perform collaborative target detection calculations. This allows the computing power of the first UAV to be used to process target images from other second UAVs, enabling the sharing of computing power resources among the UAV swarm. This collaboratively detects targets in images of other sub-regions within the target range, thereby improving the target detection efficiency within the target range area.

[0091] In some implementations, the local first UAV can receive a target image to be detected sent by any second UAV, perform target detection on the target image to obtain a collaborative target detection result, and send the collaborative target detection result back to the target edge server. For example, after step 104, the implementation may further include: receiving a target image to be detected sent by any second UAV; inputting the target image to be detected into a pre-trained target detection model to obtain the corresponding collaborative target detection result; and sending the collaborative target detection result to the target edge server, so that the target edge server sorts the collaborative target detection result with at least one historical target detection result corresponding to the corresponding second sub-region to obtain a target detection result sequence.

[0092] The target image to be detected can be an environmental image of a location point in a region taken by one of the second drones in the corresponding sub-region. For details, please refer to the description of the first image, which will not be described here.

[0093] The target detection model can be an image recognition model, for example, using yolov5m6 as the target detection model.

[0094] The collaborative target detection result is the output of the target detection model, for example, using... Indicates the target image to be detected, Indicating the results of collaborative target detection, then yolov5m6( ).

[0095] Specifically, the local first UAV can receive target images sent by any second UAV to perform collaborative target detection on the target images of other second UAVs; the target images are input into a pre-trained target detection model to obtain corresponding collaborative target detection results; the collaborative target detection results are sent to a target edge server so that the target edge server sorts the collaborative target detection results with at least one historical target detection result corresponding to the second sub-region to obtain a target detection result sequence. It should be noted that after the target edge server has sorted all the target detection results of each sub-region, it obtains a complete target detection result sequence corresponding to each sub-region. Furthermore, the complete target detection result sequences corresponding to each sub-region are merged and combined to obtain the target detection results of the entire region corresponding to the target range region.

[0096] In this way, the first image to be detected can be sent to the target second UAV corresponding to the target communication link. In this way, the first image to be detected is transmitted to the corresponding target second UAV through the target communication link with the largest bandwidth, which improves the transmission rate of the first image to be detected and saves the transmission time. Furthermore, the target second UAV can send the target detection result corresponding to the first image to be detected to the target edge server. In this way, the computing resources of the UAV swarm are shared and the target detection efficiency of the UAV swarm in the target range area is improved.

[0097] To facilitate understanding of the above embodiments, a complete example is provided below: Figure 3 This is an example diagram illustrating the task division and trajectory planning scenario for the target area provided in this application embodiment. Combined with... Figure 3 As shown, K drones are used to perform an area search task, dividing the task area into K sub-regions, with each drone responsible for one region. Figure 3 As shown, the drone flies along an S-shaped planned trajectory, with a ground coverage area of ​​Ra × Rb square meters for a single image captured by the drone. The distance between camera points along the flight path is Ra meters, the distance between flight paths is Rb meters, the drone hovers for th seconds, and the flight speed is v meters per second. It is required to... The hovering points cover a sub-region, and the image taken by the k-th drone at the l-th hovering point is... (Data volume is) The target detection result is (Data volume is) ), Let k ∈ {1, 2, …, K}. Then the time it takes for the k-th drone to complete its flight and photography is... The sum of hovering shooting time and flight time:

[0098] Among them, sub-region or drone index Hover point index .

[0099] Figure 4 This is an example diagram of the architecture of a drone swarm cooperative target detection system provided in an embodiment of this application. (Combined with...) Figure 4 As shown, the system architecture includes a drone swarm consisting of multiple drones and a target edge server. This is a common air-to-ground collaborative model of a Drone Ad Hoc Network (DAHN) consisting of K drones and one edge server. Each drone carries a camera and a computing module, which integrates LR-Wifi, 5G, and corresponding system-on-a-chip (SOC) edge computing modules. The LR-Wifi module is used to build the DAHN, enabling data sharing, task collaboration, and computation offloading among any drones; this link is defined as a drone-to-drone (D2D) link. The 5G module enables high-speed, low-latency communication between the drones and the edge server, allowing for data backhaul, computation offloading, and model updates; this link is defined as a drone-to-server (D2S) link. Let the D2S link bandwidth of drone k be... For ease of analysis, we adopt a time-slotted structure. Let the time slot of the c-th region search task be... The network consists of time slots with a duration of τ seconds, and the network state remains unchanged within each time slot. C represents the number of times the area search task is performed. The drones self-organize and network via the WiFi module on the computing power box, with each drone as a node and D2D links between drones as edges, forming a drone swarm network. It can be depicted by a graph. Among them, For a set of nodes, nodes Indicates the k-th drone. For edge set, Represents a node arrive Communication bandwidth, The communication bandwidth between any two drones is remeasured at regular intervals.

