Mine obstacle detection method and device, mine car and storage medium

Abstract

The invention discloses a mine obstacle detection method and device, a mine car and a storage medium, and the method comprises the steps: obtaining a mine road image; a mine road image is input to a target mine obstacle detection model, a corresponding mine obstacle detection result is obtained, the target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module, and the adaptive attention module is used for extracting position features of small target mine obstacles in the mine road image and sending the position features to the target mine obstacle detection model; the mine obstacle detection result comprises a first detection result used for representing that the mine road image does not contain the mine obstacle information and a second detection result used for representing that the mine road image contains the mine obstacle information. According to the method, mine obstacle detection is carried out on the mine road image based on the target mine obstacle detection model, and the detection accuracy of mine obstacle detection is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of mine obstacle detection, and in particular, relates to a mine obstacle detection method, a detection device, a mine car and a storage medium. Background Art

[0002] The open-pit mine road environment is complex and is often accompanied by mine obstacles such as rock piles, earth slopes, deep pits, retaining walls, pedestrians and auxiliary operation vehicles. In order to prevent the mine cars traveling on the mine roads from colliding with the mine obstacles, it is necessary to perform mine obstacle detection on the mine roads where the mine cars travel.

[0003] At present, mine obstacle detection is mainly performed on the collected mine road images through the YOLO (You Only Live Once, YOLO) target detection model or the Single Shot MultiBox Detector (SSD) model. Since the network model structure of the YOLO target detection model and the SSD model is relatively simple, it is impossible to effectively learn the characteristic information of small target mine obstacles in complex mine road scenes, resulting in low detection accuracy of the YOLO target detection model or the SSD model for mine obstacle detection. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a mine obstacle detection method, a detection device, a mine car and a storage medium to overcome or at least partially solve the above problems of the prior art.

[0005] In a first aspect, an embodiment of the present application provides an obstacle detection method, including: acquiring a mine road image; inputting the mine road image into a target mine obstacle detection model to obtain a corresponding mine obstacle detection result, wherein the target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module, and the adaptive attention module is used to extract position features of small target mine obstacles in the mine road image, and the mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains mine obstacle information.

[0006] In the second aspect, the embodiment of the present application provides a mine obstacle detection device, including an image acquisition module and an input module. The image acquisition module is used to acquire a mine road image; the input module is used to input the mine road image into a target mine obstacle detection model to obtain a corresponding mine obstacle detection result, the target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module, the adaptive attention module is used to extract the position features of small target mine obstacles in the mine road image, and the mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains mine obstacle information.

[0007] In a third aspect, an embodiment of the present application provides a mining car, comprising a memory; one or more processors coupled to the memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to execute the mine obstacle detection method provided in the first aspect above.

[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a program code is stored. The program code can be called by a processor to execute the mine obstacle detection method provided in the first aspect above.

[0009] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product is run on a computer device, the computer device executes the mine obstacle detection method provided in the first aspect above.

[0010] The solution provided by the present application obtains a mine road image and inputs the mine road image into a target mine obstacle detection model to obtain a corresponding mine obstacle detection result. The target mine obstacle detection model is obtained based on the YOLO target detection model and an adaptive attention module. The adaptive attention module is used to extract the position characteristics of small target mine obstacles in the mine road image. The mine obstacle detection result includes a first detection result used to characterize that the mine road image does not contain mine obstacle information, and a second detection result used to characterize that the mine road image contains mine obstacle information. The target mine obstacle detection model is constructed according to the YOLO target detection model and the adaptive attention module, which is conducive to improving the characterization ability of the target mine obstacle detection model. The target mine obstacle detection model is used to perform mine obstacle detection on the mine road image, which improves the learning ability of the target mine obstacle detection model to learn the position characteristics of small target mine obstacles, making small target mine obstacles easier to detect and improving the detection accuracy of mine obstacles. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 A schematic diagram of a scenario of a mine obstacle detection system provided in an embodiment of the present application is shown.

[0013] Figure 2 A schematic flow chart of a mine obstacle detection method provided in an embodiment of the present application is shown.

[0014] Figure 3 A structural schematic diagram of an adaptive attention module in the mine obstacle detection method provided in an embodiment of the present application is shown.

[0015] Figure 4 Another flow chart of the mine obstacle detection method provided in the embodiment of the present application is shown.

[0016] Figure 5 A structural block diagram of a mine obstacle detection device provided in an embodiment of the present application is shown.

[0017] Figure 6 A functional block diagram of a mine car provided in an embodiment of the present application is shown.

[0018] Figure 7 A computer-readable storage medium provided in an embodiment of the present application for storing or carrying a program code for implementing a mine obstacle detection method provided in an embodiment of the present application is shown.

[0019] Figure 8 A computer program product provided in an embodiment of the present application and used for storing or carrying program codes for implementing the mine obstacle detection method provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0020] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described below are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0022] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates other working conditions.

[0023] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0024] In addition, in the description of the present application, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0025] The open-pit mine road environment is complex and is often accompanied by mine obstacles such as rock piles, earth slopes, deep pits, retaining walls, pedestrians and auxiliary operation vehicles. In order to prevent the mine cars traveling on the mine roads from colliding with the mine obstacles, it is necessary to perform mine obstacle detection on the mine roads where the mine cars travel.

[0026] At present, mine obstacle detection is mainly performed on the collected mine road images through the YOLO (You Only Live Once, YOLO) target detection model or the Single Shot MultiBox Detector (SSD) model. Since the network model structure of the YOLO target detection model and the SSD model is relatively simple, it is impossible to effectively learn the characteristic information of small target mine obstacles in complex mine road scenes, resulting in low detection accuracy of the YOLO target detection model or the SSD model for mine obstacle detection.

[0027] In view of the above problems, the mine obstacle detection method, detection device, mine car and storage medium provided in the embodiments of the present application obtain a corresponding mine obstacle detection result by acquiring a mine road image and inputting the mine road image into a target mine obstacle detection model. The target mine obstacle detection model is obtained based on the YOLO target detection model and an adaptive attention module. The adaptive attention module is used to extract the position characteristics of small target mine obstacles in the mine road image. The mine obstacle detection result includes a first detection result used to characterize that the mine road image does not contain mine obstacle information, and a second detection result used to characterize that the mine road image contains mine obstacle information. The target mine obstacle detection model is constructed according to the YOLO target detection model and the adaptive attention module, which is conducive to improving the characterization ability of the target mine obstacle detection model. The target mine obstacle detection model is used to perform mine obstacle detection on the mine road image, which improves the learning ability of the target mine obstacle detection model to learn the position characteristics of small target mine obstacles, making small target mine obstacles easier to detect and improving the detection accuracy of mine obstacles.

