Target detection method, device and equipment based on aerial image of unmanned aerial vehicle

Abstract

The present application relates to the technical field of image detection, and particularly relates to a target detection method for aerial images based on a UAV, which comprises the following steps: acquiring a plurality of aerial images of a UAV and corresponding label data of each aerial image; performing image conversion processing on each aerial image according to a plurality of preset conversion type methods, respectively, to obtain a plurality of converted images corresponding to each aerial image and a conversion type identifier corresponding to each converted image; converting position data corresponding to a target pixel according to the conversion type identifier of the converted image to obtain label data of the converted image after conversion according to the corresponding conversion type; and inputting the plurality of aerial images, the corresponding label data of each aerial image, the plurality of converted images obtained after image conversion processing of each aerial image according to the plurality of preset conversion type methods, and the corresponding label data of each converted image into a neural network to be trained to obtain a target detection model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image detection, and in particular to a target detection method and device based on aerial images of unmanned aerial vehicles, an equipment and a storage medium. BACKGROUND

[0002] Currently, aerial images of unmanned aerial vehicles often encounter various problems, such as fog, mountain shadow, too dark image, too exposed image, too small target, complex background, large field of view, rotation, and a series of problems, and are affected by factors such as shooting height, flight speed, weather, different reflection angles, uneven light uniformity, and electromagnetic interference, resulting in color distortion, many noise points, and image light and dark, which seriously affects the quality of aerial images, weakens or blocks target objects, and makes it difficult to obtain important feature information of image data.

[0003] In addition, in the aerial images of unmanned aerial vehicles, there are often a large number of similar objects in the target dense area, which increases the missed detection or false positives in detection, thereby reducing the accuracy and efficiency of target detection of aerial images. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a target detection method and device based on aerial images of unmanned aerial vehicles, an equipment and a storage medium, to obtain conversion images after image conversion processing of a plurality of conversion type methods corresponding to a plurality of aerial images of unmanned aerial vehicles, and label data of each conversion image corresponding to the conversion type identifier, to convert the position data corresponding to the target pixel associated with the target detection object according to the conversion type identifier, to obtain the label data of the conversion image converted according to the corresponding conversion type, to improve the accuracy and efficiency of the label data acquisition, and to use the label data corresponding to the aerial image, a plurality of conversion images after image conversion processing of each of the aerial images according to a plurality of conversion type methods, and label data corresponding to each of the conversion images as training data, to train the neural network model, to improve the accuracy, robustness and generalization ability of the trained neural network model, and to improve the accuracy and efficiency of target detection of aerial images.

[0005] In a first aspect, the present application provides a target detection method based on aerial images of unmanned aerial vehicles, comprising the following steps:

[0006] Obtaining a plurality of aerial images of unmanned aerial vehicles and label data corresponding to each of the aerial images, wherein the aerial images are images including a preset target detection object, and the label data includes a plurality of position data corresponding to target pixels associated with the target detection object;

[0007] convert each of the aerial images according to a preset conversion type method to obtain a plurality of converted images corresponding to each of the aerial images and a conversion type identifier corresponding to each of the converted images;

[0008] convert, according to the conversion type identifier of the converted image, position data corresponding to a target pixel associated with the target detection object to obtain labeled data of the converted image converted according to the corresponding conversion type;

[0009] input the plurality of aerial images, the labeled data corresponding to each of the aerial images, the plurality of converted images obtained by converting each of the aerial images according to the preset conversion type method, and the labeled data corresponding to each of the converted images into a neural network to be trained to obtain a target detection model;

[0010] In response to a detection instruction, obtain a to-be-detected aerial image of a UAV, input the to-be-detected aerial image into the target detection model, and obtain a target detection result corresponding to the to-be-detected aerial image.

