Path planning method and device, unmanned aerial vehicle and vehicle-mounted unmanned aerial vehicle system

Abstract

The invention discloses a path planning method and device, an unmanned aerial vehicle and a vehicle-mounted unmanned aerial vehicle system, and relates to the technical field of vehicle navigation. The method comprises the steps that firstly, in response to a parking space navigation instruction, a parking lot image is acquired, and the parking lot image is used for indicating a plurality of parking spaces in part or all areas of a parking lot; performing parking space detection on the parking lot image, and determining a target empty parking space from a plurality of empty parking spaces under the condition that the plurality of empty parking spaces exist in the parking lot image; all areas of the parking lot image are divided into passing areas and non-passing areas; the passing area comprises a target empty parking space, and the non-passing area comprises other empty parking spaces and occupied parking spaces in a plurality of parking spaces; and finally, in the passing area of the parking lot image, based on the initial position marked by the unmanned aerial vehicle, generating a navigation path taking the target empty parking space as the final position. Therefore, the influence of other empty parking spaces on the trafficability of the vehicle is avoided, the accuracy of path planning is improved, and the driving safety is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle navigation, in particular to a path planning method and device, an unmanned aerial vehicle and a vehicle-mounted unmanned aerial vehicle system. BACKGROUND

[0002] In order to alleviate the problem of parking, more and more open spaces are being transformed into open parking lots. However, the infrastructure of such parking lots is relatively backward, and the lack of monitoring and guidance of empty parking spaces makes it difficult for users to know whether there are empty parking spaces in the parking lot, and also makes it impossible to know the location of the empty parking spaces.

[0003] At present, although there is a method of using an unmanned aerial vehicle to assist parking, after the unmanned aerial vehicle receives an instruction, it searches for a target parking space that can be parked in among the collected parking lot images, and plans a navigation path for the vehicle to travel to the target parking space in the entire area of the parking lot image. Since there may be other empty parking spaces or obstacles in the navigation path, the vehicle may not be able to pass through the other empty parking spaces or obstacles during the process of traveling to the target parking space according to the navigation path, and the navigation performance is poor. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a path planning method and device, an unmanned aerial vehicle and a vehicle-mounted unmanned aerial vehicle system to improve the accuracy of path planning and ensure driving safety.

[0005] In order to achieve the above-mentioned purpose, the technical solutions adopted by the embodiments of the present application are as follows: In a first aspect, the present application provides a path planning method applied to an unmanned aerial vehicle, the method comprising: in response to a parking space navigation instruction, acquiring a parking lot image, the parking lot image being used to indicate a plurality of parking spaces in part or all of the area of the parking lot; performing parking space detection on the parking lot image, and in the case where there are a plurality of empty parking spaces in the parking lot image, determining a target empty parking space from the plurality of empty parking spaces; dividing all areas of the parking lot image into a passable area and a non-passable area; the passable area includes the target empty parking space, and the non-passable area includes other empty parking spaces and parking spaces that have been occupied in the plurality of parking spaces; in the passable area of the parking lot image, based on a starting position marked by the unmanned aerial vehicle, a navigation path is generated with the target empty parking space as an end position.

[0006] In an optional implementation, the parking space detection on the parking lot image comprises: inputting the parking lot image into a parking space recognition model to output a state of each parking space in the parking lot image, the state being occupied or unoccupied; the empty parking space is the unoccupied parking space; wherein the parking space recognition model is trained according to a first sample set and a second sample set; the first sample set comprises a plurality of empty parking space images, the plurality of empty parking space images comprising images captured at different perspectives of empty parking spaces; the second sample set comprises a plurality of occupied parking space images, the plurality of occupied parking space images comprising images captured at different perspectives of occupied parking spaces.

[0007] In an optional implementation, after the parking space detection on the parking lot image, the method further comprises: providing an interactive interface according to the parking space detection result of the parking lot image; the interactive interface comprises: a plurality of empty parking spaces marked from the plurality of parking spaces; the determination of the target empty parking space from the plurality of empty parking spaces comprises: determining the target empty parking space from the plurality of empty parking spaces in response to a parking space selection operation on the interactive interface; or selecting an empty parking space matching a preset parking demand from the plurality of empty parking spaces as the target empty parking space.

[0008] In an optional implementation, the division of all regions of the parking lot image into a passable region and a non-passable region comprises: marking the other empty parking spaces as obstacles in the parking lot image to obtain a marked image; performing semantic segmentation on the marked image to obtain a first image; in the first image, the target empty parking space and the road are marked as a first pattern, and the other empty parking spaces and the occupied parking spaces are marked as a second pattern; marking an image region corresponding to the first pattern in the parking lot image as the passable region, and marking an image region corresponding to the second pattern in the parking lot image as the non-passable region.

[0009] In an optional implementation, the division of all regions of the parking lot image into a passable region and a non-passable region comprises: performing semantic segmentation on the parking lot image to obtain a second image; in the second image, the target empty parking space, the other empty parking spaces, and the road are marked as different patterns from the occupied parking spaces; modifying the pattern corresponding to the other empty parking spaces to the pattern corresponding to the occupied parking spaces in the second image according to the positions of the other empty parking spaces in the parking lot image to obtain a third image; in the third image, the target empty parking space and the road are marked as a third pattern, and the other empty parking spaces and the occupied parking spaces are marked as a fourth pattern; marking an image region corresponding to the third pattern in the parking lot image as the passable region, and marking an image region corresponding to the fourth pattern in the parking lot image as the non-passable region.

[0010] In an optional implementation, the method further includes: in a case where there is one empty parking space in the parking lot image, taking the empty parking space as a target empty parking space; performing semantic segmentation on the parking lot image to obtain a fourth image; marking the target empty parking space and a road in the fourth image as a fifth pattern, and marking an occupied parking space in the plurality of parking spaces as a sixth pattern; marking an image region corresponding to the fifth pattern in the parking lot image as the passing region, and marking an image region corresponding to the sixth pattern in the parking lot image as the non-passing region.

