Bucket wheel machine intelligent control system and method based on digital image monitoring

Abstract

The embodiment of the invention relates to the technical field of bucket wheel machine control, and provides a bucket wheel machine intelligent control system and method based on digital image monitoring, and the method comprises the steps: carrying out the image contrast normalization processing of an obtained bucket wheel machine working region image, and then carrying out the image analysis and working region state feature extraction, so as to obtain a working region state feature map; inputting the working area state feature map into a material target object feature highlighting device based on a dual optimization attention module to obtain a material target object highlighting working area state feature map; a working area state feature map is highlighted based on a material target object, a bucket wheel machine working area material semantic segmentation result is generated, the material target object is tracked based on the bucket wheel machine working area material semantic segmentation result, the advancing angle of a bucket wheel machine is controlled and adjusted, and therefore more efficient and accurate material carrying and stacking operation is achieved. And the intelligent level of the bucket wheel machine control system is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of bucket wheel machine control, and in particular to an intelligent control system and method for a bucket wheel machine based on digital image monitoring. Background Art

[0002] Bucket wheel excavator is a heavy mechanical equipment used for stacking bulk materials (such as coal, ore, etc.), and is often used in ports, mines, etc. When the bucket wheel excavator is carrying and stacking materials, it is usually necessary to accurately identify and track the location, movement trajectory and state of the materials, so as to ensure that the actual location of the materials matches the movement state of the bucket wheel excavator, so as to achieve accurate handling and stacking operations.

[0003] Traditional bucket wheel control systems usually adjust the position and posture of bucket wheel machines based on preset programs and manual intervention by operators to achieve material handling and stacking operations. However, this method is easily affected by the operator's skill level and subjective factors, resulting in unstable operation accuracy and efficiency. In addition, this method based on preset programs and manual control usually cannot accurately identify and track the location of materials, making it difficult to accurately track materials, which can lead to errors, waste, or unstable stacking during the handling process. At the same time, the traditional bucket wheel control system has limited ability to track materials, increasing the risk of collisions or accidents. Summary of the invention

[0004] The present disclosure aims to solve at least one of the problems existing in the prior art and provide an intelligent control system and method for a bucket wheel excavator based on digital image monitoring.

[0005] In one aspect of the present disclosure, there is provided an intelligent control system for a bucket wheel machine based on digital image monitoring, the intelligent control system for the bucket wheel machine comprising:

[0006] A bucket wheel machine working area image acquisition module, used to acquire an image of the bucket wheel machine working area acquired by a camera deployed on the bucket wheel machine;

[0007] A working area image contrast adjustment module, used for performing image contrast normalization processing on the bucket wheel machine working area image to obtain a contrast normalized bucket wheel machine working area image;

[0008] A working area state feature extraction module, used for performing image analysis and working area state feature extraction on the contrast normalized bucket wheel machine working area image to obtain a working area state feature map;

[0009] A material target object highlighting work area state characterization module, used for inputting the work area state feature map into a material target object feature highlighter based on a dual optimization attention module to obtain a material target object highlighting work area state feature map;

[0010] A working area material semantic segmentation module is used to generate a semantic segmentation result of the bucket wheel machine working area material based on the material target object highlighting the working area state feature map;

[0011] The bucket wheel machine travel angle control module is used to track the material target object based on the semantic segmentation result of the material in the working area of ​​the bucket wheel machine, and control and adjust the travel angle of the bucket wheel machine.

[0012] Optionally, the working area image contrast adjustment module includes:

[0013] A pixel mean and standard deviation calculation unit, used to calculate the global mean and standard deviation of the pixel values ​​in the bucket wheel machine working area image, so as to obtain the global mean of the bucket wheel machine working area image pixels and the standard deviation of the bucket wheel machine working area image pixels;

[0014] An image contrast normalization unit is used to subtract the global mean value of the bucket wheel machine working area image pixels from each pixel value in the bucket wheel machine working area image to obtain a bucket wheel machine working area pixel difference image, and then divide the bucket wheel machine working area pixel difference image pixel by pixel standard deviation of the bucket wheel machine working area image to obtain the contrast normalized bucket wheel machine working area image.

[0015] Optionally, the working area state feature extraction module is specifically used to: input the contrast-normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on a ResNet model to obtain the working area state feature map.

[0016] Optionally, the material target object highlights the working area state characterization module, including:

[0017] A working area state channel feature manifestation unit, used for inputting the working area state feature map into the channel attention optimization network in the material target object feature highlighter based on the dual optimization attention module to obtain a working area state channel saliency feature map;

[0018] A working area state space feature manifestation unit, used for inputting the working area state feature map into the spatial attention optimization network in the material target object feature highlighter based on the dual optimization attention module to obtain a working area state space saliency feature map;

[0019] The working area material target object feature highlighting unit is used to add and fuse the working area state channel salient feature map and the working area state space salient feature map position by position to obtain the material target object highlighted working area state feature map.

[0020] Optionally, the working area state channel feature visualization unit includes:

[0021] A global mean pooling processing subunit, used for performing global mean pooling processing on each feature matrix along the channel dimension in the working area state feature map to obtain a working area state feature vector;

[0022] A first convolution processing subunit is used to input the working area state feature vector into a convolution layer based on a convolution kernel of 1×3 for processing, so as to obtain a working area state channel attention vector;

[0023] A point convolution adjustment subunit, used for performing point convolution on the working area state channel attention vector to adjust the number of channels, so as to obtain a working area state channel adjustment attention vector;

[0024] A first activation processing subunit is used to perform activation processing based on a Sigmoid function on the work area state channel adjustment attention vector to obtain a work area state channel attention weight vector;

[0025] The weighted optimization subunit is used to use the various position feature values ​​in the working area state channel attention weight vector as weighting coefficients to perform weighted optimization on the various feature matrices of the working area state feature map along the channel dimension to obtain the working area state channel saliency feature map.