[0011] In a second aspect, an embodiment of the present application provides a target detection device based on aerial images of a UAV, comprising:

[0012] The obtaining module is configured to obtain a plurality of aerial images of a UAV and labeled data corresponding to each of the aerial images, wherein the aerial images are images including a preset target detection object, and the labeled data includes position data corresponding to a plurality of target pixels associated with the target detection object;

[0013] The image conversion module is configured to convert each of the aerial images according to a preset conversion type method to obtain a plurality of converted images corresponding to each of the aerial images and a conversion type identifier corresponding to each of the converted images;

[0014] The labeled data conversion module is configured to convert, according to the conversion type identifier of the converted image, position data corresponding to a target pixel associated with the target detection object to obtain labeled data of the converted image converted according to the corresponding conversion type;

[0015] The training module is configured to input the plurality of aerial images, the labeled data corresponding to each of the aerial images, the plurality of converted images obtained by converting each of the aerial images according to the preset conversion type method, and the labeled data corresponding to each of the converted images into a neural network to be trained to obtain a target detection model;

[0016] The detection module is configured to, in response to a detection instruction, acquire a to-be-detected aerial image of the UAV, input the to-be-detected aerial image into the target detection model, and acquire a target detection result corresponding to the to-be-detected aerial image.

[0017] In a third aspect, an embodiment of the present application provides a computer device, including a processor, a memory, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, the steps of the target detection method based on aerial images of a UAV according to the first aspect are implemented.

[0018] In a fourth aspect, an embodiment of the present application provides a storage medium, which stores a computer program, and when the computer program is executed by a processor, the steps of the target detection method based on aerial images of a UAV according to the first aspect are implemented.

[0019] In the embodiment of the present application, a target detection method, device, equipment and storage medium based on aerial images of a UAV are provided, a converted image obtained by performing image conversion processing on aerial images of a plurality of UAVs according to a plurality of conversion type methods and a conversion type identifier corresponding to each converted image are acquired, the position data corresponding to the target pixels associated with a target detection object are converted according to the conversion type identifier, the labeled data of the converted image converted according to the corresponding conversion type is obtained, the accuracy and efficiency of the labeled data acquisition are improved, the labeled data corresponding to the aerial images, a plurality of converted images obtained by performing image conversion processing on each of the aerial images according to a plurality of preset conversion type methods, and the labeled data corresponding to each of the converted images are used as training data to train a neural network model, the accuracy, robustness and generalization ability of the trained neural network model are improved, and thus the accuracy and efficiency of the target detection of the aerial images are improved.

[0020] For better understanding and implementation, the present application is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 A flowchart of a target detection method based on aerial images of a UAV according to an embodiment of the present application is provided.

[0022] Figure 2 A flowchart of S1 in a target detection method based on aerial images of a UAV according to an embodiment of the present application is provided.

[0023] Figure 3 A flowchart of S3 in a target detection method based on aerial images of a UAV according to an embodiment of the present application is provided.

[0024] Figure 4A flowchart of S4 in a target detection method based on aerial images of a UAV provided by an embodiment of the present application is shown in FIG. 4;

[0025] Figure 5 A structure diagram of a target detection device based on aerial images of a UAV provided by an embodiment of the present application is shown in FIG. 5;

[0026] Figure 6 A structure diagram of a computer device provided by an embodiment of the present application is shown in FIG. 6. DETAILED DESCRIPTION

[0027] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, the same numbers are used to indicate the same or similar components. The embodiments described in the following exemplary embodiments do not represent all the embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0028] The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used in the present application and the appended claims, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0029] It should be understood that although the terms first, second, third, etc. can be employed in this application to describe various information, such information should not be limited by these terms. These terms are only used to distinguish one piece of information from another. For example, a first information can also be termed a second information without departing from the scope of the present application. Similarly, a second information can also be termed a first information. The word "if" can be interpreted to mean "when" or "upon" or "in response to determining" depending on the context.

[0030] Reference will now be made to Figure 1 , Figure 1 A flowchart of a target detection method based on aerial images of a UAV provided by an embodiment of the present application is shown in FIG. 3, which includes the following steps:

[0031] S1: Obtain a plurality of aerial images of a UAV and corresponding label data of each aerial image.

[0032] The execution subject of the unmanned aerial vehicle-based aerial image target detection method is a detection device (hereinafter referred to as a detection device) of the unmanned aerial vehicle-based aerial image target detection method. In an optional embodiment, the detection device can be a computer device, which can be a server or a server cluster formed by multiple computer devices.