[0028] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0029] See also Figure 1 , which shows a schematic diagram of an application scenario of the mine obstacle detection system provided in an embodiment of the present application, which may include a processing device 100 and a mine car 200. The processing device 100 may be used to obtain a mine road image of the mine road on which the mine car 200 travels, and detect mine obstacles on the mine road based on the mine road image.

[0030] The processing device 100 may be a terminal device or a server, etc. The type of the processing device 100 is not limited here and may be specifically configured according to actual needs.

[0031] The terminal device can be a mobile terminal device (for example, any one of a mobile phone, a personal digital assistant (PDA), a tablet personal computer (Tablet PC), a laptop computer, a smart watch, a smart bracelet or a wearable device, etc.), or a fixed terminal device (desktop computer, smart panel, etc.), etc.

[0032] The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or any one of the cloud servers that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), big data or artificial intelligence platforms, etc.

[0033] The mine car 200 can be any one of an electric mine car (for example, an electric car, a battery car, etc.), a hybrid mine car (for example, a hybrid electric vehicle (HEV)), a fuel mine car or a gas mine car, etc. The type of the mine car 200 is not limited here and can be specifically set according to actual needs.

[0034] The mine obstacle may be at least any one of a pile of rocks, a soil slope, a deep pit, a retaining wall, a pedestrian or an auxiliary operation vehicle, etc. The type of the mine obstacle is not limited here and can be set according to actual needs.

[0035] In some embodiments, the processing device 100 can be set on the mine car 200 and form an integral body with the mine car 200, or the processing device 100 can be set independently from the mine car 200, etc. The setting method between the processing device 100 and the mine car 200 is not limited here, and can be set specifically according to actual needs.

[0036] See also Figure 2 , which shows a flow chart of a mine obstacle detection method provided by an embodiment of the present application. In a specific embodiment, the mine obstacle detection method can be applied to Figure 1 The processing device 100 in the mine obstacle detection system shown in FIG. 1 is taken as an example to describe the processing device 100 in the mine obstacle detection system. Figure 2 The process shown is described in detail, and the mine obstacle detection method may include the following steps S110 to S120.

[0037] Step S110: Acquire a mine road image.

[0038] In an embodiment of the present application, when a user needs to detect mine obstacles on a mine road on which a mine car travels, a detection instruction can be sent to a processing device, and the processing device receives and responds to the detection instruction to obtain a mine road image of the mine road.

[0039] In some embodiments, the processing device may be provided with a camera. When a user needs to detect mine obstacles on the mine road on which the mine car travels, a detection instruction may be sent to the processing device. The processing device receives and responds to the detection instruction, controls the camera to capture images of the mine road on which the mine car travels, and obtains corresponding mine road images.

[0040] Among them, the camera can be any one of a white light camera, an infrared camera, a laser camera, etc. The type of camera is not limited here and can be set according to actual needs.

[0041] In some embodiments, the mine obstacle detection system may further include a camera, which is connected to the processing device via a network and exchanges data with the processing device via the network. The camera may be used to capture images of the mine road on which the mine car travels.

[0042] When the user needs to detect mine obstacles on the mine road where the mine car travels, a detection instruction can be sent to the processing device. The processing device receives and responds to the detection instruction and sends an image acquisition instruction to the camera via the network. The camera receives and responds to the image acquisition instruction, performs image acquisition on the mine road where the mine car travels, obtains the corresponding mine road image, and sends the mine road image to the processing device via the network. The processing device receives the mine road image returned by the camera.

[0043] Among them, the network can be any one of a ZigBee network, a Bluetooth (BT) network, a Wireless Fidelity (Wi-Fi) network, a Thread network, a Long Range Radio (LoRa) network, a Low-Power Wide-Area Network (LPWAN), an infrared network, a Narrow Band Internet of Things (NB-IoT), a Controller Area Network (CAN), a Digital Living Network Alliance (DLNA) network, a Wide Area Network (WAN), a Local Area Network (LAN), a Metropolitan Area Network (MAN) or a Wireless Personal Area Network (WPAN), etc. The type of network is not limited here and can be set according to actual needs.

[0044] In some embodiments, the processing device may include an input panel. When a user needs to detect mine obstacles on a mine road where a mine car travels, the user may input a detection instruction on the input panel of the processing device, and the processing device receives the detection instruction through the input panel.

[0045] In some embodiments, the processing device may be provided with a voice recognition module. When a user needs to detect mine obstacles on a mine road on which a mine car travels, the user may send voice information within the voice collection range of the voice recognition module. The voice recognition module collects the voice information sent by the user, and performs voice recognition on the collected voice information. Based on the recognition result of the voice recognition, it is determined that the recognition result contains keywords for indicating the detection of mine obstacles on the mine road on which the mine car travels, such as "mine obstacle detection", and for example, "mine obstacles" and "detection", etc., and it is determined that a detection instruction for detecting mine obstacles on the mine road on which the mine car travels is received.

[0046] As an example, the voice information sent by the user is: detect mine obstacles on the mine road where the mine car travels, and the recognition result of the voice recognition contains the keyword "detect mine obstacles", then it is determined that a detection instruction for detecting mine obstacles on the mine road where the mine car travels is received.

[0047] In some embodiments, the mine obstacle detection system may further include a client, which is connected to the processing device via a network and performs data exchange with the processing device via the network.

[0048] When the user needs to detect mine obstacles on the mine road where the mine car travels, a detection instruction can be sent to the client. The client receives and responds to the detection instruction, forwards the detection instruction to the processing device through the network, and the processing device receives the detection instruction forwarded by the client.

