Novel wheel holding mechanism based on YOLOv8

Abstract

The invention relates to the technical field of intelligence, in particular to a novel wheel holding mechanism based on YOLOv8. The method includes: acquiring an aircraft nose landing gear image; performing preprocessing according to the aircraft nose landing gear image to generate an aircraft nose landing gear data set; training data and key training pictures are input into a YOLO network for training, so that an aircraft nose landing gear detection model is obtained; inputting the test data into the airplane nose landing gear detection model for detection, and outputting an airplane nose landing gear detection result; and identifying the nose landing gear of the aircraft by using the detection result of the nose landing gear of the aircraft so as to determine the accurate butt joint of the tractor wheel holding mechanism and the nose landing gear of the aircraft. An aircraft nose landing gear image is obtained, a YOLO network is used for training image data so as to obtain a detection model, and an aircraft nose landing gear detection result is output after detection of the detection model. According to the method, the nose landing gear of the airplane is identified according to the detection result of the nose landing gear of the airplane, the YOLO network is used for rapidly detecting the nose landing gear of the airplane, and then the output detection result is identified, so that the nose landing gear of the airplane is rapidly detected and identified, and the docking execution efficiency of a tractor and the airplane in an airplane traction taxiing system is ensured.

Description

Technical Field

[0001] The present invention relates to the field of intelligent technology, and in particular to a novel wheel-holding mechanism based on YOLOv8. Background Art

[0002] Aircraft tractors are an important part of the operation of modern smart airports. How the aircraft tractor's wheel-holding mechanism can quickly and accurately identify and lift the aircraft's front landing gear in the complex environment of the airport is of great significance to airport traffic and airport scheduling, improving airport operation efficiency and shortening aircraft departure time. However, the existing aircraft tractor's wheel-holding mechanism cannot meet the requirements of rapid detection and identification of the aircraft's front landing gear, which makes the current aircraft tractors unable to meet actual needs and therefore has a certain impact on the airport's operating efficiency.

[0003] In the prior art, the wheel-holding mechanism of an aircraft tractor is mainly used to identify and lift the front landing gear of an aircraft in an airport environment. This process is of great significance for airport traffic and airport scheduling, improving airport operating efficiency, and shortening aircraft departure time. However, when facing a complex airport environment, the speed and accuracy of the existing wheel-holding mechanism in identifying and lifting the front landing gear of an aircraft often cannot meet actual needs. This is mainly because the existing wheel-holding mechanism does not effectively utilize modern intelligent technologies, such as deep learning, computer vision, etc., in its design and implementation to improve the speed and accuracy of its identification and lifting of the front landing gear of the aircraft. This not only limits the working efficiency of the aircraft tractor, but also has a negative impact on the overall operating efficiency of the airport. Therefore, how to improve the recognition and lifting speed and accuracy of the wheel-holding mechanism of the aircraft tractor has become an important technical issue at present. Summary of the invention

[0004] The present invention provides a novel wheel-holding mechanism based on YOLOv8 to solve the problem that the existing wheel-holding mechanism of a tractor cannot quickly detect and identify the front landing gear of an aircraft.

[0005] According to a first aspect, the present invention provides a novel wheel holding mechanism based on YOLOv8, comprising:

[0006] Acquire the image of the aircraft's front landing gear;

[0007] Preprocessing the aircraft nose landing gear image to generate an aircraft nose landing gear dataset; wherein the aircraft nose landing gear dataset includes: test data and training data;

[0008] The training images in the training data are input into the YOLO network for training to obtain a front landing gear detection model of the aircraft;

[0009] Inputting the test data into the aircraft nose landing gear detection model for detection, and outputting the aircraft nose landing gear detection result;

[0010] The front landing gear of the aircraft is identified by utilizing the front landing gear detection result of the aircraft to determine the front landing gear of the aircraft.

[0011] By acquiring the image of the aircraft's front landing gear, the YOLO network is used to train the image data to obtain a detection model. After detection by the detection model, the detection result of the aircraft's front landing gear is output. The aircraft's front landing gear is identified by the detection result of the aircraft's front landing gear. The YOLO network is used to perform rapid aircraft front landing gear detection, and then the output detection result is identified, thereby achieving rapid detection and identification of the aircraft's front landing gear, ensuring the execution efficiency of the wheel holding mechanism in lifting the aircraft's front landing gear.