[0011] In an optional implementation, before the generating, in the passing region of the parking lot image, a navigation path with the target empty parking space as an end position based on a starting position marked by the UAV, the method further includes: in a case where there is one empty parking space in the parking lot image, taking the empty parking space as a target empty parking space; performing semantic segmentation on the parking lot image to obtain a fourth image; marking the target empty parking space and a road in the fourth image as a fifth pattern, and marking an occupied parking space in the plurality of parking spaces as a sixth pattern; marking an image region corresponding to the fifth pattern in the parking lot image as the passing region, and marking an image region corresponding to the sixth pattern in the parking lot image as the non-passing region.

[0012] In a second aspect, the present application provides a path planning device applied to a UAV, the device comprising: an acquisition module configured to acquire a parking lot image in response to a parking space navigation instruction, the parking lot image being used to indicate a plurality of parking spaces in a part or whole region of a parking lot; a determination module configured to perform parking space detection on the parking lot image, and determine a target empty parking space from the plurality of empty parking spaces in a case where there are a plurality of empty parking spaces in the parking lot image; a division module configured to divide all regions of the parking lot image into a passing region and a non-passing region; the passing region comprising the target empty parking space, and the non-passing region comprising other empty parking spaces and occupied parking spaces in the plurality of parking spaces; and a planning module configured to generate, in the passing region of the parking lot image, a navigation path with the target empty parking space as an end position based on a starting position marked by the UAV.

[0013] In a third aspect, the present application provides a UAV comprising a processor and a memory, the memory storing a computer program, and the processor implementing the path planning method of any one of the preceding embodiments when executing the computer program.

[0014] In a fourth aspect, the present application provides a vehicle-mounted UAV system comprising a vehicle and the UAV of the preceding embodiments, and the vehicle and the UAV are connected through a mobile network or a wireless network.

[0015] The path planning method, device, unmanned aerial vehicle and vehicle-mounted unmanned aerial vehicle system provided by the embodiments of the present application avoid the influence of other empty parking spaces on the passability of the vehicle, improve the accuracy of path planning, and ensure driving safety.

[0016] In order to make the above objectives, characteristics and advantages of the present application more apparent, clear and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are referred to for detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 A schematic diagram of a vehicle-mounted unmanned aerial vehicle system provided by the embodiments of the present application is shown; Figure 2 One of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 3 An empty parking space example diagram provided by the embodiments of the present application is shown; Figure 4 Another of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 5 A third of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 6 A fourth of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 7 A fifth of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 8 A sixth of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 9 A seventh of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 10 An eighth of the flowcharts of the path planning method provided by the embodiments of the present application is shown; Figure 11 A functional module diagram of the path planning device provided by the embodiments of the present application is shown; Figure 12 A block diagram of a UAV is shown. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0020] Therefore, the detailed description of the embodiments of the present application provided in the drawings below is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] It should be noted that the relational terms such as "first" and "second" and the like are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus including a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus including the element.

[0022] In the related art, although there is a way of assisting parking by using the aerial view of a UAV, because the field of view of the UAV covers a wide range and the image resolution is limited, the passability of the vehicle will be affected due to the inability to identify small obstacles. For example, in some open parking lots, there are wheel blocking rods in the parking spaces, and the wheel blocking rods are not obvious in the view of the UAV, so it will be mistaken that there is no obstacle, and thus some areas that the vehicle cannot pass through are incorrectly classified as passable areas, thereby affecting the actual passability of the vehicle. Therefore, the embodiments of the present application provide a path planning method applied to a UAV to solve the above problem.

[0023] Please refer to Figure 1This is a schematic diagram of a vehicle-mounted unmanned aerial vehicle (UAV) system provided in an embodiment of this application. It includes a UAV 101 and a vehicle 102, and the UAV 101 and vehicle 102 communicate via a mobile network or a wireless network. The mobile network can be a wide area network provided by an operator based on cellular technology. The wireless network refers to a local area network based on non-cellular technology, such as Wi-Fi (Wireless Fidelity), Bluetooth, or other types of local area networks. This embodiment of the application does not limit the communication connection method between the UAV 101 and the vehicle 102.

[0024] Please see Figure 2 This is a flowchart illustrating a path planning method provided in an embodiment of this application. This path planning method is applied to a drone and can be executed by the drone itself, by a processor used in the drone, or by a vehicle-mounted drone system that includes the drone. For example, the vehicle-mounted drone system could be... Figure 1 The system shown. The path planning method includes steps 201 to 205.

[0025] Step 201: In response to the parking space navigation command, acquire the parking lot image.

[0026] In a first optional example, the parking lot image includes the entire area of ​​the parking lot and all parking spaces within that entire area. In a second optional example, the parking lot image includes a portion of the parking lot and all parking spaces within that portion of the parking lot. These two optional examples are alternative methods provided in the embodiments of this application and should not be construed as limiting this application.

[0027] In some embodiments, parking navigation instructions can be sent from the vehicle to the drone. For example, a user can perform an interactive operation on the vehicle's interface to send parking navigation instructions to the drone. In other embodiments, parking navigation instructions can be sent from a mobile terminal, such as a smartphone, to the drone. For example, when a user needs to find an available parking space for their vehicle, they can perform an interactive operation on the mobile terminal's interface to send parking navigation instructions to the drone. These interactive operations can be implemented through voice, interactive buttons, physical buttons, etc.

[0028] For step 201, the UAV flies to the initial point according to the received parking space navigation instruction, and flies from the initial point to each collection point in the preset path according to the preset flight parameter, and takes a picture at each collection point to obtain a local image corresponding to each collection point, and then splices all the local images to obtain a parking lot image. In some embodiments, the preset path can be set to surround the entire parking lot, so that the obtained parking lot image contains all the parking spaces of the parking lot. In some other embodiments, the preset path can also be set to surround part of the parking lot, so that the obtained parking lot image contains part of the parking spaces of the parking lot.

[0029] In some embodiments, the flight parameters can include flight height, flight speed, flight distance, hovering radius, and flight direction. The flight height can be in the range of 50m-200m; the flight speed can be in the range of 20m / s-20m / s; the flight distance can be in the range of 200m-500m; the hovering radius can be in the range of 50m-150m; and the flight direction can include front, back, left, right, left front, left back, right front, and right back. It should be understood that the specific values of the flight parameters can be set according to actual conditions, and the embodiments of the present application are not limited.

