Lunar Rock Rotation Target Detection Method and Device Based on YOLO-R Model

Abstract

The present application discloses a method, device, storage medium and electronic device for detecting rotating targets of lunar falling rocks based on the YOLO-R model. The method includes: obtaining a training set of lunar remote sensing image data; wherein, the lunar remote sensing image training data set includes sliced images of lunar remote sensing images and four-point labels of lunar falling rocks; inputting the sliced images of lunar remote sensing images into an improved YOLO-R model to obtain a rotating target prediction result; constructing a loss function based on the rotating target prediction result and the four-point labels of lunar falling rocks, and training the improved YOLO-R model based on the loss function; obtaining sliced images of the lunar remote sensing image to be detected, and inputting the sliced images of the lunar remote sensing image to be detected into the trained improved YOLO-R model to obtain a rotating target prediction result of the sliced images of the lunar remote sensing image to be detected. The present application can reduce the misdetection of ordinary lunar surface rocks as lunar falling rocks and improve the accuracy of target detection.

Description

Technical Field

[0001] The present application relates to the technical field of deep space exploration, and in particular, to a method, device, storage medium, and electronic device for detecting rotating lunar rock targets based on the yolo-R model. Background Art

[0002] Lunar rocks are important surface features on the lunar surface. Collecting and detecting lunar rocks is of great significance for studying deep space topics such as the composition of extraterrestrial matter and the lunar geological environment. In recent years, with the development of artificial intelligence, the advantages of deep learning have gradually emerged. Neural networks have broad application prospects in the fields of object detection and image recognition.

[0003] The YOLO (You Only Look Once) model is a representative algorithm in the field of object detection in recent years. Its unique architecture enables it to have the ability of real-time detection while maintaining relatively high detection accuracy. It performs excellently in real-time object recognition and positioning tasks. This model realizes efficient feature extraction and fast inference through a fully convolutional neural network (CNN) architecture. The YOLO model has been trained on multiple datasets and can accurately detect and classify various targets. However, applying this model to the recognition of lunar rocks in lunar remote sensing images still faces many challenges.

[0004] However, previous object detection algorithms perform well in dealing with targets with regular shapes and fixed angles, but it is difficult for them to accurately detect the features of lunar rocks with complex and variable rotation angles on the lunar surface. Traditional lunar rock object detection algorithms have certain misdetections, which affect the performance of the model and reduce the accuracy. Summary of the Invention

[0005] Embodiments of the present application provide a method, device, storage medium, and electronic device for detecting rotating lunar rock targets based on the yolo-R model, which can improve the accuracy of object detection.

[0006] Embodiments of the present application provide a method for detecting rotating lunar rock targets based on the yolo-R model, including:

[0007] Obtain a training set of lunar remote sensing image data; wherein, the lunar remote sensing image training dataset includes sliced images of lunar remote sensing images and four-point labels of lunar rocks;

[0008] Input the sliced images of the lunar remote sensing images into an improved yolo-R model to obtain a prediction result of the rotating target;

[0009] Construct a loss function based on the prediction result of the rotating target and the four-point labels of the lunar rocks, and train the improved yolo-R model based on the loss function;

[0010] Obtain sliced images of the lunar remote sensing image to be detected, and input the sliced images of the lunar remote sensing image to be detected into the trained improved yolo-R model to obtain the rotation target prediction results of the sliced images of the lunar remote sensing image to be detected.

[0011] As a further improvement of the present invention, in the above lunar rockfall rotation target detection method based on the yolo-R model, wherein, the obtaining of the lunar remote sensing image data training set includes:

[0012] Obtain lunar remote sensing image data;

[0013] Perform orthogonal projection on the lunar remote sensing image data, and slice the orthogonally projected lunar remote sensing image data to obtain sliced images of the lunar remote sensing image;

[0014] Based on the longitude, latitude and diameter information of the lunar rockfall, obtain the four-point label of the lunar rockfall;

[0015] Construct a lunar remote sensing image data training set based on the sliced images of the lunar remote sensing image and the four-point labels of the lunar rockfall.