[0026] Optionally, the working area state space feature visualization unit includes:

[0027] A second convolution processing subunit is used to input the working area state feature map into a convolution layer based on a convolution kernel of 1×1 for processing to obtain a working area state space attention feature map;

[0028] A second activation processing subunit is used to perform activation processing based on a Sigmoid function on the working area state space attention feature map to obtain a working area state space attention weight feature map;

[0029] The position point multiplication subunit is used to perform position point multiplication on the working area state space attention weight feature map and the working area state feature map to obtain the working area state space saliency feature map.

[0030] Optionally, the working area material semantic segmentation module is specifically used to: input the material target object highlighting the working area state feature map into a classifier-based working area material semantic segmentor to obtain the bucket wheel machine working area material semantic segmentation result.

[0031] Optionally, the bucket wheel machine intelligent control system further includes a training module;

[0032] The training module is used to train the material target object feature highlighter based on the dual optimization attention module;

[0033] When the working area state feature extraction module is specifically used to input the contrast-normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on a ResNet model, the training module is further used to: train the bucket wheel machine working area state feature extractor based on a ResNet model;

[0034] When the working area material semantic segmentation mold body is used to input the material target object highlighting the working area state feature map into the classifier-based working area material semantic segmentor, the training module is also used to: train the classifier-based working area material semantic segmentor.

[0035] Optionally, the training module includes:

[0036] A training bucket wheel machine working area image acquisition unit is used to obtain images of the training bucket wheel machine working area acquired by a camera deployed on the bucket wheel machine, and real values ​​corresponding to semantic segmentation results of materials in the training bucket wheel machine working area;

[0037] A training working area image contrast adjustment unit, used for performing image contrast normalization processing on the training bucket wheel machine working area image to obtain a training contrast normalized bucket wheel machine working area image;

[0038] A training working area state feature extraction unit is used to input the training contrast normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on a ResNet model to perform image analysis and work area state feature extraction to obtain a training working area state feature map;

[0039] A training material target object highlighting work area state characterization unit, used for inputting the training work area state feature map into the material target object feature highlighter based on the dual optimization attention module to obtain a training material target object highlighting work area state feature map;

[0040] A training classification unit, used for passing the training material target object highlighting the working area state feature map through the classifier-based working area material semantic segmentor to obtain the training bucket wheel machine working area material semantic segmentation result;

[0041] A classification loss function value calculation unit is used to calculate the cross entropy loss function value between the semantic segmentation result of the material in the working area of ​​the training bucket wheel machine and the corresponding true value to obtain the classification loss function value;

[0042] A training unit is used to train the bucket wheel machine working area state feature extractor based on the ResNet model, the material target object feature highlighter based on the dual optimization attention module, and the working area material semantic segmentor based on the classifier based on the classification loss function value.

[0043] Another aspect of the present disclosure provides a bucket wheel machine intelligent control method based on digital image monitoring, the bucket wheel machine intelligent control method comprising:

[0044] Acquire an image of a working area of ​​a bucket wheel machine collected by a camera deployed on the bucket wheel machine;

[0045] Performing image contrast normalization processing on the bucket wheel machine working area image to obtain a contrast normalized bucket wheel machine working area image;

[0046] Performing image analysis and working area state feature extraction on the contrast-normalized bucket wheel machine working area image to obtain a working area state feature map;

[0047] Inputting the working area state feature map into a material target object feature highlighter based on a dual optimization attention module to obtain a material target object highlighted working area state feature map;

[0048] Based on the material target object highlighting the working area state feature map, generating a semantic segmentation result of the bucket wheel machine working area material;

[0049] Based on the semantic segmentation results of the materials in the working area of ​​the bucket wheel machine, the material target object is tracked, and the travel angle of the bucket wheel machine is controlled and adjusted.

[0050] Compared with the prior art, the present invention performs image contrast normalization on the acquired bucket wheel machine working area image, and then performs image analysis and working area state feature extraction to obtain a working area state feature map; the working area state feature map is input into a material target object feature highlighter based on a dual optimization attention module to obtain a material target object highlighted working area state feature map; based on the material target object highlighted working area state feature map, a bucket wheel machine working area material semantic segmentation result is generated, and based on the bucket wheel machine working area material semantic segmentation result, the material target object is tracked, and the bucket wheel machine travel angle is controlled and adjusted. In this way, the bucket wheel machine travel angle can be dynamically adjusted based on the image semantics of the bucket wheel machine working area monitored in real time, thereby achieving more efficient and accurate material handling and stacking operations, and improving the intelligence level of the bucket wheel machine control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0052] Figure 1 A structural block diagram of an intelligent control system for a bucket wheel machine based on digital image monitoring provided in one embodiment of the present disclosure;

[0053] Figure 2 A flowchart of a bucket wheel machine intelligent control method based on digital image monitoring provided by another embodiment of the present disclosure;

[0054] Figure 3 A structural block diagram of an electronic device provided in another embodiment of the present disclosure;

[0055] Figure 4 This is an application scenario diagram of an intelligent control system for a bucket wheel excavator based on digital image monitoring provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in each embodiment of the present disclosure, many technical details are proposed in order to enable readers to better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed for protection in the present disclosure can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present disclosure. The various embodiments can be combined and referenced with each other without contradiction.

[0057] Unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0058] Although the present disclosure makes various references to certain modules in the system according to embodiments of the present disclosure, any number of different modules can be used and run on a user terminal and / or server. The modules are illustrative only, and different aspects of the systems and methods can use different modules.