[0012] In combination with the first aspect, in a first implementation of the first aspect, the aircraft front landing gear image is preprocessed, including:

[0013] An initialization function is defined, data of the acquired aircraft nose landing gear image is enhanced, and the acquired aircraft nose landing gear image is labeled.

[0014] In combination with the first aspect, the aircraft nose landing gear image obtained in the second implementation of the first aspect is labeled, including: identifying the aircraft nose landing gear in the aircraft nose landing gear image, and obtaining the border coordinate information in the aircraft nose landing gear image.

[0015] By preprocessing the image of the front landing gear of the aircraft, it is ensured that the values ​​obtained when the image data is input into the network are biased towards the actual data, and the image integrity can be guaranteed and the duplication of image data can be avoided.

[0016] In combination with the first aspect, in a third implementation of the first aspect, training is performed by inputting training images in the training data set into a YOLO network, including:

[0017] The preprocessed image is input into the YOLO network to obtain feature maps of three different scales: the smallest receptive field and the largest scale, the medium receptive field and the medium scale, and the largest receptive field and the smallest scale;

[0018] After upsampling and feature fusion based on the feature maps of the three different scales, tensor data at the maximum scale, tensor data at the medium scale, and tensor data at the minimum scale are obtained respectively;

[0019] The loss function is used to calculate the tensor data relative to the true value, and the gradient is obtained through back propagation to update the weights. At the same time, the validation set is used for verification, and finally the detection model and detection results based on the YOLO network are obtained.

[0020] By taking advantage of the fast detection of the YOLO network, the front landing gear of the aircraft is detected, and the false detection background probability of the YOLO network is used to ensure that the front landing gear of the aircraft can be detected quickly and the detection results can be output.

[0021] In combination with the first aspect, in a fourth implementation of the first aspect, the test data set is input into an aircraft front landing gear detection model for detection, including: performing iterative detection through training data and verification data.

[0022] Iterative testing is performed through training data and validation data to improve the accuracy of aircraft nose landing gear detection in the aircraft nose landing gear detection model.

[0023] Obtaining the aircraft nose landing gear detection result, and dividing the aircraft nose landing gear detection result into an aircraft nose landing gear detection result training sample and an aircraft nose landing gear detection result test sample;

[0024] In combination with the first aspect, in a fifth implementation of the first aspect, the loss function is: binary cross entropy.

[0025] According to a second aspect, the present invention provides a novel wheel-holding mechanism, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the YOLO-based aircraft front landing gear identification method described in the first aspect or any one embodiment of the first aspect by executing the computer instructions.

[0026] According to a third aspect, an embodiment of the present invention provides a computer-readable storage medium. The storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the aircraft front landing gear recognition method based on YOLO described in the first aspect or any one embodiment of the first aspect.

[0027] Compared with the prior art, the beneficial effects of the present invention are: YOLO is an object detection algorithm based on a deep learning neural network architecture, which can ensure the accuracy of the output frame and the convergence of the algorithm regression. It can also eliminate the cumulative inspection error caused by changes in the field of view, and the target object can be identified even if the background is very complex. Applying the algorithm to the new wheel-holding mechanism for aircraft towing and taxiing can achieve rapid identification of the front landing gear of the aircraft, with fast speed and good real-time performance. It can be well adapted to changes in the airport environment and improve the efficiency of aircraft towing and departure. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings; the accompanying drawings are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0029] Figure 1 is a flowchart of a method for identifying an aircraft front landing gear based on YOLO according to an embodiment of the present invention;

[0030] Figure 2 is a flowchart of a method for identifying an aircraft front landing gear based on YOLO according to an embodiment of the present invention;

[0031] Figure 3 The present invention is based on a schematic diagram of the hardware structure of a novel wheel-holding mechanism.

[0032] Reference numerals:

[0033] 3-1 processor; 3-2 memory; 3-3 bus. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0035] The embodiment of the present invention provides a method for recognizing a front landing gear of an aircraft based on YOLO, comprising:

[0036] Figure 1 In S1-1, the image of the front landing gear of the aircraft is obtained; and the data set is packaged, which can be the surveillance video data within the airport.