[0016] As a further improvement of the present invention, in the above lunar rockfall rotation target detection method based on the yolo-R model, wherein, the improved yolo-R model includes a backbone network module, a feature extraction module and an improved head output module;

[0017] The step of inputting the sliced images of the lunar remote sensing image into the improved yolo-R model to obtain the rotation target prediction results includes:

[0018] Input the sliced images of the lunar remote sensing image into the backbone network module for feature extraction to obtain a first feature map;

[0019] Input the first feature map into the feature extraction module to perform multi-scale feature extraction and feature fusion to obtain a second feature map;

[0020] Input the second feature map into the improved head output module to obtain multiple rotation target candidate boxes;

[0021] Calculate the confidence of the multiple rotation target candidate boxes, and use the rotation target candidate box with the highest confidence as the rotation target prediction result.

[0022] As a further improvement of the present invention, in the above lunar rockfall rotation target detection method based on the yolo-R model, wherein, the improved head output module includes a raw convolutional layer, an improved convolutional layer, a pooling layer and a connection layer connected in sequence;

[0023] Inputting the second feature map into an improved head output module to obtain multiple rotation target candidate boxes, including:

[0024] Inputting the second feature map into the original convolutional layer to obtain prediction candidate boxes;

[0025] Inputting the prediction candidate boxes into the improved convolutional layer to extract angle information, and obtaining rotation target candidate boxes based on the angle information and the prediction candidate boxes;

[0026] Inputting the rotation target candidate boxes into the pooling layer for classification;

[0027] Inputting the classified rotation target candidate boxes into the connection layer for regression to obtain the final rotation target candidate boxes.

[0028] As a further improvement of the present invention, in the above-mentioned lunar meteorite rotation target detection method based on the yolo-R model, wherein, constructing a loss function based on the rotation target prediction result and the four-point label of the lunar meteorite, including:

[0029] Calculating the label area based on the four-point label of the lunar meteorite;

[0030] Calculating the predicted area based on the rotation target prediction result;

[0031] Calculating the overlapping area between the label area and the predicted area;

[0032] Calculating the loss function based on the label area, the predicted area, and the overlapping area.

[0033] As a further improvement of the present invention, in the above-mentioned lunar meteorite rotation target detection method based on the yolo-R model, wherein, the loss function is:

[0034]

[0035] Wherein, is the loss function, is the label area, is the predicted area, is the overlapping area.

[0036] As a further improvement of the present invention, in the above-mentioned lunar meteorite rotation target detection method based on the yolo-R model, wherein, as a further improvement of the present invention, in the above-mentioned lunar meteorite rotation target detection method based on the yolo-R model, wherein, calculating the overlapping area between the label area and the predicted area, including:

[0037] Calculate the left boundary, right boundary, upper boundary, and lower boundary of the four-point label and the predicted result of the rotating target respectively;

[0038] Calculate the overlapping area through the left boundary, right boundary, upper boundary, and lower boundary, which can be expressed as:

[0039]

[0040] Wherein, is the overlapping area, L is the left boundary, R is the right boundary, U is the upper boundary, D is the lower boundary.

[0041] The embodiment of the present application also provides a lunar rockfall rotating target detection device based on the yolo-R model, including:

[0042] An acquisition module for acquiring a training set of lunar remote sensing image data; wherein, the lunar remote sensing image training data set includes sliced images of lunar remote sensing images and four-point labels of lunar rockfalls;

[0043] A processing module for inputting the sliced images of the lunar remote sensing images into the improved yolo-R model to obtain a predicted result of the rotating target;

[0044] A training module for constructing a loss function based on the predicted result of the rotating target and the four-point label of the lunar rockfall, and training the improved yolo-R model based on the loss function;

[0045] A rotating target detection module for acquiring sliced images of the lunar remote sensing image to be detected, and inputting the sliced images of the lunar remote sensing image to be detected into the trained improved yolo-R model to obtain a predicted result of the rotating target of the sliced images of the lunar remote sensing image to be detected.

[0046] The embodiment of the present application also provides a computer-readable storage medium, in which multiple instructions are stored, and the instructions are suitable for being loaded by a processor to execute any one of the above-mentioned lunar rockfall rotating target detection methods based on the yolo-R model.