[0030] Step 202, detecting parking spaces in the parking lot image.

[0031] The parking space detection can be understood as a process of detecting whether there is a parking space in the image, and determining the position, state, etc. of the parking space. For example, the state of the parking space is occupied or unoccupied.

[0032] Step 203, in the case where there are multiple empty parking spaces in the parking lot image, determining a target empty parking space from the multiple empty parking spaces.

[0033] For step 202, the UAV can detect parking spaces in the parking lot image through a preset image processing algorithm or through a pre-trained network model. If it is detected that there is no empty parking space in the parking lot image, the UAV can provide an interactive interface containing no empty parking space information to let the user know that there is no empty parking space in the parking lot. If it is detected that there are multiple empty parking spaces in the parking lot image, one of the multiple empty parking spaces is selected as a target empty parking space. The target empty parking space can be understood as an empty parking space for the user to park.

[0034] For step 203, the way to determine the target empty parking space from the multiple empty parking spaces can be determined by random selection, according to a default setting, based on an interactive operation performed by the user, or according to a pre-set parking requirement.

[0035] Step 204, divide all areas of the parking lot image into a passing area and a non-passing area.

[0036] The passing area includes the target empty parking space, and the non-passing area includes other empty parking spaces and parking spaces that have been occupied in the plurality of parking spaces. The passing area can be understood as an area through which a vehicle can normally pass, and the non-passing area can be understood as an area in which an obstacle exists, causing the vehicle to be unable to pass, or an area in which the vehicle is not allowed to travel, such as a pedestrian walkway.

[0037] It can be understood that, based on the foregoing, since the wheel stop rod exists in the parking space, there is a problem of misrecognizing an area in which a vehicle cannot pass as a passable area. In order to facilitate understanding, the following will be described in conjunction with Figure 3 . Figure 3 is an example of an empty parking space provided by the present application, which includes an empty parking space 301 and an empty parking space 302, and the wheel stop rod 303 of the empty parking space 301 and the wheel stop rod 304 of the empty parking space 302 are both close to the side adjacent to both.

[0038] Since the wheel stop rod 303 and the wheel stop rod 304 are not obvious in the perspective of the unmanned aerial vehicle, it will be misrecognized that there is no obstacle between the empty parking space 301 and the empty parking space 302, that is, it will be misrecognized that the area between the two empty parking spaces is an area through which a vehicle can pass. If this area is divided into a passing area and used for path planning, it will affect the actual passability of the vehicle. Therefore, when the parking lot image is divided into areas, the target empty parking space is divided into a passing area, and other empty parking spaces and occupied parking spaces are divided into a non-passing area, so as to exclude the influence of other empty parking spaces on the passability of the vehicle.

[0039] For step 204, the unmanned aerial vehicle can divide the parking lot image into a passing area and a non-passing area through a preset image processing algorithm or through a pre-trained network model.

[0040] Step 205, in the passing area of the parking lot image, a navigation path with the target empty parking space as the end position is generated based on the starting position marked by the unmanned aerial vehicle.

[0041] For step 205, the unmanned aerial vehicle can generate a navigation path from the marked starting position to the target empty parking space through a preset path search algorithm based on the passing area.

[0042] In some embodiments, the unmanned aerial vehicle can send the navigation path to the vehicle, so that the user drives the vehicle to the target empty parking space through vehicle navigation to complete parking. In yet some embodiments, the unmanned aerial vehicle can send the navigation path to a mobile terminal, so that the user drives the vehicle to the target empty parking space according to the guidance of the mobile terminal to complete parking.

[0043] It can be understood that, by detecting parking spaces in the parking lot image, and in the case of multiple empty parking spaces, the target empty parking space is divided into a passing area and other empty parking spaces and occupied parking spaces are divided into a non-passing area, and a path to the target empty parking space is planned based on the passing area, the influence of other empty parking spaces on the passing ability of the vehicle is avoided, the accuracy of path planning is improved, and driving safety is ensured.

[0044] To improve the accuracy of parking space detection, for the above parking space detection process, an embodiment of the application provides a possible implementation. Please refer to Figure 4 , which is another flowchart of the path planning method provided by the embodiment of the application, which includes steps 401 to 405.

[0045] Step 401, in response to a parking space navigation instruction, a parking lot image is acquired.

[0046] For step 401, the basic principles and technical effects are the same as those of step 201 described above. Please refer to the corresponding contents in the above embodiments, which will not be described here.

[0047] Step 402, input the parking lot image into the parking space recognition model, and output the state of each parking space in the parking lot image, the state being occupied or unoccupied; the empty parking space is the unoccupied parking space.

[0048] The parking space recognition model is trained according to the first sample set and the second sample set; the first sample set includes a plurality of empty parking space images, and the plurality of empty parking space images include images taken at different angles of view of the empty parking space; the second sample set includes a plurality of occupied parking space images, and the plurality of occupied parking space images include images taken at different angles of view of the occupied parking space.

[0049] It can be understood that, due to the shooting angle of the unmanned aerial vehicle, the parking space in the image may be deformed, blocked, etc. Therefore, in order to improve the accuracy of parking space recognition, the parking space images taken at different angles of view are used as sample sets to train the network model.

[0050] In some embodiments, the first sample set and the second sample set can be used to train a neural network model, such as a YOLO (You Only Look Once, target detection) model, to obtain a parking space recognition model. The plurality of images in the first sample set are taken at different angles of view of the parking space when the parking space is unoccupied. In this way, the neural network model can learn the visual features of the empty parking space at various observation angles through the first sample set. The plurality of images in the second sample set are taken at different angles of view of the parking space when the parking space is occupied. In this way, the neural network model can learn various forms of occupied parking space through the second sample set.

[0051] For step 402, the parking lot image is input into the pre-trained parking space recognition model to make the parking space recognition model output the position and state of each parking space in the parking lot image, i.e., complete parking space recognition. The state is used to indicate whether the parking space is occupied, and the parking space that is not occupied is an empty parking space.