[0049] Among them, the client can be any one of a mobile client (for example, a mobile phone client, a personal digital assistant (PDA) client, a tablet computer (Tablet Personal Computer, Tablet PC) client, a laptop computer client, a smart watch client, a smart bracelet client or a wearable client, etc.) or a fixed client (for example, a desktop computer client, a smart panel client, etc.). The type of client is not limited here and can be set according to actual needs.

[0050] Step S120: inputting the mine road image into the target mine obstacle detection model to obtain the corresponding mine obstacle detection result.

[0051] In an embodiment of the present application, after acquiring a mine road image, the processing device may input the mine road image into a target mine obstacle detection model. The target mine obstacle detection model receives and responds to the mine road image, detects mine obstacles on the mine road according to the mine road image, obtains corresponding mine obstacle detection results, and outputs the mine obstacle detection results to the processing device. The processing device receives the mine obstacle detection results output by the target mine obstacle detection model. The target mine obstacle detection model may be obtained based on a YOLO target detection model and an adaptive attention module. The adaptive attention module may be used to extract position features of small target mine obstacles in the mine road image, thereby realizing construction of a target mine obstacle detection model based on the YOLO target detection model and the adaptive attention module, which is beneficial to improving the characterization capability of the target mine obstacle detection model. The target mine obstacle detection model is used to perform mine obstacle detection on the mine road image, thereby improving the learning capability of the target mine obstacle detection model to learn the position features of small target mine obstacles, making small target mine obstacles easier to detect, and improving the detection accuracy of mine obstacles.

[0052] The mine obstacle detection result may include a first detection result for indicating that the mine road image does not contain mine obstacle information, and a second detection result for indicating that the mine road image contains mine obstacle information.

[0053] The mine obstacle information may include the mine obstacle type, the mine obstacle location, the mine obstacle size, etc. The type of the mine obstacle information is not limited here and can be set according to actual needs.

[0054] In some embodiments, the adaptive attention module can be set in the cross-stage partial (CSP) module of the YOLO target detection model. The adaptive attention module can dynamically adjust the position weights of different positions in the corresponding feature map according to the position information of the small target mine obstacle in the mine road image, so that the target mine obstacle detection model pays more attention to the position features related to the small target mine obstacles. The target mine obstacle detection model can concentrate on processing the areas closely related to the small target mine obstacles, improve the perception and positioning ability of the target mine obstacle detection model for the small target mine obstacles, and improve the learning ability of the target mine obstacle detection model to learn the position features of the small target mine obstacles, so that the small target mine obstacles are easier to be detected, and the detection accuracy of the target mine obstacle detection model for detecting mine obstacles is improved.

[0055] like Figure 3As shown, the adaptive attention module may include a convolution submodule, a Sigmoid activation function submodule, a normalization submodule, a reshaping broadcast submodule, a feature weighting submodule, a fusion submodule, and a rectified linear unit (ReLU) activation function submodule connected in sequence.

[0056] The convolution submodule can be used to extract features from the mine road image to obtain a feature map; the Sigmoid activation function submodule can be used to generate a corresponding spatial attention map based on the feature map, and the attention value range of the spatial attention map is [0, 1]; the normalization submodule can be used to normalize the spatial attention map to obtain a normalized attention map.

[0057] The reshaping broadcast submodule can be used to reshape the dimension of the attention weight of the normalized attention map to the same dimension as the feature map, obtain the reshaped attention map, and broadcast the reshaped attention map.

[0058] The feature weighting submodule can be used to perform weighted processing on the feature map and the reshaped attention map to obtain a weighted feature map; the fusion submodule can be used to fuse the channels of the weighted feature map to obtain a fused feature map; the ReLU activation function submodule can be used to perform ReLU activation function processing on the fused feature map.

[0059] In some embodiments, after the processing device inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that a mine obstacle is detected on the mine road, and corresponding obstacle disposal prompt information can be generated. The obstacle disposal prompt information can be used by the user to dispose of the mine obstacle, or / and the mine car to avoid collision between the mine car and the mine obstacle, thereby improving the driving safety of the mine car.

[0060] Among them, the obstacle handling prompt information can be at least any one of sound handling prompt information, text handling prompt information or light handling prompt information, etc. The type of obstacle handling prompt information is not limited here and can be set according to actual needs.

[0061] As an implementation method, the obstacle handling prompt information can be sound handling prompt information. After the processing device inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that a mine obstacle is detected on the mine road, and corresponding sound handling prompt information can be generated.

[0062] As an implementation method, the obstacle handling prompt information can be text handling prompt information. After the processing equipment inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that a mine obstacle is detected on the mine road, and the corresponding text handling prompt information can be generated.

[0063] As an implementation method, the obstacle handling prompt information can be a lighting handling prompt information. After the processing device inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that a mine obstacle is detected on the mine road, and the corresponding lighting handling prompt information can be generated.

[0064] As an implementation method, the obstacle handling prompt information can be sound handling prompt information and text handling prompt information. After the processing device inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that the existence of mine obstacles is detected on the mine road, and corresponding sound handling prompt information and text handling prompt information can be generated.

[0065] As an implementation method, the obstacle handling prompt information can be sound handling prompt information and light handling prompt information. After the processing device inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that the existence of mine obstacles is detected on the mine road, and corresponding sound handling prompt information and light handling prompt information can be generated.

[0066] As an implementation method, the obstacle handling prompt information can be text handling prompt information and light handling prompt information. After the processing equipment inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that a mine obstacle is detected on the mine road, and corresponding text handling prompt information and light handling prompt information can be generated.

[0067] As an implementation method, the obstacle handling prompt information can be sound handling prompt information, text handling prompt information and light handling prompt information. After the processing equipment inputs the mine road image into the target mine obstacle detection model and obtains the corresponding mine obstacle detection result, when it receives the second detection result output by the target mine obstacle detection model, it indicates that the existence of mine obstacles is detected on the mine road, and the corresponding sound handling prompt information, text handling prompt information and light handling prompt information can be generated.

[0068] The solution provided in this embodiment obtains a corresponding mine obstacle detection result by acquiring a mine road image and inputting the mine road image into a target mine obstacle detection model. The target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module. The adaptive attention module is used to extract the position features of small target mine obstacles in the mine road image. The mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains mine obstacle information. The target mine obstacle detection model is constructed according to the YOLO target detection model and the adaptive attention module, which is beneficial to improving the characterization ability of the target mine obstacle detection model. The target mine obstacle detection model is used to perform mine obstacle detection on the mine road image, which improves the learning ability of the target mine obstacle detection model to learn the position features of small target mine obstacles, making small target mine obstacles easier to detect and improving the detection accuracy of mine obstacles.