[0059] The present disclosure uses a flow chart to illustrate the operations performed according to the method of the embodiments of the present disclosure. It should be understood that the preceding or following operations are not necessarily performed precisely in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0060] In view of the problems existing in the traditional bucket wheel control system, the present invention provides an intelligent control system and method for bucket wheel based on digital image monitoring, and its technical concept is as follows: the image of the working area of ​​the bucket wheel is normalized after image contrast processing, and then image analysis and working area state feature extraction are performed to obtain a working area state feature map; the working area state feature map is input into a material target object feature highlighter based on a dual optimization attention module to obtain a material target object highlighted working area state feature map; based on the material target object highlighted working area state feature map, a semantic segmentation result of the bucket wheel working area material is generated, and the material target object is tracked based on the semantic segmentation result of the bucket wheel working area material, and the travel angle of the bucket wheel can be controlled and adjusted. In this way, the travel angle of the bucket wheel can be dynamically adjusted based on the image semantics of the working area of ​​the bucket wheel monitored in real time, thereby achieving more efficient and accurate material handling and stacking operations, and improving the intelligence level of the bucket wheel control system.

[0061] One embodiment of the present disclosure relates to an intelligent control system 100 for a bucket wheel machine based on digital image monitoring, and its structure is as follows: Figure 1 As shown, it includes a bucket wheel machine working area image acquisition module 101, a working area image contrast adjustment module 102, a working area state feature extraction module 103, a material target object highlighting working area state representation module 104, a working area material semantic segmentation module 105, and a bucket wheel machine travel angle control module 106.

[0062] The bucket wheel machine working area image acquisition module 101 is used to obtain the bucket wheel machine working area image acquired by a camera deployed on the bucket wheel machine.

[0063] The working area image contrast adjustment module 102 is used to perform image contrast normalization processing on the bucket wheel machine working area image to obtain a contrast normalized bucket wheel machine working area image.

[0064] The working area state feature extraction module 103 is used to perform image analysis and working area state feature extraction on the contrast normalized bucket wheel machine working area image to obtain a working area state feature map.

[0065] The material target object highlighted working area state characterization module 104 is used to input the working area state feature map into the material target object feature highlighter based on the dual optimization attention module to obtain the material target object highlighted working area state feature map.

[0066] The working area material semantic segmentation module 105 is used to highlight the working area state feature map based on the material target object and generate the bucket wheel machine working area material semantic segmentation result.

[0067] The bucket wheel machine travel angle control module 106 is used to track the material target object based on the semantic segmentation results of the material in the bucket wheel machine working area, and control and adjust the travel angle of the bucket wheel machine.

[0068] That is to say, the bucket wheel machine intelligent control system based on digital image monitoring proposed in this embodiment can obtain the image information of the bucket wheel machine working area in real time through the high-definition camera deployed on the bucket wheel machine, and introduce the image processing and analysis algorithm based on artificial intelligence and machine vision at the back end to analyze the bucket wheel machine working area image, so as to capture the target semantic segmentation information about the bucket wheel machine working area, thereby realizing the recognition and tracking of the material target object, and controlling and adjusting the travel angle of the bucket wheel machine according to the recognition and tracking results of the material target object, so as to ensure the accurate loading and stacking of the material. That is, the bucket wheel machine intelligent control system based on digital image monitoring proposed in this embodiment can dynamically adjust the travel angle of the bucket wheel machine based on the image semantics of the bucket wheel machine working area monitored in real time, so as to cope with uncertain factors such as material position changes and loading changes, realize more efficient and accurate material handling and stacking operations, and improve the intelligence level of the bucket wheel machine control system.

[0069] It should be noted that when using a camera to actually collect images of the working area of ​​the bucket wheel machine, it may be affected by lighting conditions, etc., resulting in uneven lighting in the images of the working area of ​​the bucket wheel machine. This may reduce the contrast of the image, thereby affecting the effects of subsequent image processing and analysis. Therefore, in order to deal with the problem of uneven lighting in the working area image of the bucket wheel machine, it is often necessary to use image preprocessing technology to enhance the image quality. Among them, contrast enhancement is a common image preprocessing technology, which can adjust the contrast of the image to make the details in the image clearer and more prominent. Therefore, the bucket wheel machine intelligent control system 100 based on digital image monitoring provided in this embodiment is provided with a working area image contrast adjustment module 102, and the working area image contrast adjustment module 102 is used to perform image contrast normalization processing on the working area image of the bucket wheel machine to obtain a contrast-normalized bucket wheel machine working area image.

[0070] Exemplarily, the working area image contrast adjustment module 102 includes a pixel mean and standard deviation calculation unit and an image contrast normalization unit. The pixel mean and standard deviation calculation unit is used to calculate the global mean and standard deviation of the pixel values ​​in the bucket wheel machine working area image to obtain the global mean of the bucket wheel machine working area image pixels and the standard deviation of the bucket wheel machine working area image pixels. The image contrast normalization unit is used to subtract the global mean of the bucket wheel machine working area image pixels from each pixel value in the bucket wheel machine working area image to obtain the bucket wheel machine working area pixel difference image, and then divide the bucket wheel machine working area pixel difference image pixel by pixel standard deviation of the bucket wheel machine working area image pixel by pixel to obtain the contrast normalized bucket wheel machine working area image.

[0071] By using the pixel mean and standard deviation calculation unit and the image contrast normalization unit to perform contrast normalization processing on the bucket wheel machine working area image, the impact of global contrast changes in the bucket wheel machine working area image can be reduced, and the local details and texture information of the image can be highlighted. This can make the contrast of different areas in the bucket wheel machine working area image more consistent, which is helpful for the subsequent extraction and analysis of target features in the bucket wheel machine working area image to perform semantic segmentation and intelligent control.

[0072] Exemplarily, the working area state feature extraction module 103 is specifically used to: input the contrast-normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on the ResNet model to obtain a working area state feature map.