[0052] Step 403, in the case where there are multiple empty parking spaces in the parking lot image, a target empty parking space is determined from the multiple empty parking spaces.

[0053] Step 404, all regions of the parking lot image are divided into a passing region and a non-passing region.

[0054] The passing region includes the target empty parking space, and the non-passing region includes other empty parking spaces and parking spaces that have been occupied in the multiple parking spaces.

[0055] Step 405, in the passing region of the parking lot image, a navigation path with the target empty parking space as the end position is generated based on the starting position marked by the unmanned aerial vehicle.

[0056] For steps 403 to 405, the basic principles and the resulting technical effects are the same as those of steps 203 to 205 described above. For details, please refer to the corresponding content in the above embodiments, which will not be repeated here.

[0057] It can be understood that, by training the model through the sample set under multiple perspectives, the generalization ability of the parking space recognition model in actual application is improved, so that it can adapt to the image differences caused by the shooting angle in the parking lot environment, thereby improving the accuracy of parking space detection.

[0058] For the above process of determining the target empty parking space, the present embodiment provides a possible implementation manner, please refer to Figure 5 is another flowchart of the path planning method provided by the present embodiment, which includes steps 501 to 506.

[0059] Step 501, in response to a parking space navigation instruction, a parking lot image is acquired.

[0060] The parking lot image is used to indicate multiple parking spaces in part or all regions of the parking lot.

[0061] Step 502, parking space detection is performed on the parking lot image.

[0062] For steps 501 to 502, the basic principles and the resulting technical effects are the same as those of steps 201 to 202 described above. For details, please refer to the corresponding content in the above embodiments, which will not be repeated here.

[0063] At step 503, an interactive interface is provided according to the parking space detection result of the parking lot image.

[0064] The interactive interface includes a plurality of empty parking spaces from a plurality of parking spaces.

[0065] In some embodiments, the parking space detection result can include each empty parking space in the parking lot image, and then the UAV can directly send the interactive interface marked with the plurality of empty parking spaces to the vehicle or the mobile terminal. In yet some embodiments, the parking space detection result can include the state of each parking space in the parking lot image, and then the UAV can obtain a plurality of parking spaces with an unoccupied state according to the parking space detection result to obtain the plurality of empty parking spaces, and send the interactive interface marked with the plurality of empty parking spaces to the vehicle or the mobile terminal.

[0066] At step 504, a target empty parking space from the plurality of empty parking spaces is determined in response to a parking space selection operation on the interactive interface.

[0067] In some embodiments, the user can select the empty parking space that he or she wants to park in by performing a parking space selection operation in the interactive interface of the vehicle, and then the vehicle generates an instruction and sends it to the UAV to make the UAV determine the target empty parking space selected by the user from the plurality of empty parking spaces according to the instruction. Alternatively, the vehicle can also directly send the target empty parking space selected by the user to the UAV.

[0068] In yet some embodiments, the user can select the empty parking space that he or she wants to park in by performing a parking space selection operation in the interactive interface of the mobile terminal, and then the mobile terminal generates an instruction and sends it to the UAV to make the UAV determine the target empty parking space selected by the user from the plurality of empty parking spaces according to the instruction. Alternatively, the mobile terminal can also directly send the target empty parking space selected by the user to the UAV.

[0069] It can be understood that if the user thinks that there is no empty parking space in the interactive interface that he or she wants to park in, he or she can also perform an interactive operation in the interactive interface of the vehicle or the mobile terminal to make the vehicle or the mobile terminal send an instruction to the UAV to make the UAV reacquire the parking lot image.

[0070] At step 505, all areas of the parking lot image are divided into a passing area and a non-passing area.

[0071] The passing area includes the target empty parking space, and the non-passing area includes other empty parking spaces and occupied parking spaces in the plurality of parking spaces.

[0072] At step 506, a navigation path with the target empty parking space as an end position is generated in the passing area of the parking lot image based on the starting position marked by the UAV.

[0073] For steps 505 to 506, the basic principles and the technical effects generated are the same as those of steps 204 to 205 described above, and reference can be made to the corresponding content in the above embodiments, which will not be described here again.

[0074] It can be understood that the embodiments of the present application provide an interactive interface to let the user select the target empty parking space by himself, thereby improving the user experience and improving the flexibility of path planning.

[0075] For the above process of determining the target empty parking space, the embodiments of the present application provide another possible implementation manner, please refer to Figure 6 is another flowchart of the path planning method provided by the embodiments of the present application, which includes steps 601 to 605.

[0076] Step 601, in response to the parking space navigation instruction, obtaining a parking lot image.

[0077] The parking lot image is used to indicate a plurality of parking spaces in part or all of the area of the parking lot.

[0078] Step 602, performing parking space detection on the parking lot image.

[0079] For steps 601 to 602, the basic principles and the technical effects generated are the same as those of steps 201 to 202 described above, and reference can be made to the corresponding content in the above embodiments, which will not be described here again.

[0080] Step 603, in the case that there are a plurality of empty parking spaces in the parking lot image, selecting an empty parking space matching the preset parking demand from the plurality of empty parking spaces as the target empty parking space.

[0081] It can be understood that, in order to improve the response speed to improve the user experience, the embodiments of the present application can also pre-set the parking demand. In some embodiments, the parking demand can include a plurality of empty parking space requirements, such as the empty parking space closest to the entrance, the empty parking space closest to the exit, etc. In some other embodiments, the parking demand can include a plurality of empty parking space requirements and the priority of each empty parking space requirement, such as the priority of the empty parking space closest to the entrance is high, and the priority of the empty parking space closest to the exit is low. It should be understood that the empty parking space requirement can be set according to the actual situation, and the embodiments of the present application are not limited.

[0082] For step 603, in the case that the parking demand includes multiple empty parking space requirements, the UAV can select any empty parking space meeting the empty parking space requirement as the target empty parking space from the multiple empty parking spaces. In the case that the parking demand includes multiple empty parking space requirements and their priorities, the UAV can first select a candidate empty parking space meeting the empty parking space requirement from the multiple empty parking spaces, and if there is only one candidate empty parking space, the candidate empty parking space is selected as the target empty parking space; if there are multiple candidate empty parking spaces, the candidate empty parking space with the highest priority is selected as the target empty parking space.