[0069] See also Figure 4 , which shows a flow chart of a mine obstacle detection method provided by another embodiment of the present application. In a specific embodiment, the mine obstacle detection method can be applied to Figure 1 The processing device 100 in the mine obstacle detection system shown in FIG. 1 is taken as an example to describe the processing device 100 in the mine obstacle detection system. Figure 4 The process shown is described in detail, and the mine obstacle detection method may include the following steps S210 to S260.

[0070] Step S210: Acquire a mine road image.

[0071] In this embodiment, step S210 may refer to the contents of the corresponding steps in the aforementioned embodiments, which will not be described in detail here.

[0072] Step S220: The cross-stage local module and the adaptive attention module of the YOLO target detection model are integrated to obtain an initial mine obstacle detection model.

[0073] In this embodiment, after acquiring the mine road image, the processing device can fuse the CSP module and the adaptive attention module of the YOLO target detection model to obtain an initial mine obstacle detection model.

[0074] In some embodiments, the YOLO target detection model may be a YOLOv5 target detection model. After acquiring the mine road image, the processing device may fuse the CSP module and the adaptive attention module of the YOLOv5 target detection model to obtain an initial mine obstacle detection model.

[0075] Among them, the YOLOv5 target detection model is a high-performance target detection model. The YOLOv5 target detection model adopts a single-stage detection method based on deep learning, which is simple, fast and accurate.

[0076] The YOLOv5 target detection model can include the YOLOv5s target detection model, the YOLOv5m target detection model, the YOLOv5l target detection model, and the YOLOv5x target detection model.

[0077] In this embodiment, since the YOLOv5s target detection model has a smaller computational complexity, a smaller network depth, and a smaller width ratio, the YOLO target detection model can preferably use the YOLOv5s target detection model, which can improve the computing speed of the YOLO target detection model and the detection efficiency of the target mine obstacle detection model.

[0078] Step S230: Obtain a corresponding sample set based on the historical mine obstacle image.

[0079] In this embodiment, the processing device may obtain historical mine obstacle images, and obtain corresponding sample sets according to the historical mine obstacle images, wherein the sample sets may include training sets and test sets.

[0080] Specifically, the processing device can obtain historical mine obstacle images, and annotate the historical mine obstacle images to obtain corresponding annotated images, and divide the annotated images according to preset division rules to obtain corresponding training sets and test sets.

[0081] When the processing device annotates the historical mine obstacle image, it mainly annotates the mine obstacle type, mine obstacle position, and mine obstacle size in the historical mine obstacle image. For example, the mine obstacle type in the historical mine obstacle image is annotated with an annotation box, the corner point position of the annotation box is annotated, and the size of the annotation box is annotated.

[0082] The preset division rule is an artificial division rule. For example, the preset division rule may be an artificial division rule of training set: test set = 9:7. When the number of historical mine obstacle images is 16,000, the number of training set images is 9,000 and the number of test set images is 7,000. The preset division rule may be an artificial division rule of training set: test set = 1:1. When the number of historical mine obstacle images is 30,000, the number of training set images is 15,000 and the number of test set images is 15,000. The division method of the preset division rule is not limited here and can be set according to actual needs.

[0083] Step S240: training the initial mine obstacle detection model according to the training set to obtain a target mine obstacle detection model.

[0084] In this embodiment, after the processing device obtains the corresponding sample set based on the historical mine obstacle image, it can input the training set into the initial mine obstacle detection model. The initial mine obstacle detection model receives and responds to the training set, and is trained according to the training set to obtain the target mine obstacle detection model.

[0085] In some embodiments, after the processing device obtains the corresponding sample set based on the historical mine obstacle images, it can copy and paste the data enhancement processing on the training set to obtain the corresponding data enhancement training set, and input the data enhancement training set to the initial mine obstacle detection model. The initial mine obstacle detection model receives and responds to the data enhancement training set, and is trained according to the data enhancement training set to obtain the target mine obstacle detection model, thereby realizing the training of the initial mine obstacle detection model according to the data enhancement training set, avoiding the low robustness of the target mine obstacle detection model obtained by training the initial mine obstacle detection model with a small training set due to the small number of historical mine obstacle images, and increasing the robustness of the target mine obstacle detection model.

[0086] Among them, the processing process of copying and pasting data enhancement of the historical mine obstacle image by the processing equipment can be to copy and paste the small target mine obstacles in the historical mine obstacle image in the same historical mine obstacle image, for example, copy and paste the small target mine obstacles in the historical mine obstacle image to different positions in the same historical mine obstacle image, rotate the copied and pasted small target mine obstacles, scale the copied and pasted small target mine obstacles, or adjust the contrast of the copied and pasted small target mine obstacles, etc.

[0087] The processing process of copying and pasting data enhancement of historical mine obstacle images by the processing device can also be to copy and paste small target mine obstacles in historical mine obstacle images in different historical mine obstacle images, for example, copy and paste small target mine obstacles in historical mine obstacle images to different positions in different historical mine obstacle images, rotate the copied and pasted small target mine obstacles, scale the copied and pasted small target mine obstacles, or adjust the contrast of the copied and pasted small target mine obstacles, etc.

[0088] The processing method of copying and pasting data enhancement processing of historical mine obstacle images by the processing device is not limited here, and can be set according to actual needs.

[0089] In some embodiments, after the processing device obtains the corresponding sample set based on the historical mine obstacle image, it can copy and paste the data enhancement processing on the training set to obtain the corresponding data enhancement training set, and determine the scale-aware loss function of the initial mine obstacle detection model, and train the initial mine obstacle detection model according to the scale-aware loss function and the data enhancement training set to obtain the target mine obstacle detection model. The scale-aware loss function can be used to adjust the loss weight of the initial mine obstacle detection model according to the mine obstacle size of the small target mine obstacle corresponding to the data enhancement training set during the training of the initial mine obstacle detection model, so that the small target mine obstacles receive more attention and importance, and the detection accuracy and robustness of the target mine obstacle detection model are further improved.