[0073] Specifically, considering that after image enhancement based on contrast normalization processing, the semantic features of the bucket wheel machine working area contained in the contrast normalized bucket wheel machine working area image obtained are more distinct, therefore, in order to be able to capture the semantics of the bucket wheel machine working area, so as to perform semantic segmentation of the image, and thus identify and track the material target object, this embodiment uses the working area state feature extraction module 103 to input the contrast normalized bucket wheel machine working area image into the bucket wheel machine working area state feature extractor based on the ResNet model for feature extraction to obtain a working area state feature map.

[0074] Among them, the Convolutional Neural Networks (CNN) model in deep learning performs well in the field of image processing and can effectively extract high-level abstract semantic features from the normalized bucket wheel machine working area image. The ResNet model, as a deep residual network, has a strong feature extraction capability and can learn complex features in the image, which helps to accurately capture the state and semantic information of the bucket wheel machine working area. Specifically, the ResNet model solves the gradient vanishing problem in deep neural network training by introducing a residual learning framework, which enables the ResNet model to learn features more deeply, avoids the problem of traditional deep networks being difficult to train, and improves the model's learning ability and generalization performance.

[0075] In particular, considering that the material target object may be small-scale target information in the bucket wheel machine working area image, the semantic features of such small-scale target objects will be submerged in the semantic background of the overall image. Therefore, in order to highlight the semantic feature representation of the material target object in the bucket wheel machine working area image, thereby providing a basis for subsequent image semantic segmentation and identification and tracking of the material target object, this embodiment uses the material target object highlighting working area state representation module 104 to input the working area state feature map into the material target object feature highlighter based on the dual optimization attention module for processing, so as to obtain the material target object highlighting working area state feature map. By applying the material target object feature highlighter based on the dual optimization attention module, the feature representation of the material target object in the working area state feature map can be enhanced, so that the system can more accurately identify and track the material. This attention mechanism helps the system focus on the key semantic information of the material target object in the image and improve the accuracy and robustness of target detection.

[0076] Exemplarily, the material target object highlighting work area state characterization module 104 includes a work area state channel feature manifestation unit, a work area state space feature manifestation unit, and a work area material target object feature highlighting unit. Among them, the work area state channel feature manifestation unit is used to input the work area state feature map into the channel attention optimization network in the material target object feature highlighter based on the dual optimization attention module to obtain the work area state channel salient feature map. The work area state space feature manifestation unit is used to input the work area state feature map into the spatial attention optimization network in the material target object feature highlighter based on the dual optimization attention module to obtain the work area state space salient feature map. The work area material target object feature highlighting unit is used to add and fuse the work area state channel salient feature map and the work area state space salient feature map position by position to obtain the material target object highlighting work area state feature map.

[0077] Exemplarily, the working area state channel feature manifestation unit includes a global mean pooling processing subunit, a first convolution processing subunit, a point convolution adjustment subunit, a first activation processing subunit, and a weighted optimization subunit. Among them, the global mean pooling processing subunit is used to perform global mean pooling processing on each feature matrix along the channel dimension in the working area state feature map to obtain a working area state feature vector. The first convolution processing subunit is used to input the working area state feature vector into a convolution layer based on a convolution kernel of 1×3 for processing to obtain a working area state channel attention vector. The point convolution adjustment subunit is used to perform point convolution on the working area state channel attention vector to adjust the number of channels to obtain the working area state channel adjustment attention vector. The first activation processing subunit is used to perform activation processing based on the Sigmoid function on the working area state channel adjustment attention vector to obtain a working area state channel attention weight vector. The weighted optimization subunit is used to use the feature values ​​of each position in the attention weight vector of the working area state channel as weighting coefficients to perform weighted optimization on each feature matrix of the working area state feature map along the channel dimension to obtain the saliency feature map of the working area state channel.

[0078] In other words, the working area state channel feature manifestation unit can input the working area state feature map into the channel attention optimization network in the material target object feature highlighter based on the dual optimization attention module, and use the channel attention optimization network to process the working area state feature map through the following channel manifestation formula to obtain the working area state channel saliency feature map. The channel manifestation formula includes:

[0079] W C =σ(f 1×1 (f 1×3 (G GAP (F))))

[0080]

[0081] Among them, F is the state feature map of the working area, σ is the Sigmoid function, and f 1×1 (·) represents a convolutional layer with a kernel size of 1×1, f 3×3 (·) represents a convolutional layer with a kernel size of 1×3, G GAP (F) represents the global mean pooling processing of each feature matrix along the channel dimension in the working area state feature map, W C is the working area state channel attention weight vector, Indicates the use of W C Perform weighted processing on F based on the channel dimension, F c This is the saliency feature map of the working area state channel.

[0082] Exemplarily, the working area state space feature manifestation unit includes a second convolution processing subunit, a second activation processing subunit, and a position point multiplication subunit. Among them, the second convolution processing subunit is used to input the working area state feature map into a convolution layer based on a convolution kernel of 1×1 for processing to obtain a working area state space attention feature map. The second activation processing subunit is used to perform activation processing based on the Sigmoid function on the working area state space attention feature map to obtain a working area state space attention weight feature map. The position point multiplication subunit is used to perform position point multiplication on the working area state space attention weight feature map and the working area state feature map to obtain a working area state space saliency feature map.

[0083] In other words, the work area state spatial feature manifestation unit can input the work area state feature map into the spatial attention optimization network in the material target object feature highlighter based on the dual optimization attention module, and use the spatial attention optimization network to process the work area state feature map through the following spatial manifestation formula to obtain the work area state spatial saliency feature map. The spatial manifestation formula includes:

[0084] W S =σ(f 1×1 (F)

[0085] F S =FW S

[0086] Among them, F is the state feature map of the working area, σ is the Sigmoid function, and f 1×1 (·) represents a convolutional layer with a kernel size of 1×1, W S is the attention weight feature map of the working area state space, F S It is the saliency feature map of the working area state space.