[0047] The embodiment of the present application also provides an electronic device, including a processor and a memory, the processor is electrically connected to the memory, the memory is used for storing instructions and data, and the processor is used for the steps in any one of the above-mentioned lunar rockfall rotating target detection methods based on the yolo-R model.

[0048] The lunar rock rotation target detection method, device, storage medium and electronic device based on the yolo-R model provided by this application. By introducing a four-point label as the ground truth label, the yolo-R model is trained to obtain the rotation target prediction result. This application improves the yolo-R model, enabling it to capture the rotation features in uncertain directions around the target, reducing the misdetection of ordinary lunar surface rocks as lunar falling rocks, and improving the accuracy of target detection. At the same time, this application has a higher detection speed, can quickly respond to the requirements under different observation conditions, and provides more reliable data support for research in aspects such as the composition of extraterrestrial substances and the lunar geological environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The following will, with reference to the accompanying drawings and through a detailed description of the specific embodiments of this application, make the technical solutions and other beneficial effects of this application obvious.

[0050] Figure 1 It is a flowchart of the lunar rock rotation target detection method based on the yolo-R model provided by an embodiment of this application.

[0051] Figure 2 It is a schematic structural diagram of the improved yolo-R model provided by an embodiment of this application.

[0052] Figure 3 It is the marked lunar remote sensing image provided by an embodiment of this application.

[0053] Figure 4 It is a schematic structural diagram of the lunar rock rotation target detection device based on the yolo-R model provided by an embodiment of this application.

[0054] Figure 5 It is a schematic structural diagram of the electronic device provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The following will, with reference to the accompanying drawings in the embodiments of this application, clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of this application.

[0056] Previous object detection algorithms perform well when dealing with objects with regular shapes and fixed angles, but it is difficult for them to accurately detect the characteristics of fallen rocks on the lunar surface with complex and variable rotation angles. This is because all the prediction boxes in previous lunar fallen rock object detection algorithms are horizontal prediction boxes, without considering the rotation characteristics of the objects. Therefore, it is difficult for the neural network of traditional algorithms to learn the rotation characteristics of the objects during training, and at the same time, the rotation characteristics of the objects are not considered during output. However, the main difference between lunar rocks and lunar fallen rocks is that there are traces left by the collision with the lunar surface around the lunar fallen rocks, which is obviously a rotation characteristic with an indefinite direction.

[0057] To solve the above problems, the embodiments of the present application provide a method, device, storage medium, and electronic device for detecting rotating objects of lunar fallen rocks based on the yolo-R model. A device for detecting rotating objects of lunar fallen rocks based on the yolo-R model provided by the embodiments of the present application can be integrated in an electronic device, and the electronic device can be a device such as a terminal or a server. Among them, the terminal can include a tablet computer, a laptop computer, a personal computer (PC), a microprocessing box, or other devices, etc.

[0058] Please refer to Figure 1 , Figure 1 which is a flowchart of the method for detecting rotating objects of lunar fallen rocks based on the yolo-R model provided by the embodiments of the present application. It is applied to an electronic device, and the method for detecting rotating objects of lunar fallen rocks based on the yolo-R model includes the following steps:

[0059] S1, obtain a training set of lunar remote sensing image data; wherein, the training data set of lunar remote sensing images includes sliced images of lunar remote sensing images and four-point labels of lunar fallen rocks.

[0060] In one embodiment, step S1 includes the following steps:

[0061] S11, obtain lunar remote sensing image data.

[0062] S12, perform orthogonal projection on the lunar remote sensing image data, and slice the orthogonally projected lunar remote sensing image data to obtain sliced images of lunar remote sensing images.

[0063] S13, obtain the four-point labels of lunar fallen rocks based on the longitude, latitude, and diameter information of lunar fallen rocks.

[0064] S14, construct a training set of lunar remote sensing image data based on the sliced images of lunar remote sensing images and the four-point labels of lunar fallen rocks.