[0100] It should be noted that the drone consists of modules such as an onboard camera, main control unit, flight system, and computing power box. The main control unit provides a data interface for the onboard camera to the computing power box, through which image data captured by the camera is transmitted back to the computing power box for processing. The drones conduct collaborative search tasks through onboard WiFi networking, and the drones dynamically rebalance tasks through a swarm network based on differences in onboard computing power.

[0101] Figure 5 This is an example diagram illustrating a target detection processing scenario for a drone swarm cooperative operation provided in an embodiment of this application. Combined with... Figure 5 As shown, the drone's flight photography and image target detection tasks can be decoupled through a CPU image queue, meaning the two tasks are independent. The flight photography task is a simple task flow, and the image... Images are captured sequentially from each shooting point along the flight path and transmitted back to the CPU image queue via the drone's main control unit. middle.

[0102] Image target detection tasks are more complex, involving image data storage, target detection task offloading decisions, communication scheduling, computation scheduling, and result return to the server, among other processes, and these processes are interdependent. For the computation scheduling decision-making procedure, let... These correspond to different computing modes, including performance mode, balanced mode, and power-saving mode. Let the average power consumption of the drone k's GPU processing an image in these three modes be... The average calculation time is respectively Average power consumption and average computation time were obtained experimentally. Since setting the computation mode of the system-on-a-chip's computing box may require a restart and cannot be done online, initialization can only be performed before the scheduling algorithm begins. .

[0103] Calculation and communication scheduling are performed at the time slot level. For the d-th time slot of the c-th target detection task... , The remaining time within the time slot is initialized to... The main task processing procedure includes the following: (1) Task offloading decision: CPU offloads from CPU image queue Image extraction ,make For the task offloading decision, it indicates whether the target detection task for the l-th image of the k-th UAV is executed locally. If it is executed locally ( Then the CPU will display the image. Stored in the GPU memory queue If you need to uninstall (in the middle) Then the CPU will Store in image transmission queue In the meantime. If the computation time of the local GPU image queue... Then execute locally ( The probability of ) is The probability of uninstallation is 1- .

[0104] (2) Task offloading communication scheduling: from Image extracted from ,make For mission communication decisions, the sequence number of the UAV to be sent is indicated, and then transmitted via the corresponding D2D link (bandwidth). )send To the corresponding UAV's GPU image queue In the middle. Task unloading communication decision. The unloading link is selected based on probability, i.e., the drone to be unloaded for the specified task. The time consumed for sending each image is... The transmission will stop when there is not enough time left to send a single image.

[0105] (3) GPU computation scheduling: The GPU sequentially retrieves images from the GPU image queue. Image extracted from The calculation results were obtained by executing the object detection model (yolov5m6). ,Right now yolov5m6( The time consumed is and the target detection results Store results in the sending queue middle.

[0106] (4) Result feedback: If The CPU sends results from the queue. Extracting the result , the results Results queue sent back to the edge server In the process, the time consumed to send one image is .

[0107] (5) Since the three processes of D2D image transmission, GPU computation, and 5G network result return can be executed in parallel pipeline, the maximum time consumption can be taken. Therefore, the remaining time within the time slot... ,if If there is still time remaining in the current time slot, return to step (1) to continue task processing; otherwise, execute step (6).

[0108] (6) After the server completes the collection of all results, it sorts and organizes the data according to the sub-region and hover point number.

[0109] Through the above examples, drone swarms can achieve large-scale, efficient collaborative computing, enabling drones to break free from the limitations of cloud computing power and truly move towards on-orbit, real-time computing. This allows for large-scale, highly maneuverable "what you see is what you get," transforming drones from detection terminals into intelligent terminals and achieving low-latency collaborative computing.