[0087] That is to say, the material target object feature highlighter based on the dual optimization attention module includes a channel attention optimization network, a spatial attention optimization network and a parallel fusion network. The channel attention optimization network and the spatial attention optimization network can encode the feature map along the channel dimension and the spatial dimension respectively, so as to highlight the semantic features of the material target object in the feature map. The parallel fusion network connects and aggregates the semantic features of the channel dimension and the spatial dimension to improve the representation ability of the semantics of the material target object. In particular, the traditional channel attention module usually reduces the dimension of the features after global average pooling through the fully connected layer, which may cause the small target feature information about the material target object to be averaged and submerged in the average features with other backgrounds, thereby losing key information. To this end, the channel attention optimization network provided in this embodiment is improved on the traditional channel attention module, which replaces the fully connected layer in the original channel attention module with a convolution layer with a convolution kernel of 1×3 to retain more local information, especially for small target materials, which helps the system to more accurately identify and distinguish targets. In addition, considering that the traditional spatial attention module may lose the key features of small target materials by introducing pooling operations, thereby affecting the accuracy of detection, the spatial attention optimization network provided in this embodiment removes all pooling layers in the traditional spatial attention module, and only generates features through convolution layers with a convolution kernel of 1×1, and then activates the obtained attention features through the Sigmoid function, which helps the system focus on the areas and semantic features related to the material target objects in the image, and improves the recognition and tracking accuracy of material targets. In this way, the system can focus more on the material target objects, reduce interference with irrelevant information, and improve the accuracy and stability of the control system.

[0088] Exemplarily, the working area material semantic segmentation module 105 is specifically used to: input the material target object highlighting the working area state feature map into the classifier-based working area material semantic segmentor to obtain the bucket wheel machine working area material semantic segmentation result.

[0089] Specifically, the semantic features of the working area state after the significant enhancement of the material target object are used for classification processing, so as to achieve fine segmentation of different materials and objects in the working area of ​​the bucket wheel machine, and accurately segment different types of materials and object areas. The working area material semantic segmentation module 105 uses the working area material semantic segmentation of the classifier to semantically understand and classify different areas in the working area image according to the material target object highlighting working area state feature map. In this way, the material location and category in the working area image of the bucket wheel machine can be better understood, and more detailed and accurate information can be provided for subsequent material target tracking and decision control.

[0090] Exemplarily, the bucket wheel machine travel angle control module 106 can use the semantic segmentation results of the materials in the working area of ​​the bucket wheel machine to determine the position and boundary of the material target object, and use the target tracking algorithm (such as Kalman filter algorithm, multi-target tracking algorithm, etc.) to track the material target object, and continuously update the position information of the material target object, so as to ensure that the system can continuously track the motion trajectory of the material target object. The bucket wheel machine travel angle control module 106 can also use the position and dynamic change information of the material target object to automatically control the bucket wheel machine. Specifically, the bucket wheel machine travel angle control module 106 can first calculate the angle difference between the current position and the target position of the bucket wheel machine based on the position information of the material target object and the movement direction of the bucket wheel machine, and determine the travel angle that needs to be adjusted; then, according to the calculation result of the travel angle, formulate a corresponding control strategy, and output an adjustment signal of the control strategy, for example, use a controller to automatically control the steering system of the bucket wheel machine to adjust the corresponding angle, so that the bucket wheel machine moves toward the direction of the material target object. In this way, the bucket wheel machine travel angle control module 106 can dynamically adjust the bucket wheel machine's travel angle based on the image semantics of the bucket wheel machine's working area monitored in real time to cope with uncertain factors such as changes in material position and loading volume, thereby achieving more efficient and precise material handling and stacking operations and improving the intelligence level of the bucket wheel machine control system.

[0091] Exemplarily, the intelligent control system of the bucket wheel machine also includes a training module. The training module is used to train the material target object feature highlighter based on the dual optimization attention module. When the working area state feature extraction module is specifically used to input the contrast normalized bucket wheel machine working area image into the bucket wheel machine working area state feature extractor based on the ResNet model, the training module is also used to: train the bucket wheel machine working area state feature extractor based on the ResNet model. When the working area material semantic segmentation mold body is used to input the material target object highlighting working area state feature map into the working area material semantic segmentor based on the classifier, the training module is also used to: train the working area material semantic segmentor based on the classifier.

[0092] Exemplarily, the training module includes a training bucket wheel machine working area image acquisition unit, a training working area image contrast adjustment unit, a training working area state feature extraction unit, a training material target object highlighting working area state representation unit, a training classification unit, a classification loss function value calculation unit, and a training unit.

[0093] The training bucket wheel machine working area image acquisition unit is used to obtain the training bucket wheel machine working area image collected by the camera deployed on the bucket wheel machine, and the real value corresponding to the semantic segmentation result of the material in the training bucket wheel machine working area.

[0094] The training working area image contrast adjustment unit is used to perform image contrast normalization processing on the training bucket wheel machine working area image to obtain a training contrast normalized bucket wheel machine working area image.

[0095] The training working area state feature extraction unit is used to input the training contrast normalized bucket wheel machine working area image into the bucket wheel machine working area state feature extractor based on the ResNet model for image analysis and working area state feature extraction to obtain a training working area state feature map.

[0096] The training material target object highlighting working area state characterization unit is used to input the training working area state feature map into the material target object feature highlighter based on the dual optimization attention module to obtain the training material target object highlighting working area state feature map.

[0097] The training classification unit is used to highlight the working area state feature map of the training material target object through a classifier-based working area material semantic segmentation device to obtain the training bucket wheel machine working area material semantic segmentation result.

[0098] The classification loss function value calculation unit is used to calculate the cross entropy loss function value between the semantic segmentation result of the material in the working area of ​​the training bucket wheel machine and the corresponding true value to obtain the classification loss function value.

