Method for detecting cleaning state of bottom of mine unloading electric locomotive
An automated image-based method using a trained mineral recognition model improves the efficiency of ore car residue detection by reducing the reliance on manual inspection.
Patent Information
- Application Number
- CN202510365805.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-26
- Publication Date
- 2025-07-15
AI Technical Summary
In the prior art, the cleaning state detection efficiency of the bottom of the unloading motor vehicle box is low, and requires manual inspection, which is not very efficient.
The target mineral recognition model is adopted, and the initial image is acquired and iteratively trained, and the ore features are extracted using the encoder and decoder to generate mineral residue recognition results to achieve automatic detection.
The efficiency of cleaning status detection of the bottom of the unloading motor vehicle box is improved, manpower is saved, and detection efficiency is improved.
Smart Images
Figure CN120318751A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of ore-dumping locomotives, and particularly to a method for detecting the cleaning state of the bottom of an ore-dumping locomotive box. Background Art
[0002] In the process of mine production, locomotives are generally used for ore transportation. Since the ore contains moisture, some slag will adhere to the bottom of the ore car, and it is necessary to clean the slag at the bottom of the ore car. However, at present, it is necessary to manually check the cleaning state of the bottom of the ore-dumping locomotive box, that is, to judge whether there is mineral residue, and the efficiency is not high. Summary of the Invention
[0003] This application aims to propose a method for detecting the cleaning state of the bottom of an ore-dumping locomotive box, which can improve the efficiency of the method for checking the cleaning state of the bottom of the box.
[0004] An embodiment of this application provides a method for detecting the cleaning state of the bottom of an ore-dumping locomotive box, including:
[0005] Obtain a first initial image, where the first initial image is an image of the bottom of the ore-dumping locomotive to be detected;
[0006] Input the first initial image into a target mineral recognition model to obtain a first mineral residue recognition result, and the first mineral residue recognition result is used to indicate the cleaning state of the bottom of the ore-dumping locomotive box;
[0007] Among them, the target mineral recognition model is obtained through the following steps:
[0008] Obtain a plurality of second initial images, and a plurality of the second initial images all have mineral identifiers, and the mineral identifiers are used to mark the minerals in the second initial images;
[0009] Based on a loss function, input the plurality of second initial images into an initial mineral recognition model for model iterative training until the loss function value reaches a preset convergence value to obtain the target mineral recognition model.
[0010] According to some embodiments of this application, the initial mineral recognition model includes:
[0011] An encoder, which is used to extract features in the second initial image to obtain a first target feature image;
[0012] A decoder, which is connected to the encoder, and the decoder is used to generate a second target feature image according to the first target feature image, and the second target feature image has the same size as the second initial image;
[0013] A result output module, the result output module is connected to the decoder, and the result output module is used to output a second mineral residue recognition result according to the second target feature image.
[0014] According to some embodiments of the present application, the encoder includes:
[0015] A first downsampling module, the first downsampling module is used to perform a downsampling operation on the second initial image to obtain a first intermediate feature image;
[0016] A second downsampling module, the second downsampling module is connected to the first downsampling module, and the second downsampling module is used to perform a downsampling operation on the first intermediate feature image to obtain a second intermediate feature image;
[0017] A third downsampling module, the third downsampling module is connected to the second downsampling module, and the third downsampling module is used to perform a downsampling operation on the second intermediate feature image to obtain the first target feature image.
[0018] According to some embodiments of the present application, the first downsampling module includes a first convolutional layer, a first activation function, a second convolutional layer, a second activation function, and a first max pooling layer connected in sequence.
[0019] According to some embodiments of the present application, the second downsampling module includes a third convolutional layer, a third activation function, a fourth convolutional layer, a fourth activation function, and a second max pooling layer connected in sequence.
[0020] According to some embodiments of the present application, the third downsampling module includes a fifth convolutional layer, a fifth activation function, a sixth convolutional layer, a sixth activation function, a third max pooling layer, and a seventh convolutional layer connected in sequence.