[0065] Specifically, the lunar remote sensing image is cropped into training samples of appropriate size, and then orthogonally projected to reduce image distortion. The image is sliced, and the image is segmented and complemented into a square lunar remote sensing image of 1024×1024 size. The published authoritative manually identified lunar meteorite catalog, which contains the longitude, latitude, and diameter information of lunar meteorites, is first converted into data in the pixel coordinate system and then marked in the lunar meteorite samples. The sliced lunar remote sensing images and corresponding data labels are divided into two sets: a training set for training the model and a test set for testing the model after training is completed.

[0066] The ordinary image dataset has two-point coordinates, and the label is a horizontal rectangle. The lunar meteorite label provided in the embodiment of the present application is a four-point label, marking four points, which is a rectangle that can be rotated at any angle.

[0067] S2. Input the sliced image of the lunar remote sensing image into the improved yolo-R model to obtain the rotation target prediction result.

[0068] Figure 2 It is a schematic structural diagram of the improved yolo-R model provided in the embodiment of the present application. As Figure 2 shown, the improved yolo-R model includes a backbone network module, a feature extraction module, and an improved head output module. Step S2 specifically includes the following steps:

[0069] S21. Input the sliced image of the lunar remote sensing image into the backbone network module for feature extraction to obtain the first feature map.

[0070] S22. Input the first feature map into the feature extraction module to perform multi-scale feature extraction and feature fusion to obtain the second feature map.

[0071] S23. Input the second feature map into the improved head output module to obtain multiple rotation target candidate boxes.

[0072] Specifically, the improved head output module includes a raw convolutional layer, an improved convolutional layer, a pooling layer, and a connection layer connected in sequence. Step S23 specifically includes the following steps:

[0073] S231. Input the second feature map into the raw convolutional layer to obtain the prediction candidate boxes;

[0074] S232. Input the prediction candidate boxes into the improved convolutional layer to extract the angle information, and based on the angle information and the prediction candidate boxes, obtain the rotation target candidate boxes;

[0075] S233. Input the rotation target candidate boxes into the pooling layer for classification;

[0076] S234. Input the classified rotation target candidate boxes into the connection layer for regression to obtain the final rotation target candidate boxes.

[0077] S24. Calculate the confidence levels for multiple rotation target candidate boxes, and use the rotation target candidate box with the highest confidence level as the rotation target prediction result.

[0078] S3. Construct a loss function based on the rotation target prediction result and the four-point label of the lunar rockfall, and train the improved yolo-R model based on the loss function.

[0079] Specifically, constructing the loss function based on the rotation target prediction result and the four-point label of the lunar rockfall in step S3 includes:

[0080] S31. Calculate the area of the label region based on the four-point label of the lunar rockfall.

[0081] S32. Calculate the area of the prediction region based on the rotation target prediction result.

[0082]

[0083]

[0084] Through the input four-point label , calculate the area of the label region . In the label respectively represent the coordinates of the four points of the label rectangle after normalization processing. Through the rotation target prediction result , calculate the area of the prediction region , where respectively represent the coordinates of the four points of the prediction rectangle after normalization processing.

[0085] S33. Calculate the overlapping area between the area of the label region and the area of the prediction region.

[0086] Specifically, step S33 includes the following steps:

[0087] S331. Calculate the left boundary, right boundary, upper boundary, and lower boundary of the four-point label and the rotation target prediction result respectively.

[0088] Specifically, to calculate the overlapping area according to the coordinate information, first calculate the boundary positions of the two rectangles. In the following formula represents taking the maximum value of the values in the parentheses, represents taking the minimum value of the values in the parentheses.

[0089] Left boundary of the label rectangle: .

[0090] Right boundary of the label rectangle: 。

[0091] Upper boundary of the label rectangle: 。

[0092] Lower boundary of the label rectangle: 。

[0093] Left boundary of the prediction rectangle: 。

[0094] Right boundary of the prediction rectangle: 。

[0095] Upper boundary of the prediction rectangle: 。

[0096] Lower boundary of the label rectangle: 。

[0097] Calculate the boundaries of the overlapping area.