[0083] It can be understood that if there is no empty parking space matching the preset parking demand in the multiple empty parking spaces, the UAV can also provide an interactive interface containing no matching parking space information, so as to let the user know that there is no empty parking space meeting the parking demand in the parking lot.

[0084] Step 604, dividing all regions of the parking lot image into a passing region and a non-passing region; the passing region includes the target empty parking space, and the non-passing region includes other empty parking spaces and occupied parking spaces in the multiple parking spaces.

[0085] Step 605, in the passing region of the parking lot image, a navigation path with the target empty parking space as the end position is generated based on the starting position marked by the UAV.

[0086] For steps 604 to 605, the basic principles and technical effects thereof are the same as those of steps 204 to 205 described above, and the corresponding contents in the above embodiments can be referred to, which will not be described here again.

[0087] It can be understood that the embodiments of the present application automatically select an empty parking space matching the parking demand as the target empty parking space by setting the parking demand, so as to realize intelligent decision of empty parking space selection and improve the intelligentization of path planning.

[0088] For the process of dividing the passing region and the non-passing region, the embodiments of the present application provide a possible implementation manner, please refer to Figure 7 is another flowchart of the path planning method provided by the embodiments of the present application, which includes steps 701 to 703.

[0089] Step 701, marking other empty parking spaces as obstacles in the parking lot image to obtain a marked image.

[0090] It can be understood that, in the related art, for the parking scene, the road and the empty parking space are usually identified as non-obstacles, and the occupied parking space is identified as an obstacle, so as to perform semantic segmentation. Based on the foregoing introduction, the adjacent multiple empty parking spaces will affect the actual passability of the vehicle, so the embodiment of the application will mark other empty parking spaces as obstacles, so that other empty parking spaces and occupied parking spaces are identified as obstacles when performing semantic segmentation, so as to divide the target empty parking space and other empty parking spaces into different categories.

[0091] In some embodiments, the unmanned aerial vehicle can obtain the marked image by performing a filling operation on the parking lot image. For example, the target region for indicating other empty parking spaces in the parking lot image can be filled with a preset pattern to indicate that the target region belongs to the obstacle category, thereby obtaining the marked image.

[0092] In still other embodiments, the unmanned aerial vehicle can obtain the marked image by adding labels. For example, a first character can be used as a label of the target region for indicating other empty parking spaces in the parking lot image to indicate that the target region belongs to the obstacle category; and a second character can be used as a label of the remaining region in the parking lot image except the target region to indicate that the remaining region is a region to be processed, thereby obtaining the marked image.

[0093] Step 702, performing semantic segmentation on the marked image to obtain a first image.

[0094] In the first image, the target empty parking space and the road are marked as a first pattern, and the other empty parking spaces and the occupied parking spaces are marked as a second pattern. The first pattern and the second pattern can be understood as different visual markers, for example, the first pattern and the second pattern can be represented by different gray values or different color values.

[0095] Step 703, marking the image region corresponding to the first pattern in the parking lot image as a passable region, and marking the image region corresponding to the second pattern in the parking lot image as a non-passable region.

[0096] For step 702, the unmanned aerial vehicle can use a preset image processing algorithm or a pre-trained network model to perform semantic segmentation on the marked image to obtain the first image. The first image represents different categories by the first pattern and the second pattern. The target empty parking space and the road are marked as the first pattern, i.e., they are divided into non-obstacles; the other empty parking spaces and the occupied parking spaces are marked as the second pattern, i.e., they are divided into obstacles.

[0097] For step 703, the UAV can determine the image region corresponding to the first pattern in the parking lot image, that is, obtain the passing region, and determine the image region corresponding to the second pattern in the parking lot image, that is, obtain the non-passing region, by image fusion on the parking lot image and the first image.

[0098] It can be understood that, by marking other empty parking spaces as obstacles, the embodiment of the application divides other empty parking spaces into the obstacle category in semantic segmentation, and then divides other empty parking spaces into the non-passing region, thereby excluding the influence of other empty parking spaces on the passing ability of vehicles, improving the accuracy of path planning, and ensuring driving safety.

[0099] For the above process of dividing the passing region and the non-passing region, the embodiment of the application provides another possible implementation manner, please refer to Figure 8 is another flowchart of the path planning method provided by the embodiment of the application, which includes steps 801 to 803.

[0100] Step 801: performing semantic segmentation on the parking lot image to obtain a second image.

[0101] In the second image, the target empty parking space, the other empty parking space, and the road are marked as different patterns.

[0102] For step 801, the UAV can use a preset image processing algorithm or a pre-trained network model to perform semantic segmentation on the marked image to obtain the second image. The second image represents different categories by different patterns. The target empty parking space, the other empty parking space, and the road are marked as the same pattern, that is, they are divided into non-obstacles; the occupied parking space is marked as another pattern, that is, it is divided into obstacles.

[0103] Step 802: modifying the pattern corresponding to the other empty parking space to the pattern corresponding to the occupied parking space in the second image according to the position of the other empty parking space in the parking lot image, to obtain a third image.

[0104] In the third image, the target empty parking space and the road are marked as a third pattern, and the other empty parking space and the occupied parking space are marked as a fourth pattern. The third pattern and the fourth pattern can be understood as different visual markers, such as different gray values or different color values.

[0105] Step 803: marking the image region corresponding to the third pattern in the parking lot image as the passing region, and marking the image region corresponding to the fourth pattern in the parking lot image as the non-passing region.

[0106] It can be understood that, based on the foregoing, the adjacent multiple empty parking spaces will affect the actual passability of the vehicle, so the embodiment of the application will also modify the pattern in the second image after obtaining the second image, to modify the category of other empty parking spaces from non-obstacles to obstacles.