[0021] According to some embodiments of the present application, the decoder includes:
[0022] A first upsampling module, the first upsampling module is used to perform an upsampling operation on the first target feature image to obtain a fourth intermediate feature image;
[0023] A second upsampling module, the second upsampling module is connected to the first upsampling module, and the second upsampling module is used to perform an upsampling operation on the fourth intermediate feature image to obtain a fifth intermediate feature image;
[0024] A third upsampling module, the third upsampling module is connected to the second upsampling module, and the third upsampling module is used to perform an upsampling operation on the fifth intermediate feature image to obtain the second target feature image.
[0025] According to some embodiments of the present application, the first upsampling module includes a first transposed convolution layer, an eighth convolution layer, a seventh activation function, a ninth convolution layer, and an eighth activation function that are connected in sequence.
[0026] According to some embodiments of the present application, the second upsampling module includes a second transposed convolution layer, a tenth convolution layer, a ninth activation function, an eleventh convolution layer, and a tenth activation function that are connected in sequence.
[0027] According to some embodiments of the present application,
[0028] The third upsampling module includes a third transposed convolution layer, a twelfth convolution layer, an eleventh activation function, a thirteenth convolution layer, and a twelfth activation function that are connected in sequence.
[0029] In the embodiments of the present application, by obtaining a first initial image and inputting the first initial image into the target mineral recognition model, since the target mineral recognition model can extract the ore features in the first initial image, it is possible to confirm whether the minerals at the bottom of the ore unloading locomotive box to be detected have been cleaned, and obtain a first mineral residue recognition result, saving manpower and improving the detection efficiency.
[0030] The additional aspects and advantages of the present application will be partly given in the following description, partly will become obvious from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] The following further describes the present application with reference to the drawings and embodiments, where:
[0032] Figure 1 is a flowchart of an embodiment of the method for detecting the cleaning state of the bottom of the ore unloading locomotive box provided by the present application;
[0033] Figure 2 is a schematic diagram of the overall framework of the initial mineral recognition model in the embodiment of the method for detecting the cleaning state of the bottom of the ore unloading locomotive box provided by the present application;
[0034] Figure 3 is a schematic diagram of an embodiment of the electronic device provided by the present application.
[0035] Reference Numerals:
[0036] Electronic device 100, processor 110, memory 120. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0037] The embodiments of the present application are described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the drawings are exemplary only for explaining the present application and should not be construed as limiting the present application.
[0038] In the description of the present application, it should be understood that when it comes to orientation descriptions, such as up, down, etc., the orientation or positional relationship indicated is based on the orientation or positional relationship shown in the drawings. This is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present application.
[0039] In the description of the present application, "a plurality" refers to more than two. If there is a description of first and second, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features or implicitly specifying the sequence relationship of the indicated technical features.
[0040] In the description of the present application, unless otherwise clearly defined, terms such as "set", "installed", "connected", etc. should be understood in a broad sense. Those skilled in the art can reasonably determine the specific meanings of the above terms in the present application in combination with the specific content of the technical solution.
[0041] The embodiment of the present application provides a method for detecting the cleaning state of the bottom of an ore unloading electric locomotive carriage, as Figure 1 shown, including:
[0042] Step S100: Obtain a first initial image, where the first initial image is an image of the bottom of the ore unloading electric locomotive to be detected;
[0043] Step S200: Input the first initial image into the target mineral recognition model to obtain a first mineral residue recognition result, where the first mineral residue recognition result is used to indicate the cleaning state of the bottom of the ore unloading electric locomotive carriage;
[0044] Among them, the target mineral recognition model is obtained through the following steps:
[0045] Step S300: Obtain a plurality of second initial images, and each of the plurality of second initial images has a mineral identifier, where the mineral identifier is used to mark the minerals in the second initial image;
[0046] Step S400: Based on the loss function, input the plurality of second initial images into the initial mineral recognition model for model iterative training until the loss function value reaches a preset convergence value to obtain the target mineral recognition model.