[0110] As described above, the target detection method of the UAV swarm collaboration in this application embodiment is applied to a first UAV in a UAV swarm, which also includes multiple second UAVs. Each UAV has computing power, and each UAV in the UAV swarm can collect corresponding images in each sub-region of the target range area. Taking the first UAV as an example, the first UAV can collect the first image corresponding to the first sub-region according to a preset detection cycle and add the first image to the image detection queue. When the image detection queue contains multiple images to be detected, the current initial remaining time within the preset detection cycle is determined. Based on the preset image detection time and the initial remaining detection time, the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue is determined. When the collaborative decision identifier is an unloading decision identifier, the communication link bandwidth between the local first UAV and each second UAV and the total communication bandwidth corresponding to multiple communication links are obtained. According to the ratio between the bandwidth of each communication link and the total communication bandwidth, a target communication link is selected from multiple communication links. Finally, the first image to be detected is sent to the target second UAV corresponding to the target communication link, so that the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server.

[0111] Based on this, when the first UAV of this application acquires a first image of a first sub-region of the target area within a preset detection period, it adds the first image to the image detection queue. When the image detection queue contains a number of images to be detected, it makes a collaborative decision on the first image to be detected with the current processing progress by combining the current initial remaining time within the preset detection period with the preset image detection time. When the decision requests other second UAVs to perform collaborative calculation on the first image to be detected, it selects the target second UAV with the largest communication link bandwidth from among the multiple second UAVs, and sends the first image to be detected to the target second UAV corresponding to the target communication link for collaborative calculation. In this way, the individual UAVs in the UAV swarm can share computing resources, and the UAV swarm can collaboratively perform recognition calculation on the acquired images, thereby improving the target detection efficiency and real-time performance of the UAV swarm.

[0112] For details on the implementation of each of the above steps, please refer to the previous examples, which will not be repeated here.

[0113] To facilitate better implementation of the UAV swarm cooperative target detection method provided in this application, this application also provides a target detection device based on the aforementioned UAV swarm cooperative method. The meanings of the terms used are the same as in the aforementioned UAV swarm cooperative target detection method, and specific implementation details can be found in the descriptions within the method embodiments.

[0114] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a target detection device for drone swarm collaboration provided in an embodiment of this application. The target detection device for drone swarm collaboration is integrated into the computer equipment of this application and is applied to a first drone in a drone swarm. The drone swarm also includes multiple second drones. The target detection device for drone swarm collaboration may include a data acquisition unit 401, a determination unit 402, a communication scheduling unit 403, and a collaboration request unit 404.

[0115] The acquisition unit 401 is used to acquire a first image for a first sub-region within the target range according to a preset detection cycle, and add the first image to the image detection queue. Multiple second UAVs are used to perform target detection on other sub-regions within the target range except for the first sub-region. The determining unit 402 is used to determine the current initial remaining detection time within a preset detection period when it is detected that the image detection queue contains at least one image to be detected, and to determine the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue based on the preset image detection time and the initial remaining detection time. The communication scheduling unit 403 is used to obtain the communication link bandwidth of the communication link between the local first UAV and each second UAV and the total communication bandwidth corresponding to multiple communication links when the cooperative decision identifier is detected as the unload decision identifier, and select the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The collaborative request unit 404 is used to send the first image to be detected to the target second UAV corresponding to the target communication link, so that the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server.

[0116] In some embodiments, the determining unit 402 is further configured to: Estimate local processing capacity based on initial remaining detection time and preset image detection time; Based on the local processing capacity, determine the local processing probability and decision scheduling probability of the first image to be detected in the image detection queue with the current processing progress. The collaborative decision identifier of the first image to be detected is randomly determined based on the local processing probability and the decision scheduling probability.

[0117] In some embodiments, the determining unit 402 is further configured to: When the local processing capacity is greater than or equal to the preset processing capacity threshold, the first preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue, and the decision scheduling probability is calculated based on the local processing probability. The sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is greater than the decision scheduling probability. When the local processing capacity is less than the preset processing capacity threshold, the second preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue, and the decision scheduling probability is calculated based on the local processing probability. The sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is less than the decision scheduling probability.