[0098] Left boundary of the label rectangle: 。

[0099] Right boundary of the label rectangle: 。

[0100] Upper boundary of the label rectangle: 。

[0101] Lower boundary of the label rectangle: 。

[0102] S332. Calculate the area of the overlapping area through the left boundary, right boundary, upper boundary and lower boundary, which can be expressed as:

[0103]

[0104] Where, is the area of the overlapping area, L is the left boundary, R is the right boundary, U is the upper boundary, D is the lower boundary.

[0105] S34. Calculate the loss function based on the label area, prediction area and overlapping area.

[0106] The loss function is:

[0107]

[0108] Where, is the loss function, is the label area, is the prediction area, is the overlapping area.

[0109] Further, the potential lunar meteorite areas are verified using the authoritative manually identified lunar meteorite data, and the potential lunar meteorites meeting the preset requirements are identified and marked in the lunar remote sensing images.

[0110] Figure 3 The marked lunar remote sensing image provided by the embodiment of the present application is as Figure 3 shown. The authoritative manually identified lunar meteorite data is compared and analyzed with the predicted potential lunar meteorite areas, the number of the lunar meteorite areas matched by the two is counted, and the coincidence degree of the lunar meteorite areas of the two is calculated to evaluate the model.

[0111] S4. Obtain the sliced images of the lunar remote sensing image to be detected, input the sliced images of the lunar remote sensing image to be detected into the trained improved yolo-R model, and obtain the rotation target prediction results of the sliced images of the lunar remote sensing image to be detected.

[0112] In the present application, a four-point label is introduced as the true label to train the improved yolo-R model to obtain the rotation target prediction results. The yolo-R model is improved in the present application, so that the yolo-R model can capture the rotation features in uncertain directions around the target, improving the model accuracy. At the same time, the present application has a higher detection speed, can quickly respond to the requirements under different observation conditions, and provides more reliable data support for the research on aspects such as the composition of extraterrestrial substances and the lunar geological environment.

[0113] Compared with the prior art, the beneficial effects of the present invention also lie in:

[0114] 1. The present invention uses the improved yolo-R model to achieve efficient lunar meteorite identification: The improved yolo-R model can maintain efficient target detection performance when processing large-scale remote sensing images, and its fast feature extraction and detection capabilities are particularly suitable for the lunar meteorite identification task. It has good application prospects in the future real-time target detection related to the landing of detectors.

[0115] 2. The present invention adopts the method of rotation target detection for lunar meteorite detection. Compared with the conventional horizontal target detection, the rotation target detection can effectively capture the rotation features in uncertain directions of the target, specifically manifested as the traces left on the lunar surface by the collision around the lunar meteorite, thereby reducing the misdetection of the model that identifies ordinary lunar surface rocks as lunar meteorites and improving the accuracy of target detection.

[0116] According to the method described in the above embodiments, this embodiment will be further described from the perspective of a lunar rockfall rotation target detection device based on the yolo-R model. The lunar rockfall rotation target detection device based on the yolo-R model can be specifically implemented as an independent entity or integrated in an electronic device, which can be a device such as a terminal, a server, etc. Among them, the terminal can include a tablet computer, a notebook computer, a personal computer (PC), a microprocessing box, or other devices, etc.

[0117] Please refer to Figure 4 , Figure 4 Specifically described is the lunar rockfall rotation target detection device based on the yolo-R model provided in the embodiments of the present application, which is applied to an electronic device. The lunar rockfall rotation target detection device based on the yolo-R model may include:

[0118] An acquisition module, configured to acquire a lunar remote sensing image data training set; wherein, the lunar remote sensing image training data set includes sliced images of lunar remote sensing images and four-point labels of lunar rockfalls;

[0119] A processing module, configured to input the sliced images of the lunar remote sensing images into an improved yolo-R model to obtain a rotation target prediction result;

[0120] A training module, configured to construct a loss function based on the rotation target prediction result and the four-point labels of the lunar rockfalls, and train the improved yolo-R model based on the loss function;

[0121] A rotation target detection module, configured to acquire sliced images of a lunar remote sensing image to be detected, input the sliced images of the lunar remote sensing image to be detected into the trained improved yolo-R model, and obtain a rotation target prediction result of the sliced images of the lunar remote sensing image to be detected.