[0107] For step 802, the unmanned aerial vehicle can obtain a target region representing other empty parking spaces in the parking lot image according to the positions of other empty parking spaces in the parking lot image, and then determine an image region corresponding to the target region in the second image by image fusion of the parking lot image and the second image, to obtain a region to be modified, and then modify the pattern of the region to be modified to another pattern, to obtain a third image. The third image represents different categories through a third pattern and a fourth pattern. The target empty parking space and the road are marked as the third pattern, i.e., they are divided into non-obstacles; other empty parking spaces and occupied parking spaces are marked as the fourth pattern, i.e., they are divided into obstacles.

[0108] For step 803, the unmanned aerial vehicle can determine an image region corresponding to the third pattern in the parking lot image, i.e., obtain a passable region, and determine an image region corresponding to the fourth pattern in the parking lot image, i.e., obtain a non-passable region, by image fusion of the parking lot image and the first image.

[0109] It can be understood that, by the method of semantic segmentation and then pattern modification, the embodiment of the application divides other empty parking spaces into the category of obstacles, and then divides other empty parking spaces into non-passable regions, thereby excluding the influence of other empty parking spaces on the passability of the vehicle, improving the accuracy of path planning, and ensuring driving safety.

[0110] For the above process of generating a navigation path, the embodiment of the application provides a possible implementation manner, please refer to Figure 9 is another flowchart of the path planning method provided by the embodiment of the application, which includes steps 901 to 903.

[0111] Step 901: determining a geographical region corresponding to the passable region according to the position of the passable region in the parking lot image.

[0112] In some embodiments, the position of the passable region in the parking lot image can be represented by pixel coordinates of the passable region in the image coordinate system, and the geographical coordinates of the passable region in the world coordinate system can be calculated through coordinate conversion to obtain the geographical region corresponding to the passable region, which can be represented by a longitude range and a latitude range.

[0113] Step 902: determining the geographical position of the target empty parking space according to the position of the target empty parking space in the parking lot image.

[0114] In some embodiments, the pixel coordinates of the target empty parking space in the image coordinate system can be used to represent the position of the target empty parking space in the parking lot image, and the geographic coordinates of the target empty parking space in the world coordinate system can be calculated through coordinate conversion to obtain the geographic position of the target empty parking space, which can be represented by latitude and longitude.

[0115] Step 903, generating a navigation path according to the geographic region, the geographic position of the target empty parking space, and the starting position.

[0116] In some embodiments, the UAV can perform path planning in the geographic region based on the starting position and the geographic position of the target empty parking space to obtain at least one candidate path through a preset path search algorithm. In the case of only one candidate path, the candidate path is taken as the navigation path. In the case of multiple candidate paths, the candidate path with the shortest distance can be taken as the navigation path. Alternatively, the travel time of each candidate path can be calculated, and the candidate path with the shortest travel time can be taken as the navigation path.

[0117] It can be understood that the embodiments of the present application improve the accuracy of the navigation path by converting the travel area into a geographic region and converting the position of the target empty parking space in the image into a geographic position for path planning.

[0118] After detecting the parking spaces in the parking lot image, the embodiments of the present application also provide a possible implementation. Please refer to Figure 10 is another flowchart of the path planning method provided by the embodiments of the present application, which includes steps 1001 to 1006.

[0119] Step 1001, acquiring a parking lot image in response to a parking space navigation instruction.

[0120] The parking lot image is used to indicate a plurality of parking spaces in part or all of the region of the parking lot.

[0121] Step 1002, detecting parking spaces in the parking lot image.

[0122] For steps 1001 to 1002, the basic principles and technical effects are the same as those of steps 101 to 102 described above, and the corresponding contents in the above embodiments can be referred to, which will not be described here.

[0123] Step 1003, in the case that there is an empty parking space in the parking lot image, taking the empty parking space as a target empty parking space.

[0124] For step 1003, if it is determined that there is only one empty parking space in the parking lot image after the parking space detection on the parking lot image, the empty parking space can be taken as the target empty parking space. In some embodiments, the UAV can also send an interactive interface marked with the empty parking space to the vehicle or the mobile terminal, so as to make the user know that there is only one empty parking space. If the user does not want to select the empty parking space, the user can make the vehicle or the mobile terminal send an instruction to the UAV by performing an interactive operation in the interactive interface of the vehicle or the mobile terminal, so as to make the UAV reacquire the parking lot image.

[0125] In yet some embodiments, if the empty parking space does not match the preset parking demand, the UAV can send an interactive interface containing the no-matching parking space information to the vehicle or the mobile terminal, so as to make the user know that there is no empty parking space in the parking lot that matches the parking demand of the user.

[0126] Step 1004, performing semantic segmentation on the parking lot image to obtain a fourth image.

[0127] In the fourth image, the target empty parking space and the road are marked as a fifth pattern, and the occupied parking spaces in the plurality of parking spaces are marked as a sixth pattern.

[0128] Step 1005, marking an image region corresponding to the fifth pattern in the parking lot image as a passable region, and marking an image region corresponding to the sixth pattern in the parking lot image as a non-passable region.

[0129] For step 1004, the UAV can use a preset image processing algorithm or a pre-trained network model to perform semantic segmentation on the parking lot image to obtain the fourth image. The fourth image represents different categories through the fifth pattern and the sixth pattern. The target empty parking space and the road are marked as the fifth pattern, i.e., they are classified as non-obstacles; the occupied parking spaces are marked as the sixth pattern, i.e., they are classified as obstacles.

[0130] For step 1005, the UAV can determine the image region corresponding to the fifth pattern in the parking lot image by image fusion on the parking lot image and the fourth image, i.e., obtain the passable region, and determine the image region corresponding to the sixth pattern in the parking lot image, i.e., obtain the non-passable region.

[0131] Step 1006, generating a navigation path with the target empty parking space as an end position based on the starting position marked by the UAV in the passable region of the parking lot image.

[0132] For step 1006, the basic principles and the technical effects are the same as those of step 105 described above, and the corresponding contents in the above embodiments can be referred to, which will not be described herein again.

[0133] It can be understood that, in the case of only one empty parking space, the embodiment of the application can quickly divide the passing area and the non-passing area and perform path planning through one semantic segmentation, thereby improving the efficiency of path planning.