[0118] In some implementations, the communication scheduling unit 403 is further configured to: The communication scheduling score of the communication link between the local first UAV and each second UAV is determined based on the ratio between the bandwidth of each communication link and the total communication bandwidth. The multiple communication links are sorted in descending order of their communication scheduling scores to obtain the link sorting relationship. Based on the link ordering relationship, select the target communication link that is ranked first from multiple communication links.

[0119] In some implementations, the target detection device for UAV swarm collaboration further includes a cyclic execution unit for: Acquire the first data volume of the first image to be detected, and estimate the target transmission time of the first image to be detected from the local first UAV to the target second UAV based on the first data volume; Obtain the actual processing time of the first local drone when processing the second image to be detected in the image detection queue; Obtain the time it takes for the first local drone to transmit the current detection results back to the target edge server; Select the target unit's time duration from the target transmission duration, actual processing duration, and result return duration; The target remaining detection time within the preset detection cycle is determined based on the initial remaining detection time and the target unit's occupied time. When the remaining detection time of the target is greater than or equal to the preset time threshold, return to the step of acquiring the first image for the first sub-region within the target range area.

[0120] In some implementations, the target detection device for UAV swarm collaboration further includes a collaborative decision-making unit, used for: Determine the first data volume of the first image to be detected, and obtain the target transmission bandwidth between the local first UAV and the target second UAV; Based on the first data volume and the target transmission bandwidth, the estimated transmission time of the first image to be detected to the target second UAV is estimated. The collaborative request unit is further configured to send the first image to be detected to the target second UAV corresponding to the target communication link when the estimated transmission time is greater than or equal to the initial remaining detection time.

[0121] In some implementations, the target detection model for UAV swarm cooperation also includes a cooperative computing unit for: Receive the target image to be detected sent by any second UAV; The target image to be detected is input into a pre-trained target detection model to obtain the corresponding collaborative target detection results; The collaborative target detection results are sent to the target edge server, so that the target edge server sorts the collaborative target detection results with at least one historical target detection result corresponding to the second sub-region to obtain a target detection result sequence.

[0122] As described above, in this embodiment, when the first UAV acquires a first image of a first sub-region of the target area within a preset detection period, it adds the first image to the image detection queue. When the image detection queue contains a number of images to be detected, it makes a collaborative decision on the first image to be detected with the current processing progress by combining the current initial remaining time within the preset detection period with the preset image detection time. When the decision requests other second UAVs to perform collaborative calculation on the first image to be detected, it selects the target second UAV with the largest communication link bandwidth from among the multiple second UAVs, and sends the first image to be detected to the target second UAV corresponding to the target communication link for collaborative calculation. In this way, the individual UAVs in the UAV swarm can share computing resources, and the UAV swarm can collaboratively perform recognition calculation on the acquired images, thereby improving the target detection efficiency and real-time performance of the UAV swarm.

[0123] The specific implementation of each of the above units can be found in the previous embodiments, and will not be repeated here.

[0124] Figure 7 To implement the structural block diagram of a portion of the terminal 110 in this embodiment, the terminal 110 includes components such as a radio frequency (RF) circuit 510, a memory 515, a sensor 550, a wireless fidelity (WiFi) module 570, a processor 580, and a power supply 590. Those skilled in the art will understand that the terminal 110 structure shown in the figures does not constitute a limitation on the drone, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0125] The RF circuit 510 can be used to receive and transmit signals during information transmission or calls. In particular, it receives downlink information from the base station and processes it with the processor 580; in addition, it transmits uplink data to the base station.

[0126] The memory 515 can be used to store software programs and modules. The processor 580 executes various functional applications and data processing of the terminal by running the software programs and modules stored in the memory 515.

[0127] The processor 580 may include a system on chip (SoC), which integrates a central processing unit (CPU) and a graphics processing unit (GPU) that share memory.

[0128] In this embodiment, the processor 580 included in the terminal 110 can execute the target detection method of UAV swarm collaboration in the previous embodiment.

[0129] The terminal 110 in this application embodiment includes, but is not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc. This invention embodiment can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving.

[0130] Figure 8 This is a partial structural block diagram of a server 120 implementing an embodiment of this application. The server 120 can vary significantly due to different configurations or performance characteristics, and may include one or more central processing units (CPUs) 622 (e.g., one or more processors) and memory 632, and one or more storage media 620 (e.g., one or more mass storage devices) for storing application programs 642 or data 644. The memory 632 and storage media 620 may be temporary or persistent storage. The program stored in the storage media 620 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the server 120. Furthermore, the CPU 622 may be configured to communicate with the storage media 620 and execute the series of instruction operations in the storage media 620 on the server 120.