[0122] In specific implementation, the above-mentioned various modules and / or units can be implemented as independent entities or can be combined arbitrarily to be implemented as the same or several entities. For the specific implementation of the above-mentioned various modules and / or units, reference can be made to the method embodiments described above. For the specific beneficial effects that can be achieved, please also refer to the beneficial effects in the method embodiments described above, which will not be elaborated here.

[0123] In addition, an embodiment of the present application further provides an electronic device, which may be a device such as a computer or a tablet computer. The electronic device may implement the steps in any embodiment of the lunar rock rotation target detection method based on the yolo-R model provided by the embodiments of the present application. Therefore, the beneficial effects achievable by any lunar rock rotation target detection method based on the yolo-R model provided by the embodiments of the present invention can be achieved. For details, refer to the previous embodiments and will not be elaborated herein.

[0124] Figure 5 FIG. shows a specific structural block diagram of the electronic device provided by an embodiment of the present invention. The electronic device may be used to implement the lunar rock rotation target detection method based on the yolo-R model provided in the above embodiment. The electronic device 500 may be a device such as a terminal or a server. Among them, the terminal may include a tablet computer, a notebook computer, a personal computer (PC), a microprocessing box, or other devices, etc.

[0125] The RF circuit 510 is used to receive and transmit electromagnetic waves, realizing the mutual conversion between electromagnetic waves and electrical signals, so as to communicate with a communication network or other devices. The RF circuit 510 may include various existing circuit components for performing these functions. For example, an antenna, a radio frequency transceiver, a digital signal processor, an encryption / decryption chip, a subscriber identity module (SIM) card, a memory, and so on. The RF circuit 510 can communicate with various networks such as the Internet, an enterprise intranet, a wireless network or communicate with other devices through a wireless network. The above-mentioned wireless network may include a cellular phone network, a wireless local area network or a metropolitan area network. The above-mentioned wireless network can use various communication standards, protocols and technologies, including but not limited to Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (WCDMA), Code Division Access (CDMA), Time Division Multiple Access (TDMA), Wireless Fidelity (Wi-Fi) (such as Institute of Electrical and Electronics Engineers standards IEEE 802.11a, IEEE 802.11b, IEEE 802.11g and / or IEEE 802.11n), Voice over Internet Protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging and short messages, and any other suitable communication protocols, and may even include those protocols that have not yet been developed currently.

[0126] The memory 520 can be used to store software programs and modules, such as the corresponding program instructions / modules in the above embodiments. The processor 580 executes various functional applications and data processing by running the software programs and modules stored in the memory 520, that is, realizes functions such as taking pictures with the front camera, processing the captured images, and switching the display colors of the display content on the display screen. The memory 520 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 520 may further include a memory remotely disposed relative to the processor 580, and these remote memories can be connected to the electronic device 500 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and their combinations.

[0127] The input unit 530 can be used to receive input digital or character information, as well as generate a keyboard and a mouse related to user settings and function controls.

[0128] The display unit 540 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces, and these graphical user interfaces can be composed of graphics, text, icons, videos, and any combination thereof. The display unit 540 may include a display panel 541. Optionally, the display panel 541 can be configured in the form of an LCD (Liquid Crystal Display) or an OLED (Organic Light-Emitting Diode).

[0129] The audio circuit 560, speaker 561, and microphone 562 can provide an audio interface between the user and the electronic device 500. The audio circuit 560 can transmit the electrical signal converted from the received audio data to the speaker 561, and the speaker 561 converts it into a sound signal for output; on the other hand, the microphone 562 converts the collected sound signal into an electrical signal, which is received by the audio circuit 560 and then converted into audio data. After the audio data is output to the processor 580 for processing, it is sent through the RF circuit 510 to, for example, another terminal, or the audio data is output to the memory 520 for further processing. The audio circuit 560 may also include an earphone jack to provide communication between the peripheral earphone and the electronic device 500.