[0099] The training unit is used to train a bucket wheel machine working area state feature extractor based on the ResNet model, a material target object feature highlighter based on a dual optimization attention module, and a working area material semantic segmenter based on a classifier based on the classification loss function value.

[0100] Exemplarily, the training material target object highlighting work area state feature map is passed through a classifier-based work area material semantic segmentation device to obtain a training bucket wheel machine work area material semantic segmentation result, including: dividing each eigenvalue of the training material target object highlighting work area state feature map by the maximum eigenvalue of the training material target object highlighting work area state feature map to obtain the training material target object highlighting work area state interaction representation map; dividing the eigenvalue mean of the training material target object highlighting work area state feature map by the eigenvalue standard deviation of the training material target object highlighting work area state feature map to obtain a statistical interaction value corresponding to the training material target object highlighting work area state feature map; subtracting the corresponding eigenvalue of each eigenvalue of the training material target object highlighting work area state interaction representation map from the eigenvalue standard deviation of the training material target object highlighting work area state feature map; After counting the interaction values, the base-two logarithm of the absolute value of each position is calculated to obtain the training material target object salient working area state interaction information representation diagram; each eigenvalue of the training material target object salient working area state interaction representation diagram is added with the corresponding statistical interaction value and then multiplied by the predetermined weight hyperparameter to obtain the training material target object salient working area state interaction mode representation diagram; the training material target object salient working area state interaction information representation diagram and the training material target object salient working area state interaction mode representation diagram are added to obtain the optimized material target object salient working area state feature diagram; and the optimized material target object salient working area state feature diagram is input into the classifier-based working area material semantic segmentor to obtain the training bucket wheel excavator working area material semantic segmentation result.

[0101] That is to say, in order to improve the feature expression effect of the multi-dimensional attention enhanced fusion feature distribution representation of the image semantic features of the bucket wheel machine working area image in the training material target object highlighting work area state feature map, this embodiment uses a short sequence of interactive representation of probabilistic statistical features represented by the mean and the standard value and the interactive representation of the distribution of eigenvalues ​​and maximum eigenvalues ​​as latent variable features as the sub-manifold potential model under the complex manifold network of the training material target object highlighting work area state feature map, to compare the obtained training material target object highlighting work area state feature map based on the training material target object highlighting work area state interactive information representation map with the training material target object highlighting work area state feature map. The latent motif feature information pattern and feature distribution pattern of the state interaction pattern representation graph are used as its global structure inference unit, so as to reconstruct the complex manifold structure of the state feature graph of the training material target object highlighting the working area based on the connection in the form of the global structure latent motif dictionary, so as to improve the generation and evolution understanding ability of the model for the manifold structure corresponding to the features under the complex feature representation dimension in the iterative process, and improve the feature expression effect of the complex feature representation dimension of the state feature graph of the training material target object highlighting the working area during the model iteration process, and then improve the accuracy of the semantic segmentation result of the bucket wheel machine working area material obtained by the classifier-based working area material semantic segmentation device of the training material target object highlighting the working area state feature graph. In this way, the material target object can be identified and tracked more accurately, so that the travel angle of the bucket wheel machine can be more accurately controlled to ensure the accurate loading and stacking of materials, and the intelligent level of the bucket wheel machine control system can be improved.

[0102] The bucket wheel machine intelligent control system based on digital image monitoring provided by the embodiment of the present disclosure acquires the image information of the bucket wheel machine working area in real time by deploying a high-definition camera on the bucket wheel machine, and introduces an image processing and analysis algorithm based on artificial intelligence and machine vision at the back end to analyze the image of the bucket wheel machine working area, so as to capture the target semantic segmentation information about the bucket wheel machine working area, thereby realizing the recognition and tracking of the material target object, and controlling and adjusting the travel angle of the bucket wheel machine according to the recognition and tracking results of the material target object, so as to ensure the accurate loading and stacking of the material by controlling and adjusting the travel angle of the bucket wheel machine. That is, the bucket wheel machine intelligent control system based on digital image monitoring proposed in this embodiment can dynamically adjust the travel angle of the bucket wheel machine based on the image semantics of the bucket wheel machine working area monitored in real time, so as to cope with uncertain factors such as material position changes and loading changes, realize more efficient and accurate material handling and stacking operations, and improve the intelligence level of the bucket wheel machine control system.

[0103] Another embodiment of the present disclosure relates to an intelligent control method for a bucket wheel machine based on digital image monitoring, the process of which is as follows: Figure 2As shown, including:

[0104] Step S201, obtaining an image of a working area of ​​a bucket wheel machine collected by a camera deployed on the bucket wheel machine.

[0105] Step S202, performing image contrast normalization processing on the bucket wheel machine working area image to obtain a contrast normalized bucket wheel machine working area image.

[0106] Step S203, performing image analysis and working area state feature extraction on the contrast normalized bucket wheel machine working area image to obtain a working area state feature map.

[0107] Step S204, inputting the working area state feature map into the material target object feature highlighter based on the dual optimization attention module to obtain the material target object highlighted working area state feature map.

[0108] Step S205, based on the material target object highlighting the working area state feature map, generate the bucket wheel machine working area material semantic segmentation result.

[0109] Step S206, based on the semantic segmentation result of the material in the working area of ​​the bucket wheel machine, the material target object is tracked, and the travel angle of the bucket wheel machine is controlled and adjusted.

[0110] The specific implementation steps of each step in the bucket wheel machine intelligent control method based on digital image monitoring provided in the embodiment of the present disclosure can be referred to the specific implementation method of each module in the bucket wheel machine intelligent control system based on digital image monitoring provided in the embodiment of the present disclosure, and will not be repeated here.