[0131] Server 120 may also include one or more power supplies 626, one or more wired or wireless network interfaces 650, one or more input / output interfaces 658, and / or one or more operating systems 641, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0132] The central processing unit 622 in server 120 can be used to execute the target detection method of drone swarm collaboration according to the embodiments of this application.

[0133] This application also provides a computer-readable storage medium for storing program code for executing the drone swarm cooperative target detection method of the foregoing embodiments.

[0134] This application also provides a computer program product, which includes a computer program. A processor of a computer device reads and executes the computer program, causing the computer device to perform the aforementioned target detection method for UAV swarm cooperation.

[0135] Furthermore, the terms “comprising” and “including”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, apparatus, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are expressly listed, but may include other steps or units that are not expressly listed or that are inherent to such process, method, product or device.

[0136] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0137] It should be understood that in the description of the embodiments of this application, "multiple" means two or more, "greater than", "less than", "exceeding" etc. are understood to exclude the number itself, and "above", "below", "within" etc. are understood to include the number itself.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0139] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0140] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0141] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.

[0143] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0144] The above is a detailed description of the embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A target detection method for unmanned aerial vehicle (UAV) swarm cooperation, characterized in that, The method involves applying a first drone in a drone swarm, the drone swarm also including multiple second drones, the method comprising: According to a preset detection cycle, a first image is acquired for the first sub-region within the target range area, and the first image is added to the image detection queue. The multiple second UAVs are used to perform target detection on other sub-regions within the target range area other than the first sub-region. When it is detected that the image detection queue contains at least one image to be detected, the initial remaining detection time within the preset detection period is determined, and the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue is determined according to the preset image detection time and the initial remaining detection time. When the collaborative decision identifier is detected as an unload decision identifier, the communication link bandwidth between the local first UAV and each second UAV and the total communication bandwidth corresponding to the multiple communication links are obtained, and the target communication link is selected according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The first image to be detected is sent to the target second UAV corresponding to the target communication link, so that the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server.

2. The target detection method for UAV swarm cooperation according to claim 1, characterized in that, The step of determining the collaborative decision identifier of the first image to be detected in the image detection queue with the current processing progress based on the preset image detection time and the initial remaining detection time includes: Based on the initial remaining detection time and the preset image detection time, the local processing capacity is estimated; Based on the local processing capacity, determine the local processing probability and decision scheduling probability of the first image to be detected in the image detection queue with the current processing progress. The collaborative decision identifier of the first image to be detected is randomly determined according to the local processing probability and the decision scheduling probability.

3. The target detection method for UAV swarm cooperation according to claim 2, characterized in that, The step of determining the local processing probability and decision scheduling probability of the first image to be detected in the image detection queue at its current progress based on the local processing capacity includes: When the local processing capacity is greater than or equal to a preset processing threshold, the first preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue, and the decision scheduling probability is calculated based on the local processing probability, wherein the sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is greater than the decision scheduling probability. When the local processing capacity is less than the preset processing capacity threshold, the second preset scalar is determined as the local processing probability of the first image to be detected in the current progress of the image detection queue, and the decision scheduling probability is calculated based on the local processing probability, wherein the sum of the local processing probability and the decision scheduling probability is 1, and the local processing probability is less than the decision scheduling probability.

4. The target detection method for UAV swarm cooperation according to claim 1, characterized in that, Selecting a target communication link based on the ratio between the bandwidth of each communication link and the total communication bandwidth includes: Based on the ratio between the bandwidth of each communication link and the total communication bandwidth, a communication scheduling score is determined for the communication link between the local first UAV and each second UAV. According to the communication scheduling score in descending order, the multiple communication links are sorted to obtain the link sorting relationship; According to the link sorting relationship, the target communication link that is ranked first among the multiple communication links is selected.