[0037] Figure 1 In S1-2, preprocessing is performed on the aircraft nose landing gear image to generate an aircraft nose landing gear dataset; wherein the aircraft nose landing gear dataset includes: test data and training data; wherein the preprocessing includes: defining an initialization function, performing image data enhancement on the acquired aircraft nose landing gear image, and labeling the acquired aircraft nose landing gear image.

[0038] Defining the initialization function includes: selecting test parameter data, setting the image size of the parameter data, preferably, the image size is a multiple of 32. Setting the data enhancement method, setting the training data input size, selecting the sampling method, and the number of training data.

[0039] Data image enhancement includes: randomly flipping the data image horizontally, randomly cropping the data image, and randomly rotating the data image horizontally to ensure the consistency of the data image.

[0040] Labeling the front landing gear image of the aircraft may be performed by obtaining the data target image of the front landing gear of the aircraft, obtaining the coordinate information of the target image and the category information of the target image (such as the front landing gear of the aircraft and the main landing gear of the aircraft);

[0041] Figure 1 In S1-3, the training images in the training data are input into the YOLO network for training to obtain the aircraft front landing gear detection model; the preprocessed image data is input into the YOLO network, the YOLO network sets the border for the training images in the training data, selects the detection target through the border, marks it, and finally outputs the detection result.

[0042] Figure 1 In S1-4, the test data is input into the aircraft nose landing gear detection model for detection, and the aircraft nose landing gear detection result is output;

[0043] The test data is tested through the generated detection model, and the test results are output. The test results include a selection box image with a mark or recognition result. The front landing gear of the aircraft is determined by identifying the data in the selection box image, so that rapid identification and detection can be achieved, ensuring the efficiency of the tractor's wheel-holding mechanism in lifting the front landing gear of the aircraft.

[0044] By acquiring the image of the aircraft's front landing gear, the image data is trained using the YOLO network to obtain a detection model. After detection by the detection model, the detection result of the aircraft's front landing gear is output. By identifying the detection result of the aircraft's front landing gear, the YOLO network is used to perform rapid aircraft front landing gear detection. After that, the output detection result is identified, thereby achieving rapid detection and identification of the aircraft's front landing gear, ensuring the efficiency of the tractor's wheel-holding mechanism in lifting the aircraft's front landing gear.

[0045] The embodiment of the present invention provides a method for recognizing the front landing gear of an aircraft based on YOLO, such as Figure 2 As shown, the training images in the training data set are input into the YOLO network for training, including:

[0046] Figure 2In S2-1, the preprocessed image is input into the YOLO network to obtain three different scale feature maps with the smallest receptive field and the largest scale, the medium receptive field and the medium scale, and the largest receptive field and the smallest scale; for example: on the smallest 13x13 feature map (largest receptive field), its larger feature selection box can be (116x90), (156x198), (373x326), which is suitable for detecting larger objects. On the medium 26*26 feature map (medium receptive field), its medium feature selection box can be (30x61), (62x45), (59x119), which is suitable for detecting medium-sized objects. On the larger 52x52 feature map (smaller receptive field), its smaller feature selection box can be (10x13), (16x30), (33x23), which is suitable for detecting smaller objects.

[0047] Figure 2 In S2-2, after upsampling and feature fusion based on the feature maps of the three different scales, tensor data at the maximum scale, tensor data at the medium scale, and tensor data at the minimum scale are obtained respectively; wherein, upsampling can be obtained by bilinear interpolation and transposed convolution, and feature fusion requires the use of a feature pyramid (FNC) to fuse features of different scales, wherein separate multi-scale detection is also required on the fused feature maps of different scales to ensure the accuracy of the final detection result.

[0048] Figure 2 In S2-3, the tensor data is calculated relative to the true value through the loss function, and the gradient is obtained through back propagation, the weight is updated, and the verification set is used for verification, and finally the detection model and detection results based on the YOLO network are obtained. By using the verification set for verification, the accuracy of the training data can be guaranteed, so as to ensure the accuracy of the recognition results, which is beneficial to improve the execution efficiency of detection and recognition. Among them, the YOLO network used is YOLOv8.

[0049] The LOYO algorithm is used to quickly detect and identify the front landing gear of an aircraft. The LOYO algorithm can directly and quickly output the detection results to ensure the efficiency of the detection execution, and can also meet the requirements of real-time aircraft front landing gear identification, which is beneficial to the rapid identification and detection of the front landing gear of the aircraft towed by the tractor, and improves the ability of the tractor's wheel holding mechanism to grasp the aircraft front landing gear.