[0047] In the embodiment of the present application, by obtaining the first initial image and inputting the first initial image into the target mineral recognition model, since the target mineral recognition model can extract the ore features in the first initial image, it is possible to confirm whether the minerals at the bottom of the ore unloading electric locomotive to be detected have been cleaned, obtain the first mineral residue recognition result, save manpower, and improve the detection efficiency.
[0048] In the above step S100, by installing a camera in the working area, performing spatial calibration on the camera, and using the fixed area where the carriage of the ore-unloading locomotive appears in the image as the target detection area, the first initial image is obtained through the camera.
[0049] In the above step S300, the second initial image is obtained by performing mineral annotation on the image of the carriage of the ore-unloading locomotive. For example, using the labelme annotation tool to mark the mineral pixels in the carriage image to obtain the second initial image.
[0050] In the above step S400, the loss function can adopt the cross-entropy loss function to calculate the difference between the prediction result and the true label. The formula of the cross-entropy loss function is as follows:
[0051] L = -∑(ylog(p) + (1 - y)log(1 - p)),
[0052] where y is the true label and p is the predicted probability of the initial mineral recognition model.
[0053] Use an optimizer, such as the Adam optimizer, to update the weights according to the gradients calculated by the cross-entropy loss function. The Adam optimizer combines the advantages of the momentum and RMSProp algorithms, can automatically adjust the learning rate, and improve the training efficiency.
[0054] The formula for weight update is:
[0055]
[0056] where θ is the weight, η is the learning rate, and are the first-order moment estimate and the second-order moment estimate respectively, and ∈ is a small constant used to prevent division by zero. Calculate the gradient of the loss function with respect to each weight of the initial mineral recognition model through backpropagation.
[0057] In some embodiments of the present application, the initial mineral recognition model includes:
[0058] An encoder, which is used to extract features from the second initial image to obtain the first target feature image;
[0059] A decoder, which is connected to the encoder. The decoder is used to generate a second target feature image according to the first target feature image, and the size of the second target feature image is the same as that of the second initial image;
[0060] A result output module, which is connected to the decoder. The result output module is used to output the second mineral residue recognition result according to the second target feature image.
[0061] In this embodiment, the encoder performs a downsampling operation on the second initial image to extract the mineral features in the second initial image, obtaining the first target feature image. The decoder performs an upsampling operation on the first target feature image to restore the first target feature image to the same size as the second initial image, obtaining the second target feature image. Finally, the result output module outputs the second mineral residue recognition result according to the second target feature image.
[0062] The result output module uses a 1x1 convolutional layer to convolve the second target feature image into an image with 2 channels (binary image), obtaining the second mineral residue recognition result. In the second mineral residue recognition result, the pixels with a value of 1 are the positions of the minerals in the image pixels.
[0063] In some embodiments of the present application, the second initial image is first subjected to 1x1 convolution to change the number of channels to 1, and then input into the encoder.
[0064] In some embodiments of the present application, as Figure 2 shown, the encoder includes:
[0065] The first downsampling module is used to perform a downsampling operation on the second initial image to obtain the first intermediate feature image;
[0066] The second downsampling module is connected to the first downsampling module and is used to perform a downsampling operation on the first intermediate feature image to obtain the second intermediate feature image;
[0067] The third downsampling module is connected to the second downsampling module and is used to perform a downsampling operation on the second intermediate feature image to obtain the first target feature image.
[0068] In this embodiment, the first downsampling module, the second downsampling module, and the third downsampling module of the encoder are used to perform three downsampling operations on the second initial image in sequence, so as to extract the mineral features in the second initial image and obtain the first target feature image.