5. The target detection method for UAV swarm cooperation according to any one of claims 1 to 4, characterized in that, After sending the first image to be detected to the target second UAV corresponding to the target communication link, the method further includes: Acquire the first data volume of the first image to be detected, and estimate the target transmission time of the first image to be detected from the local first UAV to the target second UAV based on the first data volume; Obtain the actual processing time of the local first UAV when processing the second image to be detected in the image detection queue; Obtain the time it takes for the local first drone to transmit the current detection results back to the target edge server; Select the target unit occupancy time from the target transmission time, the actual processing time, and the result return time; The target remaining detection time within the preset detection period is determined based on the initial remaining detection time and the target unit occupancy time. When the remaining detection time of the target is greater than or equal to the preset time threshold, return to the step of acquiring the first image for the first sub-region within the target range area.

6. The target detection method for UAV swarm cooperation according to claim 1, characterized in that, Before sending the first image to be detected to the target second UAV corresponding to the target communication link, the method further includes: Determine the first data volume of the first image to be detected, and obtain the target transmission bandwidth between the local first UAV and the target second UAV; Based on the first data volume and the target transmission bandwidth, the estimated transmission time of the first image to be detected to the target second UAV is estimated. The step of sending the first image to be detected to the target second UAV corresponding to the target communication link includes: When the estimated transmission duration is greater than or equal to the initial remaining detection duration, the first image to be detected is sent to the target second UAV corresponding to the target communication link.

7. The target detection method for UAV swarm cooperation according to claim 1, characterized in that, The method further includes: Receive the target image to be detected sent by any second UAV; The target image to be detected is input into a pre-trained target detection model to obtain the corresponding collaborative target detection results; The collaborative target detection results are sent to the target edge server, so that the target edge server sorts the collaborative target detection results with at least one historical target detection result corresponding to the second sub-region to obtain a target detection result sequence.

8. A target detection system for unmanned aerial vehicle (UAV) swarm cooperation, characterized in that, The target detection system includes a first UAV and multiple second UAVs, including: The first UAV acquires a first image of a first sub-region within the target range according to a preset detection cycle, and adds the first image to an image detection queue. The plurality of second UAVs are used to perform target detection on other sub-regions within the target range besides the first sub-region. When the first UAV detects that the image detection queue contains at least one image to be detected, it determines the current initial remaining detection time within the preset detection period, and determines the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue based on the preset image detection time and the initial remaining detection time. When the first UAV detects that the collaborative decision identifier is an unload decision identifier, it obtains the communication link bandwidth of the communication link between the local first UAV and each second UAV, as well as the total communication bandwidth corresponding to the multiple communication links, and selects the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth. The first UAV sends the first image to be detected to the target second UAV corresponding to the target communication link; The target second UAV performs target detection on the first image to be detected using a pre-trained target detection model, obtains the target detection result, and sends the target detection result to the target edge server; The target edge server receives the target detection result and sorts the target detection result with at least one historical target detection result corresponding to the first sub-region to obtain a target detection result sequence. The historical target detection result is the detection result sent by the first UAV and / or any one of the second UAVs for a historical first image within the first sub-region.

9. A target detection device for unmanned aerial vehicle (UAV) swarm collaboration, characterized in that, The method involves applying a first drone in a drone swarm, the drone swarm also including multiple second drones, the method comprising: The acquisition unit is used to acquire a first image for a first sub-region within the target range according to a preset detection cycle, and add the first image to the image detection queue. The plurality of second UAVs are used to perform target detection on other sub-regions within the target range except for the first sub-region. The determining unit is configured to, when detecting that the image detection queue contains at least one image to be detected, determine the current initial remaining detection time within the preset detection period, and determine the collaborative decision identifier of the first image to be detected with the current processing progress in the image detection queue based on the preset image detection time and the initial remaining detection time; The communication scheduling unit is used to, when the cooperative decision identifier is detected as an unload decision identifier, obtain the communication link bandwidth of the communication link between the local first UAV and each second UAV and the total communication bandwidth corresponding to the multiple communication links, and select the target communication link according to the ratio between the bandwidth of each communication link and the total communication bandwidth; The collaborative request unit is used to send the first image to be detected to the target second UAV corresponding to the target communication link, so that the target second UAV sends the target detection result corresponding to the first image to be detected to the target edge server.

10. A computer device, characterized in that, The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the target detection method for UAV swarm collaboration as described in any one of claims 1 to 7.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to execute the UAV swarm cooperative target detection method according to any one of claims 1 to 7.

Images