[0130] The electronic device 500 can help the user receive requests, send information, etc. through the transmission module 570 (such as a Wi-Fi module), and it provides the user with wireless broadband Internet access. Although the transmission module 570 is shown in the figure, it can be understood that it does not belong to the essential components of the electronic device 500 and can be completely omitted within the scope of not changing the essence of the invention according to needs.

[0131] The processor 580 is the control center of the electronic device 500, connecting various parts of the entire mobile phone through various interfaces and circuits. By running or executing software programs and / or modules stored in the memory 520, and by calling the data stored in the memory 520, it executes various functions of the electronic device 500 and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 580 may include one or more processing cores; in some embodiments, the processor 580 may integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor may not be integrated into the processor 580 either.

[0132] The electronic device 500 also includes a power supply 590 (such as a battery) for supplying power to each component. In some embodiments, the power supply can be logically connected to the processor 580 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power supply 590 may also include any components such as one or more DC or AC power supplies, a recharge system, a power failure detection circuit, a power converter or inverter, and a power status indicator.

[0133] Although not shown, the electronic device 500 also includes a camera (such as a front camera, a rear camera), a Bluetooth module, etc., which will not be elaborated here. Specifically, in this embodiment, the display unit of the electronic device is a touch screen display, and the mobile terminal also includes a memory, and one or more programs, where one or more programs are stored in the memory and are configured to be executed by one or more processors. One or more programs include instructions for performing the following operations:

[0134] Obtain a training set of lunar remote sensing image data; wherein, the lunar remote sensing image training data set includes sliced images of lunar remote sensing images and four-point labels of lunar rockfalls;

[0135] Input the sliced images of the lunar remote sensing images into the improved yolo-R model to obtain a rotation target prediction result;

[0136] Construct a loss function based on the rotation target prediction result and the four-point labels of the lunar rockfalls, and train the improved yolo-R model based on the loss function;

[0137] Obtain sliced images of the lunar remote sensing image to be detected, and input the sliced images of the lunar remote sensing image to be detected into the trained improved yolo-R model to obtain a rotation target prediction result of the sliced images of the lunar remote sensing image to be detected.

[0138] In specific implementation, each of the above modules can be implemented as an independent entity, or can be arbitrarily combined and implemented as the same or several entities. For the specific implementation of each of the above modules, reference can be made to the foregoing method embodiments, which will not be elaborated herein.

[0139] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling relevant hardware through instructions. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. For this purpose, an embodiment of the present invention provides a storage medium, in which multiple instructions are stored, and these instructions can be loaded by a processor to execute the steps of any one of the embodiments of the lunar falling rock rotation target detection method based on the yolo-R model provided by the embodiments of the present invention.

[0140] Among them, the computer-readable storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

[0141] Since the instructions stored in this storage medium can execute the steps of any one of the embodiments of the lunar falling rock rotation target detection method based on the yolo-R model provided by the embodiments of the present invention, the beneficial effects achievable by any of the lunar falling rock rotation target detection methods based on the yolo-R model provided by the embodiments of the present invention can be achieved. For details, refer to the foregoing embodiments, which will not be elaborated herein.

[0142] The above has introduced in detail a lunar falling rock rotation target detection method, device, storage medium, and electronic device provided by the embodiments of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for detecting rotating lunar rockfall targets based on the yolo-R model, characterized in that, The method includes: Obtaining a training set of lunar remote sensing image data; wherein, the training data set of lunar remote sensing images includes sliced images of lunar remote sensing images and four-point labels of lunar fallen rocks; Inputting the sliced images of the lunar remote sensing images into an improved yolo-R model to obtain a rotation target prediction result; the improved yolo-R model includes a backbone network module, a feature extraction module, and an improved head output module; The processing process of the improved yolo-R model is: inputting the sliced images of the lunar remote sensing images into the backbone network module to obtain a first feature map; inputting the first feature map into the feature extraction module to obtain a second feature map; inputting the second feature map into the improved head output module to obtain multiple rotation target candidate boxes; calculating the confidence of the multiple rotation target candidate boxes, and taking the rotation target candidate box with the highest confidence as the rotation target prediction result; The improved head output module includes a raw convolutional layer, an improved convolutional layer, a pooling layer, and a connection layer connected in sequence; The processing process of the improved head output module is: inputting the second feature map into the raw convolutional layer to obtain prediction candidate boxes; inputting the prediction candidate boxes into the improved convolutional layer to extract angle information, and obtaining rotation target candidate boxes based on the angle information and the prediction candidate boxes; inputting the rotation target candidate boxes into the pooling layer for classification; inputting the classified rotation target candidate boxes into the connection layer to obtain the final rotation target candidate boxes; Constructing a loss function based on the rotation target prediction result and the four-point label of the lunar fallen rock, and training the improved yolo-R model based on the loss function; Obtaining a sliced image of the lunar remote sensing image to be detected, and inputting the sliced image of the lunar remote sensing image to be detected into the trained improved yolo-R model to obtain a rotation target prediction result of the sliced image of the lunar remote sensing image to be detected.