[0050] The embodiment of the present invention also provides a new type of wheel holding mechanism, such as Figure 3 As shown, the wheel holding mechanism terminal may include a processor 3-1 and a memory 3-2, wherein the processor 3-1 and the memory 3-2 may be connected via a bus 3-3 or other means. Figure 3 The connection via bus 3-3 is taken as an example.

[0051] The processor 3-1 may be a central processing unit (CPU). The processor 3-1 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0052] The memory 3-2 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the vehicle display device key shielding method in the embodiment of the present invention. The processor 3-1 executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory 3-2, that is, the YOLO-based aircraft front landing gear recognition method in the above method embodiment is implemented.

[0053] The memory 3-2 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor 3-1, etc. In addition, the memory 3-2 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 3-2 may optionally include a memory remotely arranged relative to the processor 3-1, and these remote memories may be connected to the processor 3-1 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0054] The one or more modules are stored in the memory 3-2 and when executed by the processor 3-1, perform the following steps: Figure 1 , Figure 2 The aircraft front landing gear recognition method based on YOLO in the illustrated embodiment.

[0055] For details of the above-mentioned wheel holding mechanism terminal, please refer to Figure 1 , Figure 2 The corresponding related descriptions and effects in the illustrated embodiments can be understood and will not be repeated here.

[0056] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk (HDD) or a solid state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0057] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A new wheel-holding mechanism based on YOLOv8, characterized in that: Acquire an image of the front landing gear of an aircraft. Perform preprocessing on the image of the front landing gear of an aircraft to generate a data set of the front landing gear of an aircraft: wherein the data set of the front landing gear of an aircraft includes test data and training data; the training images in the training data are input into the YOLO network for training to obtain a detection model of the front landing gear of an aircraft. The detection model of the front landing gear of an aircraft includes an EfficientRep network structure and a Rep-PAN network structure: the EfficientRep network structure is used to extract multi-scale features of the image; the Rep-PAN network structure is used to perform information fusion based on the multi-scale features using top-down and bottom-up paths to obtain fused features. The test data is input into the detection model of the front landing gear of an aircraft for detection, and the detection result of the front landing gear of an aircraft is output; the detection result of the front landing gear of an aircraft is used to identify the front landing gear of an aircraft to determine the front landing gear of the aircraft.

2. The method according to claim 1, characterized in that Preprocessing is performed on the image of the front landing gear of the aircraft, an initialization function is defined, image data enhancement is performed on the acquired image of the front landing gear of the aircraft, and the acquired image of the front landing gear of the aircraft is labeled.

3. The method according to claim 2, characterized in that The acquired aircraft nose landing gear image is labeled, the aircraft nose landing gear in the aircraft nose landing gear is identified, and the frame coordinates and category information in the aircraft nose landing gear image are acquired.

4. The method according to claim 1, characterized in that The training images in the training data set are input into the YOLO network for training, and the preprocessed images are input into the YOLO grid to obtain feature maps of three different scales, namely, the smallest receptive field and the largest scale, the medium receptive field and the medium scale, and the largest receptive field and the smallest scale; upsampling and feature fusion are performed based on the feature maps of the three different scales to obtain tensor data at the largest scale, tensor data at the medium scale, and tensor data at the smallest scale respectively; the tensor data is calculated relative to the true value through the loss function, and the gradient is obtained through back propagation, the weight is updated, and verification is performed using the verification set, and finally a detection model and detection results based on the YOLO network are obtained.

5. The method according to claim 4, characterized in that The test data set is input into the aircraft nose landing gear detection model for detection, including: performing iterative detection through training data and verification data.

6. The method according to claim 4, characterized in that The front landing gear of the aircraft is identified by using the front landing gear detection result of the aircraft to determine the front landing gear of the aircraft, the front landing gear detection result of the aircraft is obtained, and the front landing gear result of the aircraft is divided into a training sample of the front landing gear result of the aircraft and a test sample of the front landing gear detection result of the aircraft.

7. The method according to claim 4, characterized in that The loss function is: binary cross entropy.

8. According to claim 1, it is characterized in that: The memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the YOLO-based aircraft front landing gear recognition method according to any one of claims 1 to 7 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the aircraft front landing gear recognition method based on YOLO according to any one of claims 1-7.

Images