[0090] Among them, after the processing equipment obtains the corresponding sample set based on the historical mine obstacle image, it can copy and paste the data enhancement processing on the training set to obtain the corresponding data enhanced training set, and input the test set to the initial mine obstacle detection model. The initial mine obstacle detection model receives and responds to the test set, performs testing according to the test set, obtains the corresponding prediction box, and calculates the overlap between the real box and the prediction box corresponding to the test set, and determines the weight of the positioning loss function of the initial mine obstacle detection model based on the overlap, and determines the scale perception loss function of the initial mine obstacle detection model based on the weight of the positioning loss function and the positioning loss function.

[0091] The processing device determines the weight of the positioning loss function of the initial mine obstacle detection model according to the overlap between the real frame and the predicted frame corresponding to the test set, which can be a rule for setting the weight of the positioning loss function according to a preset rule. For example, when the overlap between the real frame and the predicted frame corresponding to the test set is small, the weight of the positioning loss function can be set to be large, indicating that the initial mine obstacle detection model pays more attention to the predicted frame that overlaps less with the real frame; when the overlap between the real frame and the predicted frame corresponding to the test set is large, the weight of the positioning loss function can be set to be small, indicating that the initial mine obstacle detection model pays less attention to the predicted frame that overlaps more with the real frame.

[0092] The processing device can calculate the weight of the positioning loss function and the product of the positioning loss function to obtain the scale perception loss function of the initial mine obstacle detection model.

[0093] In some embodiments, after the processing device obtains the corresponding sample set based on the historical mine obstacle image, it can copy and paste the data enhancement processing on the training set to obtain the corresponding data enhancement training set, and obtain the pixel area of ​​the real frame corresponding to the test set, and determine whether the historical mine obstacle corresponding to the test set is a small target mine obstacle according to the pixel area, and when it is determined that the historical mine obstacle corresponding to the test set is a small target mine obstacle according to the pixel area, the test set is input into the initial mine obstacle detection model to obtain the corresponding prediction frame, and the overlap between the real frame corresponding to the test set and the prediction frame is calculated, and the initial mine obstacle is determined according to the overlap. The weight of the positioning loss function of the obstacle detection model is determined, and the scale-aware loss function of the initial mine obstacle detection model is determined according to the weight of the positioning loss function and the positioning loss function. The initial mine obstacle detection model is trained according to the scale-aware loss function and the data enhancement training set to obtain the target mine obstacle detection model. When the historical mine obstacle is determined to be a small target mine obstacle, the loss weight of the initial mine obstacle detection model is adjusted according to the mine obstacle size of the small target mine obstacle corresponding to the data enhancement training set, thereby improving the detection accuracy and robustness of the target mine obstacle detection model.

[0094] Among them, when the pixel area is smaller than the preset pixel area threshold, the historical mine obstacle corresponding to the test set is determined to be a small target mine obstacle; when the pixel area is greater than or equal to the preset pixel area threshold, the historical mine obstacle corresponding to the test set is determined to be a non-small target mine obstacle.

[0095] The preset pixel area threshold can be used to characterize the maximum pixel area value of a small target mine obstacle. The preset pixel area value can be a pixel area value preset by a user, or a pixel area value automatically generated by a processing device according to a detection process of detecting mine obstacles multiple times. For example, the preset pixel area threshold can be 32 pixels * 32 pixels, the preset pixel area threshold can also be 50 pixels * 50 pixels, the preset pixel area threshold can also be 100 pixels * 100 pixels, etc. The setting method of the preset pixel area threshold is not limited here, and can be set according to actual needs.

[0096] In some embodiments, after the processing device obtains the corresponding sample set based on the historical mine obstacle image, it can copy and paste the data enhancement processing on the training set to obtain the corresponding data enhanced training set, and obtain the pixel area of ​​the real frame corresponding to the test set, and determine whether the historical mine obstacle corresponding to the test set is a small target mine obstacle according to the pixel area, and when it is determined according to the pixel area that the historical mine obstacle corresponding to the test set is a non-small target mine obstacle, determine the positioning loss function as the scale perception loss function of the initial mine obstacle detection model, and train the initial mine obstacle detection model according to the scale perception loss function and the data enhanced training set to obtain the target mine obstacle detection model, so as to avoid the increase of calculation amount caused by adjusting the damage weight of the large target mine obstacle during the training of the initial mine obstacle detection model, thereby improving the training efficiency of the initial mine obstacle detection model.

[0097] Step S250: Determine whether the target mine obstacle detection model converges.

[0098] In this embodiment, after the processing device trains the initial mine obstacle detection model according to the training set to obtain the target mine obstacle detection model, it can determine whether the target mine obstacle detection model converges.

[0099] Specifically, after the processing equipment trains the initial mine obstacle detection model according to the training set to obtain the target mine obstacle detection model, the test set can be input into the target mine obstacle detection model. The target mine obstacle detection model receives and responds to the test set, tests the test set, obtains the corresponding test results, and determines the test accuracy corresponding to the test results based on the test results and the test set, and determines whether the target mine obstacle detection model converges based on the test accuracy.

[0100] When the test accuracy is greater than or equal to the preset accuracy threshold, it is determined that the target mine obstacle detection model has converged; when the test accuracy is less than the preset accuracy threshold, it is determined that the target mine obstacle detection model has not converged.

[0101] Among them, the preset accuracy threshold can be used to characterize the minimum accuracy corresponding to the convergence of the target mine obstacle detection model. The preset accuracy threshold can be an accuracy pre-set by the user, or it can be an accuracy automatically generated by the processing equipment according to the detection process of multiple mine obstacle detections, etc. The setting method of the preset accuracy threshold is not limited here, and it can be set according to actual needs.

[0102] Step S260: When it is determined that the target mine obstacle detection model converges, the mine road image is input into the target mine obstacle detection model to obtain a corresponding mine obstacle detection result.