2. The method for detecting the rotating target of lunar falling rocks based on the yolo-R model according to claim 1, wherein The obtaining of the training set of lunar remote sensing image data includes: Obtaining lunar remote sensing image data; Performing orthogonal projection on the lunar remote sensing image data, and slicing the orthogonally projected lunar remote sensing image data to obtain sliced images of the lunar remote sensing images; Based on the longitude, latitude, and diameter information of the lunar fallen rock, obtaining the four-point label of the lunar fallen rock; Constructing a training data set of lunar remote sensing image data based on the sliced images of the lunar remote sensing images and the four-point labels of the lunar fallen rocks.

3. The lunar rock rotation target detection method based on the yolo-R model according to claim 1, characterized in that The constructing of the loss function based on the rotation target prediction result and the four-point label of the lunar fallen rock includes: Calculating the label area based on the four-point label of the lunar fallen rock; Calculating the predicted area based on the rotation target prediction result; Calculating the overlapping area between the label area and the predicted area; Calculating the loss function based on the label area, the predicted area, and the overlapping area.

4. The method for detecting the rotating target of lunar falling rocks based on the yolo-R model according to claim 3, characterized in that, The loss function is: Among them, is the loss function, is the area of the label region, is the area of the prediction region, is the area of the overlapping region.

5. The method for detecting the rotating target of lunar falling rocks based on the yolo-R model according to claim 3, characterized in that, The calculating of the overlapping area between the label area and the predicted area includes: Calculating the left boundary, right boundary, upper boundary, and lower boundary of the four-point label and the rotation target prediction result respectively; Calculate the overlapping area through the left boundary, the right boundary, the upper boundary, and the lower boundary, which can be expressed as: Among them, is the area of the overlapping region, L is the left boundary, R is the right boundary, U is the upper boundary, D is the lower boundary.

6. A lunar rock rotation target detection device based on the YOLO-R model, the lunar rock rotation target detection device based on the YOLO-R model is used to implement the lunar rock rotation target detection method based on the YOLO-R model described in claim 1, characterized in that, Including: An acquisition module, configured to acquire a training set of lunar remote sensing image data; wherein, the lunar remote sensing image training data set includes sliced images of lunar remote sensing images and four-point labels of lunar rockfalls; A processing module, configured to input the sliced images of the lunar remote sensing images into an improved yolo-R model to obtain a rotation target prediction result; A training module, configured to construct a loss function based on the rotation target prediction result and the four-point labels of the lunar rockfalls, and train the improved yolo-R model based on the loss function; A rotation target detection module, configured to acquire sliced images of the lunar remote sensing image to be detected, and input the sliced images of the lunar remote sensing image to be detected into the trained improved yolo-R model to obtain a rotation target prediction result of the sliced images of the lunar remote sensing image to be detected.

7. A computer-readable storage medium, characterized in that, Multiple instructions are stored in the computer-readable storage medium, and the instructions are suitable for being loaded by a processor to execute the lunar rockfall rotation target detection method based on the yolo-R model according to any one of claims 1 to 5.

8. An electronic device, characterized in that, Including a processor and a memory, the processor is electrically connected to the memory, the memory is used to store instructions and data, and the processor is used to execute the steps in the lunar rockfall rotation target detection method based on the yolo-R model according to any one of claims 1 to 5.

Images