[0033] The aerial image is an image including a preset target detection object. In this embodiment, the detection device controls the unmanned aerial vehicle to perform flight operations on a preset inspection route to obtain video collection data on the inspection route.

[0034] In order to improve the operation efficiency, the detection device clips the video collection data to remove video collection data that does not include the target detection object, obtains the video collection data after clipping, performs frame extraction processing on the video collection data after clipping, and obtains a plurality of images including the preset target detection object as the aerial image. The target detection object can be a person, a vehicle, a ship, a building, etc.

[0035] The marking data includes position data corresponding to a plurality of target pixels associated with the detection object. For example, the marking data can be an XML file. By setting a bndbox (feature information box), each detection object in the aerial image is framed with a rectangular information box, and the pixels of the aerial image corresponding to the upper left corner and the lower right corner of the rectangular information box are extracted to obtain the position data corresponding to the pixels as the marking data corresponding to the aerial image.

[0036] Please refer to Figure 2 , Figure 2 The flowchart of S1 of the unmanned aerial vehicle-based aerial image target detection method provided in an embodiment of the present application further includes the following step S101.

[0037] S101: Perform interference judgment processing on the plurality of aerial images to extract target aerial images from the plurality of aerial images.

[0038] In this embodiment, the detection device performs interference judgment processing on the plurality of aerial images to extract target aerial images from the plurality of aerial images.

[0039] Specifically, the detection device binarizes the plurality of aerial images, extracts object contours from the binarized aerial images, determines whether a pixel in the binarized aerial image is a disturbance pixel by judging the number of same pixels around the pixel according to the thickness of the contour, and determines that the aerial image is a disturbance image if the number of disturbance pixels in the aerial image is greater than or equal to a preset threshold of the number of disturbance pixels, and discards the disturbance image; if the number of disturbance pixels in the aerial image is less than the preset threshold of the number of disturbance pixels, the aerial image is determined to be a non-disturbance image, which is used as the target aerial image for extraction.

[0040] S2: performing image conversion processing on each of the aerial images according to a plurality of preset conversion type methods to obtain a plurality of converted images corresponding to each of the aerial images and a conversion type identifier corresponding to each of the converted images.

[0041] In this embodiment, the detection device performs image conversion processing on each of the aerial images according to a plurality of preset conversion type methods to obtain a plurality of converted images corresponding to each of the aerial images, and obtains a conversion type identifier corresponding to each of the converted images according to the conversion type of the converted image, wherein the image conversion processing step includes scaling, rotation, tilting, flipping, gray scale adjustment, contrast adjustment, Gaussian blur, and median blur processing.

[0042] Specifically, the scaling aerial image size to obtain a scaled image, stretching the aerial image size to obtain a stretched image, adjusting the tilt angle to obtain a tilted image, adjusting the gray scale value of the image to obtain different gray scale adjustment images, flipping the image horizontally and vertically to obtain flipped images at different positions, adjusting the contrast of the image to obtain contrast adjustment images at different contrasts, adjusting by denoising and filtering methods to eliminate image noise and obtain a smooth image, Gaussian blur to obtain a Gaussian blur image, and median blur to obtain a median blur image.

[0043] S3: converting the position data corresponding to the target pixel associated with the target detection object according to the conversion type identifier of the converted image to obtain labeled data of the converted image converted according to the corresponding conversion type.

[0044] In this embodiment, the prediction device converts the position data corresponding to the target pixel associated with the target detection object according to the conversion type identifier of the converted image to obtain labeled data of the converted image converted according to the corresponding conversion type.

[0045] In an optional embodiment, the mark data is a tree structure, including a root node, the root node is provided with a connected sub-node, each sub-node is provided with a branched sub-element, the sub-node stores identity of a plurality of detected objects in the aerial image, and the sub-element stores position data corresponding to target pixels associated with the detected objects.

[0046] Referring to Figure 3 , Figure 3 The flowchart of S3 in the target detection method based on aerial image of unmanned aerial vehicle provided by an embodiment of the present application is shown in FIG. 3, including steps S301-S302, which are specifically as follows.