[0103] In this embodiment, when the processing device determines that the target mine obstacle detection model has converged, the mine road image can be input into the target mine obstacle detection model. The target mine obstacle detection model receives and responds to the mine road image, detects mine obstacles on the mine road according to the mine road image, obtains corresponding mine obstacle detection results, and outputs the mine obstacle detection results to the processing device. The processing device receives the mine obstacle detection results output by the target mine obstacle detection model, and implements mine obstacle detection on the mine road image based on the target mine obstacle detection model when it is determined that the target mine obstacle detection model has converged, thereby ensuring the stability of the target mine obstacle detection model and improving the detection credibility of mine obstacles.

[0104] The solution provided in this embodiment obtains an initial mine obstacle detection model by acquiring a mine road image and fusing the cross-stage local module and the adaptive attention module of the YOLO target detection model, and obtains a corresponding sample set based on the historical mine obstacle image, and trains the initial mine obstacle detection model based on the training set to obtain a target mine obstacle detection model, and determines whether the target mine obstacle detection model converges, and when it is determined that the target mine obstacle detection model converges, inputs the mine road image into the target mine obstacle detection model to obtain the corresponding mine obstacle detection result, and implements mine obstacle detection on the mine road image based on the target mine obstacle detection model when it is determined that the target mine obstacle detection model converges, thereby ensuring the stability of the target mine obstacle detection model and improving the detection credibility of the mine obstacle detection.

[0105] See also Figure 5 , which shows a mine obstacle detection device 300 provided by an embodiment of the present application. The mine obstacle detection device 300 can be applied to Figure 1 The processing device 100 in the mine obstacle detection system shown in FIG. 1 is taken as an example to describe the processing device 100 in the mine obstacle detection system. Figure 5The mine obstacle detection device 300 shown is described in detail. The mine obstacle detection device 300 may include an image acquisition module 310 and an input module 320 .

[0106] The image acquisition module 310 can be used to acquire a mine road image; the input module 320 can be used to input the mine road image into a target mine obstacle detection model to obtain a corresponding mine obstacle detection result. The target mine obstacle detection model can be obtained based on a YOLO target detection model and an adaptive attention module. The adaptive attention module can be used to extract the position features of small target mine obstacles in the mine road image. The mine obstacle detection result may include a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains mine obstacle information.

[0107] In some embodiments, the mine obstacle detection device 300 may further include a fusion module, a sample acquisition module, a training module, and a first determination module.

[0108] The fusion module can be used for inputting the mine road image into the target mine obstacle detection model through the input module 320, and before obtaining the corresponding mine obstacle detection result, the cross-stage local module and the adaptive attention module of the YOLO target detection model are integrated to obtain the initial mine obstacle detection model; the sample acquisition module can be used for obtaining the corresponding sample set based on the historical mine obstacle image, and the sample set can at least include the training set; the training module can be used for training the initial mine obstacle detection model based on the training set to obtain the target mine obstacle detection model; the first determination module can be used to determine whether the target mine obstacle detection model converges.

[0109] In some implementations, the input module 320 may include a first input unit.

[0110] The first input unit may be used to input the mine road image into the target mine obstacle detection model to obtain a corresponding mine obstacle detection result when it is determined that the target mine obstacle detection model has converged.

[0111] In some embodiments, the mine obstacle detection device 300 may further include an enhanced processing module.

[0112] The enhancement processing module can be used in the training module to train the initial mine obstacle detection model according to the training set. Before obtaining the target mine obstacle detection model, the training set is subjected to copy-and-paste data enhancement processing to obtain the corresponding data enhancement training set.

[0113] In some implementations, the training module may include a training unit.

[0114] The training unit can be used to train the initial mine obstacle detection model according to the data enhancement training set to obtain the target mine obstacle detection model.

[0115] In some embodiments, the mine obstacle detection device 300 may further include a second determination module.

[0116] The second determination module can be used for the training unit to train the initial mine obstacle detection model according to the data enhancement training set, and before obtaining the target mine obstacle detection model, determine the scale perception loss function of the initial mine obstacle detection model.

[0117] In some implementations, a training unit may include a training sub-unit.

[0118] The training subunit can be used to train the initial mine obstacle detection model according to the scale-aware loss function and the data enhancement training set to obtain the target mine obstacle detection model.

[0119] In some implementations, the sample set may further include a test set, and the second determination module may include a second input unit, a calculation unit, a first determination unit, and a second determination unit.

[0120] The second input unit can be used to input the test set into the initial mine obstacle detection model to obtain the corresponding prediction frame; the calculation unit can be used to calculate the overlap between the real frame and the prediction frame corresponding to the test set; the first determination unit can be used to determine the weight of the positioning loss function of the initial mine obstacle detection model according to the overlap; the second determination unit can be used to determine the scale perception loss function of the initial mine obstacle detection model according to the weight and the positioning loss function.

[0121] In some embodiments, the mine obstacle detection device 300 may further include an area acquisition module and a third determination module.

[0122] The area acquisition module can be used for the second input unit to input the test set into the initial mine obstacle detection model, and before obtaining the corresponding prediction frame, obtain the pixel area of ​​the real frame corresponding to the test set; the third determination module can be used to determine whether the historical mine obstacle corresponding to the test set is a small target mine obstacle based on the pixel area.

[0123] In some embodiments, the second input unit may include an input sub-unit.

[0124] The input subunit can be used to input the test set into the initial mine obstacle detection model to obtain the corresponding prediction box when it is determined based on the pixel area that the historical mine obstacle corresponding to the test set is a small target mine obstacle.

[0125] In some implementations, the third determining module may include a third determining unit and a fourth determining unit.

[0126] The third determination unit can be used to determine that the historical mine obstacle corresponding to the test set is a small target mine obstacle when the pixel area is less than a preset pixel area threshold; the fourth determination unit can be used to determine that the historical mine obstacle corresponding to the test set is a non-small target mine obstacle when the pixel area is greater than or equal to the preset pixel area threshold.

[0127] In some embodiments, the mine obstacle detection device 300 may further include a fourth determination module.

[0128] The fourth determination module can be used to determine that the positioning loss function is the scale perception loss function of the initial mine obstacle detection model when the historical mine obstacle corresponding to the test set is determined to be a non-small target mine obstacle based on the pixel area.

[0129] In some embodiments, the adaptive attention module may include a convolution submodule, a sigmoid activation function submodule, a normalization submodule, a reshaping broadcast submodule, a feature weighting submodule, a fusion submodule, and a rectified linear unit activation function submodule connected in sequence.