[0134] In order to perform the corresponding steps in the above-mentioned embodiments and various possible manners, an implementation manner of the path planning method device is given below. Please refer to Figure 11 , which is a functional module diagram of the path planning method device 1100 provided by the embodiment of the application. It should be noted that the path planning method device 1100 provided by the embodiment has the same basic principles and technical effects as the above-mentioned embodiments, and for brief description, the embodiment is not mentioned in the above-mentioned embodiments. For brief description, the corresponding contents in the above-mentioned embodiments can be referred to. The path planning method device 1100 comprises: The acquisition module 1101 is configured to acquire a parking lot image in response to a parking space navigation instruction, and the parking lot image is configured to indicate a plurality of parking spaces in part or all areas of the parking lot.

[0135] The determination module 1102 is configured to perform parking space detection on the parking lot image, and in the case that there are a plurality of empty parking spaces in the parking lot image, determine a target empty parking space from the plurality of empty parking spaces.

[0136] The division module 1103 is configured to divide all areas of the parking lot image into a passing area and a non-passing area; the passing area comprises the target empty parking space, and the non-passing area comprises other empty parking spaces and parking spaces that have been occupied in the plurality of parking spaces.

[0137] The planning module 1104 is configured to generate a navigation path with the target empty parking space as an end position based on a starting position marked by the unmanned aerial vehicle in the passing area of the parking lot image.

[0138] Optionally, the determination module 1102 is specifically configured to: input the parking lot image into a parking space recognition model, output the state of each parking space in the parking lot image, and the state is occupied or unoccupied; the empty parking space is the unoccupied parking space; wherein the parking space recognition model is trained according to a first sample set and a second sample set; the first sample set comprises a plurality of empty parking space images, and the plurality of empty parking space images comprise images taken at different angles of view of the empty parking space; the second sample set comprises a plurality of occupied parking space images, and the plurality of occupied parking space images comprise images taken at different angles of view of the occupied parking space.

[0139] Optionally, the determination module 1102 is specifically configured to: provide an interactive interface according to the parking space detection result of the parking lot image; the interactive interface comprises: a plurality of empty parking spaces marked from a plurality of parking spaces; in response to a parking space selection operation on the interactive interface, determine a target empty parking space in the plurality of empty parking spaces; or select an empty parking space matching a preset parking demand from the plurality of empty parking spaces as the target empty parking space.

[0140] Optionally, the dividing module 1103 is specifically configured to: mark other empty parking spaces as obstacles in the parking lot image to obtain a marked image; perform semantic segmentation on the marked image to obtain a first image; in the first image, the target empty parking space and the road are marked as a first pattern, and the other empty parking spaces and the occupied parking spaces are marked as a second pattern; mark an image region corresponding to the first pattern in the parking lot image as a passing region, and mark an image region corresponding to the second pattern in the parking lot image as a non-passing region.

[0141] Optionally, the dividing module 1103 is specifically configured to: perform semantic segmentation on the parking lot image to obtain a second image; in the second image, the target empty parking space, the other empty parking spaces, and the road and the occupied parking spaces are marked as different patterns; according to the positions of the other empty parking spaces in the parking lot image, modify the patterns corresponding to the other empty parking spaces in the second image to the pattern corresponding to the occupied parking spaces to obtain a third image; in the third image, the target empty parking space and the road are marked as a third pattern, and the other empty parking spaces and the occupied parking spaces are marked as a fourth pattern; mark an image region corresponding to the third pattern in the parking lot image as a passing region, and mark an image region corresponding to the fourth pattern in the parking lot image as a non-passing region.

[0142] Optionally, the planning module 1104 is specifically configured to: determine a geographical region corresponding to the passing region according to the position of the passing region in the parking lot image; determine a geographical position of the target empty parking space according to the position of the target empty parking space in the parking lot image; and generate the navigation path according to the geographical region, the geographical position of the target empty parking space, and the starting position.

[0143] Optionally, the determining module 1102 is further configured to: in a case where there is one empty parking space in the parking lot image, take the empty parking space as the target empty parking space; and the dividing module 1103 is further configured to: perform semantic segmentation on the parking lot image to obtain a fourth image; in the fourth image, the target empty parking space and the road are marked as a fifth pattern, and the occupied parking spaces in the plurality of parking spaces are marked as a sixth pattern; mark an image region corresponding to the fifth pattern in the parking lot image as a passing region, and mark an image region corresponding to the sixth pattern in the parking lot image as a non-passing region.

[0144] The embodiments of the present application also provide a UAV, please refer to Figure 12 is a block schematic diagram of the UAV provided by the embodiments of the present application. The UAV 1200 includes a processor 1201, a memory 1202, and a communication module 1203, and each element is electrically connected to each other directly or indirectly to realize the transmission or interaction of data. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines.

[0145] The processor 1201 is used to read / write data or programs stored in the memory 1202 and perform corresponding functions. It can be a general-purpose processor, including CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0146] The memory 1202 is used to store programs or data, and it can be RAM (Random Access Memory), ROM (Read Only Memory), PROM (Programmable Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electric Erasable Programmable Read-Only Memory), etc.

[0147] The communication module 1203 is used to communicate with other devices.

[0148] Understandable, Figure 12 The structure shown is only a schematic diagram of the UAV 1200. The UAV 1200 may also include components that are larger than... Figure 12 The more or fewer components shown, or having the same Figure 12 The different configurations shown. Figure 12 The components shown can be implemented using hardware, software, or a combination thereof. For example, to achieve the corresponding functions of the drone, the drone 1200 may also include a navigation module, a shooting module, a lidar module, etc.

[0149] The UAV provided in this application embodiment has a memory that stores a computer program. When the processor executes the computer program, it implements the path planning method disclosed in this application embodiment.

[0150] This application also provides a vehicle-mounted unmanned aerial vehicle (UAV) system, which includes a vehicle and the UAV provided in this application embodiment. For example, the vehicle-mounted UAV could be... Figure 1 The system shown.

[0151] The embodiment of the present application further provides a storage medium, which stores a computer program, and the computer program is executed by a processor to realize the path planning method disclosed by the embodiment of the present application.