[0047] S301: Obtain the identity of the target detection object input by the user, search from the root node of the mark data of the aerial image according to the identity of the target detection object, obtain the target sub-node matched with the identity of the target detection object, and obtain the position data corresponding to the target pixels associated with the target detection object from the sub-elements branched from the target sub-node.

[0048] In order to improve the efficiency of obtaining the mark data of the target detection object, in the embodiment, the detection device obtains the identity of the target detection object input by the user, wherein the identity is a unique IP identity, which can be a number, a letter or the like.

[0049] According to the identity of the target detection object, search from the root node of the mark data of the aerial image, obtain the target sub-node matched with the identity of the target detection object, and obtain the position data corresponding to the target pixels associated with the target detection object from the sub-elements branched from the target sub-node.

[0050] S302: According to the conversion type identifier of the conversion image, obtain the proportional coefficient corresponding to the conversion type identifier, convert the position data corresponding to the target pixels associated with the target detection object in the mark data according to the proportional coefficient corresponding to the conversion type identifier, and obtain the mark data corresponding to the conversion image of each different conversion type.

[0051] The proportion coefficient is a parameter in the image conversion processing step. Since the conversion types of the converted images are different, the proportion coefficients of the corresponding image conversion processing steps are also different. In order to accurately obtain the label data corresponding to the converted image and improve the efficiency of obtaining the label data, in this embodiment, the proportion coefficient corresponding to the conversion type identifier is obtained according to the conversion type identifier of the converted image. The position data corresponding to the target pixel associated with the target detection object in the label data is converted according to the proportion coefficient corresponding to the conversion type identifier, and the label data corresponding to the converted image of each different conversion type is obtained. The workload of constructing the label data corresponding to the converted image of each different conversion type is effectively reduced, the labeling error caused by manual labeling is avoided, and the efficiency and accuracy of data augmentation are improved.

[0052] S4: inputting a plurality of aerial images, the label data corresponding to each aerial image, a plurality of converted images obtained by performing image conversion processing on each aerial image according to a plurality of preset conversion type methods, and the label data corresponding to each converted image into a neural network to be trained to obtain a target detection model.

[0053] The neural network model to be trained adopts a YOLOv5 (You Only Look Once) model. The YOLOv5 model is based on an open source framework Pytorch. The Pytorch framework integrates many neural network modules and calling functions to define and compose a detection algorithm.

[0054] By redefining the target detection as a classification and regression problem, the entire image is input into a neural network module, the image is divided into a grid, and the class probability of each grid and the generated detection rectangular frame are predicted.

[0055] In this embodiment, the detection device inputs a plurality of aerial images, the label data corresponding to each aerial image, a plurality of converted images obtained by performing image conversion processing on each aerial image according to a plurality of preset conversion type methods, and the label data corresponding to each converted image into a neural network to be trained to obtain a target detection model, thereby improving the robustness and generalization ability of the target detection model and effectively reducing the overfitting of the target detection model.

[0056] In an optional embodiment, the neural network to be trained includes an image cropping module and a detection and recognition module connected in sequence. Please refer to Figure 4 , Figure 4 The flowchart of S4 in the target detection method based on aerial images of a UAV provided by an embodiment of the present application includes steps S401-S402, and specifically as follows:

[0057] S401: Obtain the feature cropped image corresponding to the aerial image and the feature cropped image corresponding to the converted image output by the image cropping module according to the aerial image, the label data corresponding to the aerial image, the converted image, the label data corresponding to the converted image, and the image cropping module.

[0058] In this embodiment, the aerial image, the label data corresponding to the aerial image, the converted image, and the label data corresponding to the converted image are input into the image cropping module to obtain the feature cropped image corresponding to the aerial image and the feature cropped image corresponding to the converted image output by the image cropping module. This effectively reduces the redundancy of data input and improves the efficiency of model training. Compared with the converted image, the feature information in the feature cropped image accounts for a higher proportion. Using the feature cropped image as training data enables the model to better learn image features and improve the accuracy of the target detection model.

[0059] S402: Input the feature cropped image corresponding to the aerial image and the feature cropped image corresponding to the converted image into the detection and recognition module for iterative training to obtain the target detection model.