[0130] The convolution submodule can be used to extract features from the mine road image to obtain a feature map; the Sigmoid activation function submodule can be used to generate a corresponding spatial attention map based on the feature map, and the attention value range of the spatial attention map can be [0, 1]; the normalization submodule can be used to normalize the spatial attention map to obtain a normalized attention map.

[0131] The reshaping broadcast submodule can be used to reshape the dimension of the attention weight of the normalized attention map to the same dimension as the feature map, obtain the reshaped attention map, and broadcast the reshaped attention map.

[0132] The feature weighting submodule can be used to perform weighted processing on the feature map and the reshaped attention map to obtain a weighted feature map; the fusion submodule can be used to fuse the channels of the weighted feature map to obtain a fused feature map; the rectified linear unit activation function submodule can be used to perform rectified linear unit activation function processing on the fused feature map.

[0133] The solution provided in this embodiment obtains a corresponding mine obstacle detection result by acquiring a mine road image and inputting the mine road image into a target mine obstacle detection model. The target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module. The adaptive attention module is used to extract the position features of small target mine obstacles in the mine road image. The mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains mine obstacle information. The target mine obstacle detection model is constructed according to the YOLO target detection model and the adaptive attention module, which is beneficial to improving the characterization ability of the target mine obstacle detection model. The target mine obstacle detection model is used to perform mine obstacle detection on the mine road image, which improves the learning ability of the target mine obstacle detection model to learn the position features of small target mine obstacles, making small target mine obstacles easier to detect and improving the detection accuracy of mine obstacles.

[0134] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. Any processing method described in the method embodiment can be implemented by the corresponding processing module in the device embodiment, and will not be repeated one by one in the device embodiment.

[0135] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0136] See also Figure 6 , which shows a functional block diagram of a mine cart 400 provided in an embodiment of the present application, the mine cart 400 may include one or more of the following components: a memory 410, a processor 420, and one or more applications, wherein the one or more applications may be stored in the memory 410 and configured to be executed by the one or more processors 420, and the one or more applications are configured to execute the method described in the aforementioned method embodiment.

[0137] The memory 410 may include a random access memory (RAM) or a read-only memory (Read-Only Memory). The memory 410 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 410 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system and instructions for implementing at least one function (such as obtaining a mine road image, inputting a mine road image, obtaining a mine obstacle detection result, extracting location features, fusing cross-stage local modules and an adaptive attention module, obtaining an initial mine obstacle detection model, obtaining a sample set, training an initial mine obstacle detection model, obtaining a target mine obstacle detection model, determining whether the target mine obstacle detection model converges, determining whether the target mine obstacle detection model converges, copying and pasting data enhancement processing, and obtaining a data enhancement training set). , determine the scale-aware loss function, input the test set, obtain the prediction box, calculate the overlap, determine the weight, obtain the pixel area, determine whether the target mine obstacle is a small target mine obstacle, determine whether the target mine obstacle is a small target mine obstacle, determine whether the target mine obstacle is a non-small target mine obstacle, extract features, obtain feature maps, generate spatial attention maps, normalize operations, obtain normalized attention maps, reshape dimensions, obtain reshaped attention maps, broadcast reshaped attention maps, weighted processing, obtain weighted feature maps, fuse channels, obtain fused feature maps, and corrected linear unit activation function processing, etc.), instructions for implementing the following various method embodiments, etc. The storage data area can also store data created by the mine car 400 during use (such as mine road images, target mine obstacle detection models, mine obstacle detection results, YOLO target detection models, adaptive attention modules, small target mine obstacles, location features, mine obstacle information, first detection results, second detection results, cross-stage local modules, initial mine obstacle detection models, historical mine obstacle images, sample sets, training sets, data enhancement training sets, scale-aware loss functions, target mine obstacle information, test sets, prediction boxes, true boxes, overlap, positioning loss functions, weights, pixel areas, preset pixel area thresholds, non-small target mine obstacles, convolution submodules, Sigmoid activation function submodules, normalization submodules, reshaping broadcast submodules, feature weighting submodules, fusion submodules, rectified linear unit activation function submodules, feature maps, spatial attention maps, attention value ranges, normalized attention maps, attention weights, dimensions, reshaping attention maps, weighted feature maps, channels, and fused feature maps), etc.

[0138] The processor 420 may include one or more processing cores. The processor 420 uses various interfaces and lines to connect various parts of the entire mining car 400, and executes various functions and processes data of the mining car 400 by running or executing instructions, programs, code sets or instruction sets stored in the memory 410, and calling data stored in the memory 410. Optionally, the processor 420 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 420 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU), and a modem. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing display content; and the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 420, but may be implemented separately through a communication chip.

[0139] Please refer to Figure 7 , which shows a structural block diagram of a computer-readable storage medium provided in an embodiment of the present application. The computer-readable storage medium 500 stores a program code 510, which can be called by a processor to execute the method described in the above method embodiment.

[0140] The computer readable storage medium 500 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium 500 includes a non-transitory computer-readable storage medium. The computer readable storage medium 500 has storage space for program code 510 that performs any method steps in the above method. These program codes can be read from or written to one or more computer program products. The program code 510 can be compressed, for example, in an appropriate form.

[0141] Please refer to Figure 8, which shows a structural block diagram of a computer program product 600 provided in an embodiment of the present application. The computer program product 600 includes a computer program / instruction 610, and the computer program / instruction 610 is stored in a computer-readable storage medium of a computer device. When the computer program product 600 is run on a computer device, a processor of the computer device reads the computer program / instruction 610 from the computer-readable storage medium, and the processor executes the computer program / instruction 610, so that the computer device performs the method described in the above method embodiment.