[0069] In some embodiments of the present application, the first downsampling module includes a first convolutional layer, a first activation function, a second convolutional layer, a second activation function, and a first max pooling layer connected in sequence.
[0070] In this embodiment, the convolutional kernels of the first convolutional layer and the second convolutional layer are 3×3 in size, with a stride of 1, a padding of 0, and an output channel number of 64. By performing two 3x3 convolutions to extract the mineral features in the second initial image, an image with a larger receptive field can be obtained, and the information of the image is more comprehensive. Then, the first max pooling layer is used to strengthen the features with larger weights and remove unimportant features through downsampling to obtain the first intermediate feature image. The pooling kernel of the first max pooling layer is 2×2 in size, with a stride of 2. The first convolutional layer and the second convolutional layer can also use convolutional kernels of 5x5 size.
[0071] In some embodiments of the present application, the second downsampling module includes a third convolutional layer, a third activation function, a fourth convolutional layer, a fourth activation function, and a second max pooling layer connected in sequence.
[0072] In this embodiment, the convolutional kernels of the third convolutional layer and the fourth convolutional layer are 3×3 in size, with a stride of 1, a padding of 0, and an output channel number of 128. By performing two 3x3 convolutions to extract the mineral features in the first intermediate feature image, an image with a larger receptive field can be obtained, and the information of the image is more comprehensive. Then, the second max pooling layer is used to strengthen the features with larger weights and remove unimportant features through downsampling to obtain the second intermediate feature image. The pooling kernel of the second max pooling layer is 2×2 in size, with a stride of 2.
[0073] In some embodiments of the present application, the third downsampling module includes a fifth convolutional layer, a fifth activation function, a sixth convolutional layer, a sixth activation function, a third max pooling layer, and a seventh convolutional layer connected in sequence.
[0074] In this embodiment, the convolutional kernels of the fifth convolutional layer and the sixth convolutional layer are 3×3 in size, with a stride of 1, a padding of 0, and an output channel number of 256. By performing two 3x3 convolutions to extract the mineral features in the second intermediate feature image, an image with a larger receptive field can be obtained, and the information of the image is more comprehensive. Then, the third max pooling layer is used to strengthen the features with larger weights and remove unimportant features through downsampling, and through the seventh convolutional layer, the first target feature image is obtained. The pooling kernel of the third max pooling layer is 2×2 in size, with a stride of 2. The convolutional kernel of the seventh convolutional layer is 3×3 in size, with a stride of 1, a padding of 0, and an output channel number of 1024.
[0075] In some embodiments of the present application, as Figure 2 shown, the decoder includes:
[0076] A first upsampling module, which is used to perform an upsampling operation on the first target feature image to obtain a fourth intermediate feature image;
[0077] The second upsampling module is connected to the first upsampling module. The second upsampling module is used to perform an upsampling operation on the fourth intermediate feature image to obtain a fifth intermediate feature image;
[0078] The third upsampling module is connected to the second upsampling module. The third upsampling module is used to perform an upsampling operation on the fifth intermediate feature image to obtain a second target feature image.
[0079] In this embodiment, by using the first upsampling module, the second upsampling module, and the third upsampling module of the decoder, an upsampling operation is sequentially performed on the first target feature image to restore the first target feature image to the same size as the second initial image, thereby obtaining a second target feature image.
[0080] In some embodiments of the present application, the first upsampling module includes a first transposed convolution layer, an eighth convolution layer, a seventh activation function, a ninth convolution layer, and an eighth activation function that are sequentially connected.
[0081] In some embodiments of the present application, the convolution kernel size of the first transposed convolution layer is 2×2, the stride is 2, the padding is 0, and the number of output channels is 512. The convolution kernel sizes of the eighth convolution layer and the ninth convolution layer are 3×3, the stride is 1, the padding is 0, and the number of output channels is 512. The first target feature image is restored by using the first transposed convolution layer. The used convolution kernel size is 2x2 and the stride is 2, which can enlarge the first target feature image. Moreover, the smaller convolution and stride can reduce the generation of 0 pixels and increase useless information. The upsampling operation inevitably generates 0 pixel values. Therefore, the eighth convolution layer and the ninth convolution layer are used again to reduce the influence of the 0 pixel positions, thereby obtaining a fourth intermediate feature image.