[0060] In this embodiment, the detection device inputs the feature cropped image corresponding to the aerial image and the feature cropped image corresponding to the converted image into the detection and recognition module for iterative training to obtain a plurality of trained neural network models, and obtains a target neural network model from the plurality of trained neural network models as the target detection model. Specifically, the detection device extracts feature vectors of features in the feature cropped image using the detection and recognition module in the neural network model, captures multi-scale context information and object boundary information through edge, color, shape, and other related information of the target detection object, and performs iterative training to obtain the target detection model.

[0061] S4: In response to a detection instruction, obtain a to-be-detected aerial image of a UAV, input the to-be-detected aerial image into the target detection model, and obtain a target detection result corresponding to the to-be-detected aerial image.

[0062] The detection instruction is issued by a user and received by the detection device.

[0063] In this embodiment, the detection device obtains a detection instruction sent by a user, obtains a to-be-detected aerial image of a UAV, inputs the to-be-detected aerial image into the target detection model, and obtains a target detection result corresponding to the to-be-detected aerial image.

[0064] In an optional embodiment, the detection device acquires a target detection identifier corresponding to the to-be-detected aerial image according to the target detection result, and returns to the display interface of the detection device to display and label the recognition identifier on the to-be-detected aerial image.

[0065] Please refer to Figure 5 , Figure 5 The structure diagram of the target detection device based on aerial images of a UAV is provided for an embodiment of the present application. The device can realize all or part of the target detection device based on aerial images of a UAV through software, hardware, or a combination of both. The device 5 includes:

[0066] The acquisition module 51 is configured to acquire a plurality of aerial images of a UAV and corresponding label data of each aerial image. The aerial images are images including a preset target detection object, and the label data includes position data corresponding to a plurality of target pixels associated with the target detection object.

[0067] The image conversion module 52 is configured to perform image conversion processing on each aerial image according to a plurality of preset conversion type methods, to obtain a plurality of converted images corresponding to each aerial image and a conversion type identifier corresponding to each converted image.

[0068] The label data conversion module 53 is configured to convert the position data corresponding to the target pixels associated with the target detection object according to the conversion type identifier of the converted image, to obtain label data of the converted image converted according to the corresponding conversion type.

[0069] The training module 54 is configured to input a plurality of aerial images, corresponding label data of each aerial image, a plurality of converted images obtained by performing image conversion processing on each aerial image according to a plurality of preset conversion type methods, and corresponding label data of each converted image to a neural network to be trained, to obtain a target detection model.

[0070] The detection module 55 is configured to acquire a to-be-detected aerial image of a UAV in response to a detection instruction, input the to-be-detected aerial image into the target detection model, and acquire a target detection result corresponding to the to-be-detected aerial image.

[0071] In the embodiment, the acquisition module acquires a plurality of aerial images of the unmanned aerial vehicle and corresponding label data of each aerial image, wherein the aerial image is an image including a preset target detection object, and the label data includes position data corresponding to a plurality of target pixels associated with the target detection object; the image conversion module converts each aerial image according to a plurality of preset conversion type methods to obtain a plurality of converted images corresponding to each aerial image and a conversion type identifier corresponding to each converted image; the label data conversion module converts the position data corresponding to the target pixels associated with the target detection object according to the conversion type identifier of the converted image to obtain label data of the converted image converted according to the corresponding conversion type; the training module inputs the plurality of aerial images, the corresponding label data of each aerial image, the plurality of converted images obtained by converting each aerial image according to the plurality of preset conversion type methods, and the corresponding label data of each converted image into a neural network to be trained to obtain a target detection model; and the detection module acquires a to-be-detected aerial image of the unmanned aerial vehicle in response to a detection instruction, inputs the to-be-detected aerial image into the target detection model, and acquires a target detection result corresponding to the to-be-detected aerial image.