[0111] Compared with the prior art, the intelligent control method of the bucket wheel machine based on digital image monitoring provided by the embodiment of the present disclosure can dynamically adjust the travel angle of the bucket wheel machine based on the image semantics of the bucket wheel machine's working area monitored in real time to cope with uncertain factors such as changes in material position and loading volume, thereby achieving more efficient and precise material handling and stacking operations and improving the intelligence level of the bucket wheel machine control system.

[0112] Another embodiment of the present disclosure relates to an electronic device 600, whose structure is as follows: Figure 3 The terminal device corresponding to the electronic device 600 may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and fixed terminals such as digital televisions, desktop computers, and the like. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0113] like Figure 3 As shown, the electronic device 600 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage device 608 to the random access memory (RAM) 603, so as to realize the bucket wheel machine intelligent control method based on digital image monitoring described in the above embodiment. In RAM603, various programs and data required for the electronic device 600 to perform various operations are also stored. The processing device 601, ROM602 and RAM603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0114] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 600 is shown to have various devices, but it should be understood that this does not require the electronic device 600 to implement or have all of the devices shown, and the electronic device 600 may instead implement or have more or fewer devices.

[0115] Another embodiment of the present disclosure relates to a computer program product, including a computer program, the computer program including program code for executing the bucket wheel machine intelligent control method based on digital image monitoring described in the above embodiment, and when the computer program is executed by a processor, the bucket wheel machine intelligent control method based on digital image monitoring described in the above embodiment is implemented. The computer program can be downloaded and installed from the network through a communication device 609, or installed from a storage device 608, or installed from a ROM 602, and when the computer program is executed by a processing device 601, some or all of the above steps defined in the bucket wheel machine intelligent control method based on digital image monitoring described in the above embodiment are implemented.

[0116] Another embodiment of the present disclosure relates to a computer-readable medium, which carries the computer program described in the above embodiment, and when the computer program is executed, it implements the bucket wheel machine intelligent control method based on digital image monitoring described in the above embodiment.

[0117] Among them, the computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium or any combination of a computer-readable signal medium and a computer-readable storage medium. The computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium containing or storing a computer program, which can be used by or in combination with an instruction execution system, device or device. In this embodiment, the computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal can take a variety of forms, and can include, but is not limited to, an electromagnetic signal, an optical signal, or any suitable combination of an electromagnetic signal and an optical signal. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0118] The computer-readable medium may be included in the electronic device, or may exist independently but not installed in the electronic device.

[0119] Exemplarily, one or more programming languages ​​or a combination thereof may be used to write computer program codes for executing the intelligent control method for bucket wheel machine based on digital image monitoring described in the above embodiment. The above programming languages ​​may include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. The computer program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0120] Exemplarily, the client and server including the user computer, the remote computer can communicate using any currently known or future developed network protocol such as HTTP (Hyper Test Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., communication network). Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), internets (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.

[0121] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0122] The modules involved in the embodiments described in the present disclosure may be implemented by software or hardware. The name of a module does not limit the module itself in some cases. For example, a bucket wheel machine working area image acquisition module may also be described as a "module for acquiring a working area image of a bucket wheel machine".

[0123] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0124] Figure 4 The application scenario diagram of the bucket wheel machine intelligent control system based on digital image monitoring and the bucket wheel machine intelligent control method based on digital image monitoring provided by the embodiment of the present disclosure is shown. Figure 4 As shown, in this application scenario, first, an image of the working area of ​​the bucket wheel machine collected by a camera deployed on the bucket wheel machine is obtained. For example, the image of the working area of ​​the bucket wheel machine can be Figure 4 Then, the obtained bucket wheel machine working area image is input into a server deployed with a bucket wheel machine intelligent control method based on digital image monitoring. For example, the server may be Figure 4 As shown in the figure, the server can process the image of the bucket wheel machine working area based on the bucket wheel machine intelligent control method based on digital image monitoring, track the material target object based on the semantic segmentation result of the bucket wheel machine working area material, and control and adjust the travel angle of the bucket wheel machine according to the tracking result.

[0125] Those skilled in the art will appreciate that the above-mentioned embodiments are specific embodiments for implementing the present disclosure, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present disclosure.

Claims

1. An intelligent control system for bucket wheel excavator based on digital image monitoring, characterized in that: The bucket wheel machine intelligent control system comprises: A bucket wheel machine working area image acquisition module, used to acquire an image of the bucket wheel machine working area acquired by a camera deployed on the bucket wheel machine; A working area image contrast adjustment module, used for performing image contrast normalization processing on the bucket wheel machine working area image to obtain a contrast normalized bucket wheel machine working area image; A working area state feature extraction module, used for performing image analysis and working area state feature extraction on the contrast normalized bucket wheel machine working area image to obtain a working area state feature map; A material target object highlighting work area state characterization module, used for inputting the work area state feature map into a material target object feature highlighter based on a dual optimization attention module to obtain a material target object highlighting work area state feature map; A working area material semantic segmentation module is used to generate a semantic segmentation result of the bucket wheel machine working area material based on the material target object highlighting the working area state feature map; The bucket wheel machine travel angle control module is used to track the material target object based on the semantic segmentation result of the material in the working area of ​​the bucket wheel machine, and control and adjust the travel angle of the bucket wheel machine.

2. The intelligent control system of bucket wheel excavator according to claim 1 is characterized in that: The working area image contrast adjustment module comprises: A pixel mean and standard deviation calculation unit, used to calculate the global mean and standard deviation of the pixel values ​​in the bucket wheel machine working area image, so as to obtain the global mean of the bucket wheel machine working area image pixels and the standard deviation of the bucket wheel machine working area image pixels; An image contrast normalization unit is used to subtract the global mean value of the bucket wheel machine working area image pixels from each pixel value in the bucket wheel machine working area image to obtain a bucket wheel machine working area pixel difference image, and then divide the bucket wheel machine working area pixel difference image pixel by pixel standard deviation of the bucket wheel machine working area image to obtain the contrast normalized bucket wheel machine working area image.