[0082] In some embodiments of the present application, the second upsampling module includes a second transposed convolution layer, a tenth convolution layer, a ninth activation function, an eleventh convolution layer, and a tenth activation function that are sequentially connected.
[0083] In some embodiments of the present application, the convolution kernel size of the second transposed convolution layer is 2×2, the stride is 2, the padding is 0, and the number of output channels is 256. The convolution kernel sizes of the tenth convolution layer and the eleventh convolution layer are 3×3, the stride is 1, the padding is 0, and the number of output channels is 256. The fourth intermediate feature image is restored by using the second transposed convolution layer. The used convolution kernel size is 2x2 and the stride is 2, which can enlarge the fourth intermediate feature image. Moreover, the smaller convolution and stride can reduce the generation of 0 pixels and increase useless information. The upsampling operation inevitably generates 0 pixel values. Therefore, the tenth convolution layer and the eleventh convolution layer are used again to reduce the influence of the 0 pixel positions, thereby obtaining a fifth intermediate feature image.
[0084] In some embodiments of the present application, the third upsampling module includes a third transposed convolution layer, a twelfth convolution layer, an eleventh activation function, a thirteenth convolution layer, and a twelfth activation function connected in sequence.
[0085] In some embodiments of the present application, the convolution kernel size of the third transposed convolution layer is 2×2, the stride is 2, the padding is 0, and the number of output channels is 128. The convolution kernel sizes of the twelfth convolution layer and the thirteenth convolution layer are 3×3, the stride is 1, the padding is 0, and the number of output channels is 128. The fifth intermediate feature image is restored using the third transposed convolution layer. The convolution kernel size used is 2x2 and the stride is 2, which can enlarge the fifth intermediate feature image. Moreover, the smaller convolution and stride can reduce the generation of 0 pixels and increase useless information. The upsampling operation inevitably generates 0 pixel values. Therefore, the twelfth convolution layer and the thirteenth convolution layer are used again to reduce the influence of the 0 pixel positions, and the second target feature image is obtained.
[0086] In addition, an embodiment of the present application provides an electronic device 100, as Figure 3 shown, including:
[0087] At least one processor 110;
[0088] At least one memory 120 for storing at least one program;
[0089] When the at least one program is executed by the at least one processor 110, the ore unloading locomotive bottom cleaning state detection method as described above is implemented.
[0090] The electronic device 100 provided by the embodiment of the present application can implement each process implemented by the above method embodiment and achieve the same beneficial effects. To avoid repetition, it will not be elaborated here.
[0091] In addition, an embodiment of the present application provides a computer-readable storage medium, in which a processor-executable program is stored. When the processor-executable program is executed by a processor, it is used to implement the ore unloading locomotive bottom cleaning state detection method as described above.
[0092] The computer-readable storage medium provided by the embodiment of the present application can implement each process implemented by the above method embodiment and achieve the same beneficial effects. To avoid repetition, it will not be elaborated here.
[0093] Those of ordinary skill in the art will understand that all or some of the steps and systems disclosed above can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVDs) or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.
[0094] The embodiments of the present application have been described in detail above with reference to the accompanying drawings. However, the present application is not limited to the above embodiments, and various changes can be made without departing from the spirit of the present application within the scope of knowledge possessed by those of ordinary skill in the relevant art.