[0072] The plurality of converted images obtained by converting the aerial images of the plurality of unmanned aerial vehicles according to a plurality of conversion type methods and the conversion type identifier corresponding to each converted image are used to convert the position data corresponding to the target pixels associated with the target detection object, to obtain label data of the converted image converted according to the corresponding conversion type, thereby improving the accuracy and efficiency of label data acquisition. The label data corresponding to the aerial image, the plurality of converted images obtained by converting each aerial image according to the plurality of preset conversion type methods, and the corresponding label data of each converted image are used as training data to train the neural network model, thereby improving the accuracy, robustness, and generalization ability of the trained neural network model, and improving the accuracy and efficiency of target detection of the aerial image.

[0073] Please refer to Figure 6 , Figure 6 The computer device provided in an embodiment of the present application includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. The computer device can store a plurality of instructions, which are suitable for being loaded and executed by the processor 61 to perform the method steps of the embodiments shown in Figures 1 to 4 . The specific execution process can refer to the specific description of the embodiments shown in Figures 1 to 4 . Details are not described herein.

[0074] The processor 61 can include one or more processing cores. The processor 61 connects various parts within the server by various interfaces and lines, executes various functions and processes data of the target detection apparatus 5 based on aerial images of a UAV by running or executing instructions, programs, code sets or instruction sets stored in the memory 62, and calling data in the memory 62. Optionally, the processor 61 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programable logic array (PLA). The processor 61 can be integrated with one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes operating systems, user interfaces, and application programs; the GPU is responsible for rendering and drawing the content displayed on the touch display; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 61, but can be realized by a separate chip.

[0075] The memory 62 can include a random access memory (RAM) and a read-only memory (ROM). Optionally, the memory 62 includes a non-transitory computer-readable storage medium. The memory 62 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 62 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function (such as touch instructions, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 62 can also be at least one storage device located away from the aforementioned processor 61.

[0076] The embodiments of the present application also provide a storage medium, which can store a plurality of instructions. The instructions are suitable for being loaded and executed by a processor to implement the method steps of the embodiments of the present application. Figures 1 to 4 The specific implementation process of the method steps of the embodiments shown in the above-mentioned method embodiments can be referred to the specific description of the embodiments shown in the above-mentioned method embodiments, and will not be repeated here. Figures 1 to 4 The specific implementation process of the method steps of the embodiments shown in the above-mentioned method embodiments can be referred to the specific description of the embodiments shown in the above-mentioned method embodiments, and will not be repeated here.

[0077] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or software. In addition, the specific name of each functional unit and module is only for the convenience of mutual distinction, and does not limit the protection scope of the present application. The specific working process of the unit and module in the above system can be referred to the corresponding process in the foregoing method embodiment, which will not be described here.

[0078] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0079] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the algorithm. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0080] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other ways. For example, the above-described apparatus / terminal device embodiments are only schematic. The division of the modules or units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.

[0081] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0082] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0083] The integrated module / unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, an executable file, or some intermediate form.

[0084] The present application is not limited to the above-described embodiments, and various modifications or changes can be made to the present application without departing from the spirit and scope of the present application. Therefore, it is intended that the present application encompass all such modifications and changes and fall within the scope of the appended claims and their equivalents.

Claims

1. A target detection method based on aerial images captured by unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Acquire several aerial images taken by a drone and the corresponding labeling data for each aerial image, wherein the aerial images are images including preset target detection objects, and the labeling data includes position data corresponding to several target pixels associated with the target detection objects; Each aerial image is processed by several preset conversion methods to obtain several converted images corresponding to each aerial image and a conversion type identifier corresponding to each converted image; The system obtains the identity identifier of the target detection object input by the user. Based on the identity identifier of the target detection object, it retrieves the target child node that matches the identity identifier of the target detection object from the root node of the aerial image's labeled data, and obtains the position data corresponding to the target pixel associated with the target detection object from the child elements of the target child node's branch. Based on the conversion type identifier of the converted image, the scaling factor corresponding to the conversion type identifier is obtained, wherein the scaling factor is a parameter in the image conversion processing step. Based on the scaling factor corresponding to the conversion type identifier, the position data corresponding to the target pixels associated with the target detection object in the marker data is converted to obtain the marker data corresponding to the converted images of each different conversion type. The marker data is a tree structure, including a root node, the root node is provided with connected child nodes, each child node is provided with branch child elements, the child nodes store the identity identifiers of several detected objects in the aerial image, and the child elements store the position data corresponding to the target pixels associated with the detected objects. Several aerial images, the label data corresponding to each aerial image, several converted images after each aerial image has been processed by several preset conversion methods, and the label data corresponding to each converted image are input into the neural network to be trained for training to obtain the target detection model. In response to a detection command, an aerial image of the UAV to be detected is acquired, the aerial image to be detected is input into the target detection model, and the target detection result corresponding to the aerial image to be detected is obtained.