3. The intelligent control system of bucket wheel excavator according to claim 1 is characterized in that: The working area state feature extraction module is specifically used to: input the contrast normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on the ResNet model to obtain the working area state feature map.

4. The bucket wheel excavator intelligent control system according to claim 1, characterized in that: The material target object highlights the working area state representation module, including: A working area state channel feature manifestation unit, used for inputting the working area state feature map into the channel attention optimization network in the material target object feature highlighter based on the dual optimization attention module to obtain a working area state channel saliency feature map; A working area state space feature manifestation unit, used for inputting the working area state feature map into the spatial attention optimization network in the material target object feature highlighter based on the dual optimization attention module to obtain a working area state space saliency feature map; The working area material target object feature highlighting unit is used to add and fuse the working area state channel salient feature map and the working area state space salient feature map position by position to obtain the material target object highlighted working area state feature map.

5. The intelligent control system of bucket wheel excavator according to claim 4 is characterized in that: The working area state channel feature manifestation unit includes: A global mean pooling processing subunit, used for performing global mean pooling processing on each feature matrix along the channel dimension in the working area state feature map to obtain a working area state feature vector; A first convolution processing subunit is used to input the working area state feature vector into a convolution layer based on a convolution kernel of 1×3 for processing, so as to obtain a working area state channel attention vector; A point convolution adjustment subunit, used for performing point convolution on the working area state channel attention vector to adjust the number of channels, so as to obtain a working area state channel adjustment attention vector; A first activation processing subunit is used to perform activation processing based on a Sigmoid function on the work area state channel adjustment attention vector to obtain a work area state channel attention weight vector; The weighted optimization subunit is used to use the various position feature values ​​in the working area state channel attention weight vector as weighting coefficients to perform weighted optimization on the various feature matrices of the working area state feature map along the channel dimension to obtain the working area state channel saliency feature map.

6. The intelligent control system of bucket wheel excavator according to claim 4 is characterized in that: The working area state space feature manifestation unit comprises: A second convolution processing subunit is used to input the working area state feature map into a convolution layer based on a convolution kernel of 1×1 for processing to obtain a working area state space attention feature map; A second activation processing subunit is used to perform activation processing based on a Sigmoid function on the working area state space attention feature map to obtain a working area state space attention weight feature map; The position point multiplication subunit is used to perform position point multiplication on the working area state space attention weight feature map and the working area state feature map to obtain the working area state space saliency feature map.

7. The intelligent control system of bucket wheel excavator according to claim 1 is characterized in that: The working area material semantic segmentation module is specifically used to: input the material target object highlighting the working area state feature map into the working area material semantic segmentor based on the classifier to obtain the semantic segmentation result of the bucket wheel machine working area material.

8. The bucket wheel machine intelligent control system according to claim 3 or 7, characterized in that: The bucket wheel machine intelligent control system also includes a training module; The training module is used to train the material target object feature highlighter based on the dual optimization attention module; When the working area state feature extraction module is specifically used to input the contrast-normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on a ResNet model, the training module is further used to: train the bucket wheel machine working area state feature extractor based on a ResNet model; When the working area material semantic segmentation mold body is used to input the material target object highlighting the working area state feature map into the classifier-based working area material semantic segmentor, the training module is also used to: train the classifier-based working area material semantic segmentor.

9. The intelligent control system of bucket wheel excavator according to claim 8, characterized in that: The training module comprises: A training bucket wheel machine working area image acquisition unit is used to obtain images of the training bucket wheel machine working area acquired by a camera deployed on the bucket wheel machine, and real values ​​corresponding to semantic segmentation results of materials in the training bucket wheel machine working area; A training working area image contrast adjustment unit, used for performing image contrast normalization processing on the training bucket wheel machine working area image to obtain a training contrast normalized bucket wheel machine working area image; A training working area state feature extraction unit is used to input the training contrast normalized bucket wheel machine working area image into a bucket wheel machine working area state feature extractor based on a ResNet model to perform image analysis and work area state feature extraction to obtain a training working area state feature map; A training material target object highlighting work area state characterization unit, used for inputting the training work area state feature map into the material target object feature highlighter based on the dual optimization attention module to obtain a training material target object highlighting work area state feature map; A training classification unit, used for passing the training material target object highlighting the working area state feature map through the classifier-based working area material semantic segmentor to obtain the training bucket wheel machine working area material semantic segmentation result; A classification loss function value calculation unit is used to calculate the cross entropy loss function value between the semantic segmentation result of the material in the working area of ​​the training bucket wheel machine and the corresponding true value to obtain the classification loss function value; A training unit is used to train the bucket wheel machine working area state feature extractor based on the ResNet model, the material target object feature highlighter based on the dual optimization attention module, and the working area material semantic segmentor based on the classifier based on the classification loss function value.

10. A bucket wheel machine intelligent control method based on digital image monitoring, characterized in that: The bucket wheel machine intelligent control method comprises: Acquire an image of a working area of ​​a bucket wheel machine collected by a camera deployed on the bucket wheel machine; Performing image contrast normalization processing on the bucket wheel machine working area image to obtain a contrast normalized bucket wheel machine working area image; Performing image analysis and working area state feature extraction on the contrast-normalized bucket wheel machine working area image to obtain a working area state feature map; Inputting the working area state feature map into a material target object feature highlighter based on a dual optimization attention module to obtain a material target object highlighted working area state feature map; Based on the material target object highlighting the working area state feature map, generating a semantic segmentation result of the bucket wheel machine working area material; Based on the semantic segmentation results of the materials in the working area of ​​the bucket wheel machine, the material target object is tracked, and the travel angle of the bucket wheel machine is controlled and adjusted.

Images