[0142] The solution provided in this embodiment obtains a corresponding mine obstacle detection result by acquiring a mine road image and inputting the mine road image into a target mine obstacle detection model. The target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module. The adaptive attention module is used to extract the position features of small target mine obstacles in the mine road image. The mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains mine obstacle information. The target mine obstacle detection model is constructed according to the YOLO target detection model and the adaptive attention module, which is beneficial to improving the characterization ability of the target mine obstacle detection model. The target mine obstacle detection model is used to perform mine obstacle detection on the mine road image, which improves the learning ability of the target mine obstacle detection model to learn the position features of small target mine obstacles, making small target mine obstacles easier to detect and improving the detection accuracy of mine obstacles.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A mine obstacle detection method, characterized in that: include: Acquire mine road images; The mine road image is input into a target mine obstacle detection model to obtain a corresponding mine obstacle detection result, wherein the target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module, wherein the adaptive attention module is used to extract position features of small target mine obstacles in the mine road image, and the mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains the mine obstacle information.

2. The mine obstacle detection method according to claim 1, characterized in that: Before inputting the mine road image into the target mine obstacle detection model to obtain the corresponding mine obstacle detection result, the mine obstacle detection method further includes: The cross-stage local module and adaptive attention module of the YOLO target detection model are integrated to obtain the initial mine obstacle detection model; According to the historical mine obstacle image, a corresponding sample set is obtained, wherein the sample set at least includes a training set; Training the initial mine obstacle detection model according to the training set to obtain a target mine obstacle detection model; Determining whether the target mine obstacle detection model converges; The inputting the mine road image into the target mine obstacle detection model to obtain the corresponding mine obstacle detection result includes: When it is determined that the target mine obstacle detection model converges, the mine road image is input into the target mine obstacle detection model to obtain a corresponding mine obstacle detection result.

3. The mine obstacle detection method according to claim 2, characterized in that: Before the initial mine obstacle detection model is trained according to the training set to obtain the target mine obstacle detection model, the mine obstacle detection method further includes: Performing copy-and-paste data enhancement processing on the training set to obtain a corresponding data enhanced training set; The training of the initial mine obstacle detection model according to the training set to obtain a target mine obstacle detection model includes: The initial mine obstacle detection model is trained according to the data enhancement training set to obtain a target mine obstacle detection model.

4. The mine obstacle detection method according to claim 3, characterized in that: Before the initial mine obstacle detection model is trained according to the data enhancement training set to obtain the target mine obstacle detection model, the mine obstacle detection method further includes: Determining a scale-aware loss function of the initial mine obstacle detection model; The initial mine obstacle detection model is trained according to the data enhancement training set to obtain a target mine obstacle detection model, including: The initial mine obstacle detection model is trained according to the scale-aware loss function and the data enhancement training set to obtain a target mine obstacle detection model.

5. The mine obstacle detection method according to claim 4, characterized in that: The sample set also includes a test set, and the determining of the scale-aware loss function of the initial mine obstacle detection model includes: Inputting the test set into the initial mine obstacle detection model to obtain a corresponding prediction box; Calculate the overlap between the real frame and the predicted frame corresponding to the test set; Determining a weight of a positioning loss function of the initial mine obstacle detection model according to the overlap; A scale perception loss function of the initial mine obstacle detection model is determined according to the weight and the positioning loss function.

6. The mine obstacle detection method according to claim 5, characterized in that: Before inputting the test set into the initial mine obstacle detection model to obtain the corresponding prediction frame, the mine obstacle detection method further includes: Obtaining the pixel area of ​​the real frame corresponding to the test set; Determining whether the historical mine obstacle corresponding to the test set is the small target mine obstacle according to the pixel area; The step of inputting the test set into the initial mine obstacle detection model to obtain a corresponding prediction frame includes: When it is determined according to the pixel area that the historical mine obstacle corresponding to the test set is the small target mine obstacle, the test set is input into the initial mine obstacle detection model to obtain a corresponding prediction box.

7. The mine obstacle detection method according to claim 6, characterized in that: The determining, according to the pixel area, whether the historical mine obstacle corresponding to the test set is the small target mine obstacle includes: When the pixel area is smaller than a preset pixel area threshold, determining that the historical mine obstacle corresponding to the test set is the small target mine obstacle; When the pixel area is greater than or equal to a preset pixel area threshold, it is determined that the historical mine obstacle corresponding to the test set is a non-small target mine obstacle.

8. The mine obstacle detection method according to claim 6, characterized in that: Also includes: When it is determined according to the pixel area that the historical mine obstacle corresponding to the test set is a non-small target mine obstacle, the positioning loss function is determined to be the scale perception loss function of the initial mine obstacle detection model.

9. The mine obstacle detection method according to any one of claims 1 to 8, characterized in that: The adaptive attention module includes a convolution submodule, a Sigmoid activation function submodule, a normalization submodule, a reshaping broadcast submodule, a feature weighting submodule, a fusion submodule and a rectified linear unit activation function submodule connected in sequence; The convolution submodule is used to extract features from the mine road image to obtain a feature map; the Sigmoid activation function submodule is used to generate a corresponding spatial attention map according to the feature map, and the attention value range of the spatial attention map is [0, 1]; The normalization submodule is used to perform a normalization operation on the spatial attention map to obtain a normalized attention map; The reshaping broadcast submodule is used to reshape the dimension of the attention weight of the normalized attention map to be the same as the dimension of the feature map, obtain a reshaped attention map, and broadcast the reshaped attention map; The feature weighting submodule is used to perform weighted processing on the feature map and the reshaped attention map to obtain a weighted feature map; the fusion submodule is used to fuse the channels of the weighted feature map to obtain a fused feature map; the corrected linear unit activation function submodule is used to perform corrected linear unit activation function processing on the fused feature map.

10. A mine obstacle detection device, characterized in that: include: An image acquisition module, used for acquiring mine road images; An input module is used to input the mine road image into a target mine obstacle detection model to obtain a corresponding mine obstacle detection result, wherein the target mine obstacle detection model is obtained based on a YOLO target detection model and an adaptive attention module, wherein the adaptive attention module is used to extract position features of small target mine obstacles in the mine road image, and the mine obstacle detection result includes a first detection result for characterizing that the mine road image does not contain mine obstacle information, and a second detection result for characterizing that the mine road image contains the mine obstacle information.

11. A mining car, characterized in that: include: Memory; One or more processors coupled to the memory; One or more applications, wherein the one or more applications are stored in the memory and are configured to be executed by one or more processors, and the one or more applications are configured to execute the mine obstacle detection method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, and the program codes can be called by a processor to execute the mine obstacle detection method according to any one of claims 1 to 9.

Images