Claims
1. A method for detecting the cleaning state of the bottom of an ore-discharging electric locomotive, characterized in that, Including: Obtain a first initial image, where the first initial image is the bottom image of the ore-discharging locomotive to be detected; Input the first initial image into a target mineral recognition model to obtain a first mineral residue recognition result, where the first mineral residue recognition result is used to indicate the cleaning state of the bottom of the ore-discharging locomotive; Among them, the target mineral recognition model is obtained through the following steps: Obtain a plurality of second initial images, and a plurality of the second initial images all have mineral identifiers, where the mineral identifiers are used to mark the minerals in the second initial images; Based on a loss function, input the plurality of second initial images into an initial mineral recognition model for model iterative training until the loss function value reaches a preset convergence value to obtain the target mineral recognition model.
2. The method for detecting the cleaning state of the bottom of the ore unloading electric locomotive box according to claim 1, characterized in that, The initial mineral recognition model includes: An encoder, where the encoder is used to extract features in the second initial image to obtain a first target feature image; A decoder, where the decoder is connected to the encoder, and the decoder is used to generate a second target feature image according to the first target feature image, and the second target feature image has the same size as the second initial image; A result output module, where the result output module is connected to the decoder, and the result output module is used to output a second mineral residue recognition result according to the second target feature image.
3. The method for detecting the cleaning state of the bottom of the ore-discharging electric locomotive according to claim 2, wherein The encoder includes: A first downsampling module, where the first downsampling module is used to perform a downsampling operation on the second initial image to obtain a first intermediate feature image; A second downsampling module, where the second downsampling module is connected to the first downsampling module, and the second downsampling module is used to perform a downsampling operation on the first intermediate feature image to obtain a second intermediate feature image; A third downsampling module, where the third downsampling module is connected to the second downsampling module, and the third downsampling module is used to perform a downsampling operation on the second intermediate feature image to obtain the first target feature image.
4. The method for detecting the cleaning state of the bottom of the ore-discharging locomotive according to claim 3, wherein: The first downsampling module includes a first convolutional layer, a first activation function, a second convolutional layer, a second activation function, and a first max-pooling layer connected in sequence.
5. The method for detecting the cleaning state of the bottom of the ore-discharging locomotive according to claim 3, wherein: The second downsampling module includes a third convolutional layer, a third activation function, a fourth convolutional layer, a fourth activation function, and a second max-pooling layer connected in sequence.
6. The method for detecting the cleaning state of the bottom of the ore-discharging locomotive according to claim 3, wherein: The third downsampling module includes a fifth convolutional layer, a fifth activation function, a sixth convolutional layer, a sixth activation function, a third max-pooling layer, and a seventh convolutional layer connected in sequence.
7. The method for detecting the cleaning state of the bottom of the ore unloading electric locomotive box according to claim 2, characterized in that, The decoder includes: A first upsampling module, where the first upsampling module is used to perform an upsampling operation on the first target feature image to obtain a fourth intermediate feature image; A second upsampling module, the second upsampling module is connected to the first upsampling module, and the second upsampling module is configured to perform an upsampling operation on the fourth intermediate feature image to obtain a fifth intermediate feature image; A third upsampling module, the third upsampling module is connected to the second upsampling module, and the third upsampling module is configured to perform an upsampling operation on the fifth intermediate feature image to obtain the second target feature image.
8. The method for detecting the cleaning state of the bottom of the ore unloading locomotive box according to claim 7, wherein: The first upsampling module includes a first transposed convolution layer, an eighth convolution layer, a seventh activation function, a ninth convolution layer, and an eighth activation function connected in sequence.
9. The method for detecting the cleaning state of the bottom of the ore unloading locomotive box according to claim 7, wherein: The second upsampling module includes a second transposed convolution layer, a tenth convolution layer, a ninth activation function, an eleventh convolution layer, and a tenth activation function connected in sequence.
10. The method for detecting the cleaning state of the bottom of the ore unloading locomotive box according to claim 9, wherein: The third upsampling module includes a third transposed convolution layer, a twelfth convolution layer, an eleventh activation function, a thirteenth convolution layer, and a twelfth activation function connected in sequence.