[0152] In several embodiments provided by the present application, it should be understood that the disclosed apparatus and method can also be implemented by other manners. The apparatus embodiments described above are merely schematic, for example, the flow charts and block diagrams in the drawings show the possible implementation architecture, function and operation of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flow chart or block diagram can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logic function. It should also be noted that in some alternative implementation manners, the functions noted in the blocks can also occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can also be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flow chart, and the combination of blocks in the block diagram and / or flow chart, can be implemented by a special hardware-based system for executing the specified function or action, or can be implemented by a combination of special hardware and computer instructions.

[0153] In addition, each functional module in the embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0154] If the functions are realized in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0155] The above descriptions are only the preferred embodiments of the present application, and are not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A path planning method, characterized in that, Applied to drones, the method includes: In response to a parking space navigation command, an image of the parking lot is acquired, the parking lot image being used to indicate multiple parking spaces in part or all of the parking lot area; Parking space detection is performed on the parking lot image, and if there are multiple empty parking spaces in the parking lot image, a target empty parking space is determined from the multiple empty parking spaces; The parking lot image is divided into a passable area and a non-passable area; the passable area includes the target empty parking space, and the non-passable area includes other empty parking spaces and the parking spaces that are already occupied among the multiple parking spaces; Within the traversable area of ​​the parking lot image, a navigation path is generated with the target empty parking space as the endpoint, based on the starting position marked by the drone.

2. The path planning method according to claim 1, characterized in that, The step of detecting parking spaces in the parking lot image includes: The parking lot image is input into the parking space recognition model, and the status of each parking space in the parking lot image is output, which is either occupied or unoccupied; the empty parking space is an unoccupied parking space. The parking space recognition model is trained based on a first sample set and a second sample set. The first sample set includes multiple images of empty parking spaces, which are images taken from different perspectives of the empty parking spaces. The second sample set includes multiple images of occupied parking spaces, which are images taken from different perspectives of the occupied parking spaces.

3. The path planning method according to claim 1, characterized in that, After performing parking space detection on the parking lot image, the method further includes: Based on the parking space detection results of the parking lot image, an interactive interface is provided; the interactive interface includes: multiple empty parking spaces marked from the multiple parking spaces; The step of determining the target empty parking space from the plurality of empty parking spaces includes: In response to a parking space selection operation on the interactive interface, a target available parking space is determined from the plurality of available parking spaces; or, Select the empty parking space that matches the preset parking demand from the plurality of empty parking spaces as the target empty parking space.

4. The path planning method according to claim 1, characterized in that, The step of dividing all areas of the parking lot image into passable and non-passable areas includes: In the parking lot image, the other empty parking spaces are marked as obstacles to obtain a marked image; The marked image is semantically segmented to obtain a first image; in the first image, the target empty parking space and the road are marked as a first pattern, and the other empty parking spaces and the occupied parking spaces are marked as a second pattern. The image area in the parking lot image corresponding to the first pattern is marked as the passage area, and the image area in the parking lot image corresponding to the second pattern is marked as the non-passage area.

5. The path planning method according to claim 1, characterized in that, The step of dividing all areas of the parking lot image into passable and non-passable areas includes: The parking lot image is semantically segmented to obtain a second image; in the second image, the target empty parking space, the other empty parking spaces, and the road are marked with different patterns than the occupied parking spaces. Based on the location of other empty parking spaces in the parking lot image, the pattern corresponding to the other empty parking spaces in the second image is modified to the pattern corresponding to the occupied parking space to obtain a third image; in the third image, the target empty parking space and the road are marked as the third pattern, and the other empty parking spaces and the occupied parking spaces are marked as the fourth pattern. The image area in the parking lot image corresponding to the third pattern is marked as the passage area, and the image area in the parking lot image corresponding to the fourth pattern is marked as the non-passage area.

6. The path planning method according to claim 1, characterized in that, In the passage area of ​​the parking lot image, based on the starting position marked by the drone, a navigation path is generated with the target empty parking space as the destination, including: Based on the location of the passage area in the parking lot image, determine the geographical region corresponding to the passage area; Based on the location of the target empty parking space in the parking lot image, determine the geographical location of the target empty parking space; The navigation path is generated based on the geographical region, the geographical location of the target empty parking space, and the starting position.

7. The path planning method according to claim 1, characterized in that, Before generating a navigation path with the target empty parking space as the endpoint, based on the starting position marked by the drone within the traffic area of ​​the parking lot image, the method further includes: If there is an empty parking space in the parking lot image, the empty parking space is taken as the target empty parking space; The parking lot image is semantically segmented to obtain a fourth image; in the fourth image, the target empty parking space and the road are marked as a fifth pattern, and the occupied parking space among the multiple parking spaces is marked as a sixth pattern; The image area in the parking lot image corresponding to the fifth pattern is marked as the passage area, and the image area in the parking lot image corresponding to the sixth pattern is marked as the non-passage area.

8. A path planning device, characterized in that, The device, applied to drones, includes: The acquisition module is used to acquire a parking lot image in response to a parking space navigation command. The parking lot image is used to indicate multiple parking spaces in part or all of the parking lot area. The determination module is used to detect parking spaces in the parking lot image and, if there are multiple empty parking spaces in the parking lot image, determine the target empty parking space from the multiple empty parking spaces; A segmentation module is used to divide all areas of the parking lot image into a passable area and a non-passable area; the passable area includes the target empty parking space, and the non-passable area includes other empty parking spaces and the parking spaces that are already occupied among the multiple parking spaces; The planning module is used to generate a navigation path with the target empty parking space as the destination, based on the starting position marked by the drone in the traffic area of ​​the parking lot image.

9. A drone, characterized in that, It includes a processor and a memory, the memory storing a computer program, and when the processor executes the computer program, it implements the path planning method according to any one of claims 1-7.

10. A vehicle-mounted unmanned aerial vehicle system, characterized in that, include: The vehicle and the drone as described in claim 9; The vehicle and the drone communicate with each other via a mobile network or wireless network.

Images