2. The target detection method based on UAV aerial images according to claim 1, characterized in that: The neural network model to be trained includes an image cropping module and a detection and recognition module connected in sequence.

3. The target detection method based on UAV aerial images according to claim 2, characterized in that, The step of inputting a plurality of aerial images, the label data corresponding to each aerial image, a plurality of transformed images obtained by processing each aerial image according to a plurality of preset transformation methods, and the label data corresponding to each transformed image into a neural network to be trained for training to obtain a target detection model includes the following steps: Based on the aerial image, the corresponding marker data of the aerial image, the converted image, the corresponding marker data of the converted image, and the image cropping module, the feature cropping image corresponding to the aerial image and the feature cropping image corresponding to the converted image output by the image cropping module are obtained. The feature-cropped image corresponding to the aerial image and the feature-cropped image corresponding to the transformed image are input into the detection and recognition module for iterative training to obtain the target detection model.

4. The target detection method based on UAV aerial images according to claim 1, characterized in that, The acquisition of several aerial images from the drone also includes the following steps: Interference detection processing is performed on the plurality of aerial images, and the target aerial image is extracted from the plurality of aerial images.

5. The target detection method based on UAV aerial images according to claim 1, characterized in that: The conversion methods include scaling, rotation, tilting, flipping, grayscale adjustment, contrast adjustment, noise filtering, Gaussian blur, and median blur processing; the converted images include scaled images, rotated images, tilted images, flipped images, grayscale adjusted images, contrast adjusted images, smoothed images, Gaussian blurred images, and median blurred images.

6. A target detection device based on aerial images captured by unmanned aerial vehicles, characterized in that, include: The acquisition module is used to acquire several aerial images of the UAV and the corresponding labeling data of each aerial image. The aerial images are images that include a preset target detection object, and the labeling data includes position data of several target pixels associated with the target detection object. The image conversion module is used to perform image conversion processing on each of the aerial images according to several preset conversion type methods, so as to obtain several converted images corresponding to each aerial image and a conversion type identifier corresponding to each converted image; The tag data conversion module is used to obtain the identity identifier of the target detection object input by the user, and according to the identity identifier of the target detection object, to retrieve the target child node that matches the identity identifier of the target detection object from the root node of the tag data of the aerial image, and to obtain the position data corresponding to the target pixel associated with the target detection object from the sub-elements of the branch of the target child node. Based on the conversion type identifier of the converted image, the scaling factor corresponding to the conversion type identifier is obtained, wherein the scaling factor is a parameter in the image conversion processing step. Based on the scaling factor corresponding to the conversion type identifier, the position data corresponding to the target pixels associated with the target detection object in the marker data is converted to obtain the marker data corresponding to the converted images of each different conversion type. The marker data is a tree structure, including a root node, the root node is provided with connected child nodes, each child node is provided with branch child elements, the child nodes store the identity identifiers of several detected objects in the aerial image, and the child elements store the position data corresponding to the target pixels associated with the detected objects. The training module is used to input several aerial images, the label data corresponding to each aerial image, several transformed images after each aerial image has been processed by several preset transformation methods, and the label data corresponding to each transformed image into the neural network to be trained for training, so as to obtain the target detection model. The detection module is used to respond to a detection command, acquire an aerial image of the UAV to be detected, input the aerial image to be detected into the target detection model, and obtain the target detection result corresponding to the aerial image to be detected.

7. A computer device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor; the computer program, when executed by the processor, implements the steps of the target detection method based on aerial images from unmanned aerial vehicles as described in any one of claims 1 to 5.

8. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the steps of the target detection method based on aerial images from unmanned aerial vehicles as described in any one of claims 1 to 5.

Images