Intelligent remote control water cannon system for mine watering cart and use method of intelligent remote control water cannon system

Abstract

The invention belongs to the technical field of strip mine watering carts, and discloses an intelligent remote control water cannon system for a mine watering cart and a using method of the intelligent remote control water cannon system. The water cannon unit comprises a base, a platform, a water supply pipe, a pitching mechanism which is horizontally arranged and rotationally connected with the water supply pipe, a launching pipe fixedly connected with the pitching mechanism and a second fixing frame which is arranged above the platform and fixedly connected with the platform; a hydraulic cylinder is arranged on the second fixing frame, the output end of the hydraulic cylinder is fixedly connected with a telescopic rod, and the other end of the telescopic rod is hinged to the launching tube; the control unit comprises a CPU, a watering cart monitoring system, an automatic and manual switching module, a mobile signal analog conversion control module, an operation part, a vehicle-mounted camera, an image recognition device, a water cannon switching control module and a control panel. According to the water spraying vehicle, operation convenience, high automation and an excellent spraying effect can be achieved, so that the automation level of the water spraying vehicle is improved, and the spraying efficiency is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of open-pit mine sprinkler trucks, and particularly to an intelligent remote control water cannon system for mine sprinkler trucks and a usage method thereof. Background Art

[0002] In the mine operation environment, the operation of the water cannon of traditional mine sprinkler trucks mostly relies on manual control. The operator needs to operate on the vehicle or in the close range, which not only exposes the operator to the complex and potentially dangerous environment in the mine, such as respiratory diseases caused by dust diffusion, personal injuries caused by ore splashing, etc., but also in some special scenarios, such as high-temperature, high-noise areas or sections with leakage of toxic and harmful gases, manual operation is difficult to effectively carry out. At the same time, manual operation is limited by the reaction speed and operation accuracy of people, and it is difficult to quickly and accurately adjust parameters such as the spraying angle and flow rate of the water cannon to meet the requirements of complex and variable dust suppression, fire extinguishing and other operations. In addition, with the expansion of mine scale and the increasing trend of automated operation, the traditional manual control method has been difficult to meet the requirements of efficient, safe and intelligent mine operation. Therefore, the research and development of a remote control water cannon system for mine sprinkler trucks has extremely important practical significance and urgency.

[0003] Chinese Utility Model Patent: The publication number is "CN217399511U", and the name is "A water cannon for a sprinkler truck with automatic adjustment", which discloses a water cannon for a sprinkler truck with automatic adjustment, including a water delivery pipe, a butterfly valve, an explosion-proof electric valve, a first connecting pipe, a second connecting pipe, a water spraying cylinder, a support seat, a protective shell, and a first stepping motor. In this technical solution, the water delivery pipe and the first connecting pipe are rotatably connected in the horizontal direction through a first connecting bearing, and a driven gear is fixed on the outer ring of the first connecting bearing. Support seats are horizontally and vertically connected and fixed on the vertical rod and the cross rod of the first connecting pipe respectively. A first stepping motor is fixed on the horizontally arranged support seat, so that the water spraying cylinder can rotate and adjust the angle in the horizontal direction; the second connecting pipe is rotatably connected with the first connecting pipe in the vertical direction through a second connecting bearing, and a driven gear is fixed on the outer ring of the second connecting bearing. A second stepping motor is fixed on the vertically arranged support seat, so that the water spraying cylinder can rotate and adjust the angle in the vertical direction. It can solve the problem that the traditional manual control is difficult to meet the requirements of efficient and safe mine operation to a certain extent. However, this technical solution can only adjust the height and angle of the water cannon system, and cannot adjust the water spraying amount and water spraying pressure, so as to adapt to different operation requirements. Summary of the Invention

[0004] In order to solve the problem in the above-mentioned existing technology that the water cannon system cannot automatically adjust the water spraying amount and water spraying pressure, and thus cannot adapt to different operation requirements, the present invention proposes an intelligent remote control water cannon system for a mine use sprinkler truck and a usage method, aiming to provide a remote control water cannon system for a mine use sprinkler truck that is convenient to operate, has a high degree of automation, good spraying effect, can realize vertical pitching and horizontal rotation actions, full-automatic spraying, is safe and reliable, improves the spraying efficiency of the open-pit mine sprinkler truck, reduces the labor cost, and improves the economic benefits of the coal mine.

[0005] The present invention is realized through the following technical solutions: It includes a water cannon unit and a control unit for controlling the firing of the water cannon unit. The water cannon unit includes a base arranged inside the sprinkler truck, a platform arranged above the base, a water supply pipe passing through the base and the platform and connected to the water tank inside the sprinkler truck, a pitching mechanism horizontally arranged and rotatably connected to the water supply pipe, a firing pipe fixedly connected to the pitching mechanism, and a second fixing frame arranged above the platform and fixedly connected to the platform. The base passes through the vehicle body of the sprinkler truck and is fixedly connected to the sprinkler truck; a hydraulic cylinder is arranged on the second fixing frame, the output end of the hydraulic cylinder is fixedly connected with a telescopic rod, and the other end of the telescopic rod is hinged to the firing pipe; the control unit includes:

[0006] A CPU, which provides computing power support for the entire system and conveys control commands;

[0007] A sprinkler truck monitoring system, which is signal-connected to the CPU and is used for transmitting monitoring data;

[0008] An automatic-manual switching module, with its input end signal-connected to the operating component and its sending end signal-connected to the CPU, is used for receiving the control signal of the operating component and transmitting it to the CPU;

[0009] A mobile signal analog conversion control module, which is signal-connected to the CPU and the operating component, is used for receiving the control signal of the operating component and transmitting it to the CPU;

[0010] An operating component, which is used for manually performing pitching actions and rotation actions on the water cannon unit;

[0011] A vehicle-mounted camera, which is used for taking real-time images of the operation site and sending them to the CPU;

[0012] An image recognition device, which is used for receiving the image data transmitted by the CPU to perform image recognition, judging the pollution situation of the operation site, and sending the judgment result to the CPU;

[0013] A water cannon switching control module, which is signal-connected to the CPU, is used for receiving the action control command of the CPU for the water cannon unit and sending the corresponding control command to the water cannon unit;

[0014] The control panel is installed in the cab of the sprinkler truck and is used to receive the images and monitoring data transmitted by the CPU and send the driver's operation instructions to the CPU.

[0015] As a further preference, the water cannon unit further includes a motor disposed inside the base, a first gear fixedly connected to the fixed end of the motor, a rotating shaft disposed at the center of the bottom of the platform and fixedly connected thereto, and a second gear fixedly connected to the rotating shaft. The output end of the motor faces the platform vertically, the rotating shaft is parallel to the output end of the motor, and the second gear meshes with the first gear.

[0016] As a further preference, the sprinkler truck monitoring system consists of a pressure sensor, a flow sensor, a liquid level sensor, a mileage sensor, and a water filling times counter, all of which are signal-connected to the CPU; the liquid level sensor is disposed inside the water tank of the sprinkler truck, the mileage sensor is disposed in the cab, the pressure sensor is disposed inside the launch tube, the flow sensor is disposed inside the water supply pipe, and the water filling times counter is disposed at the water inlet of the water tank of the sprinkler truck.

[0017] As a further preference, the control unit further includes an alarm, and the alarm is signal-connected to the CPU.

[0018] As a further preference, the control unit further includes an information memory, and the information memory is signal-connected to the CPU.

[0019] As a further preference, the hinge point of the telescopic rod and the launch tube is far from the pitching mechanism.

[0020] As a further preference, a first fixing bracket may be provided between the telescopic rod and the launch tube.

[0021] The present invention also provides a usage method applicable to the intelligent remote control water cannon system of the mining sprinkler truck, including the following steps:

[0022] S1: During the driving process of the mining sprinkler truck, use the on-vehicle camera to collect images of the mine working area from different angles;

[0023] S2: Transmit the images captured by the on-vehicle camera to the image recognition device through the CPU. The image recognition device uses a deep learning algorithm to perform learning processing on the images captured by the on-vehicle camera and enhance the images captured by the on-vehicle camera to obtain enhanced images;

[0024] S21: For the input images, use a matrix as the convolution kernel, use the values of the matrix as the weight parameters of the neural network, and move the convolution kernel from the upper left corner of the input image from left to right and from top to bottom with a step size of 1 to perform convolution operations to obtain a number of convolution feature matrices. The formula for the value of each element in a single convolution feature matrix is as follows:

[0025] h = f(W (1) X small + b (1) )

[0026] where X small represents the small image extracted from the large image by the convolution kernel, W (1) and b (1) represent the weight matrix and bias term corresponding to the first convolution kernel, and h is the element value at the position corresponding to X small in the first convolution feature matrix, and f(*) is the activation function;

[0027] During the convolution process, the spatial domain function and the range kernel function are combined to blur the image captured by the vehicle-mounted camera. The formula is as follows:

[0028]

[0029] where is for the purpose of normalization; σ s is the standard deviation of the spatial domain Gaussian function, σ r is the standard deviation of the range Gaussian function, and Ω represents the domain of convolution;

[0030] S22: Construct the final recognition model based on the weakly supervised learning data augmentation system, the attention network, and the Mobilenet-v2 network, and obtain the augmented image for the input image. The formula is as follows:

[0031]

[0032] where f(*) is the convolution function, and A i represents a component of the recognition object;

[0033] S23: Use the loss function to quantitatively evaluate the augmented image.

[0034] S3: Perform recognition and judgment on the augmented image obtained in step S2, determine the decontamination, dust reduction, and avoidance operations, and transmit the judgment result to the CPU;

[0035] S31: Compare the augmented image obtained in step S2 with the preset clean area image features, where the image features include color, texture, clarity, and contrast. If there is a color deviation from the preset clean area and the texture is rough in the augmented image, it is determined as a dirty area and decontamination operations are performed;

[0036] S311: Represent the dirt level with n, which is divided into three categories: n = 1 for mild dirt, n = 2 for moderate dirt, and n = 3 for severe dirt; the corresponding water volumes for the three levels are P 1 , P2 , P 3 , where P 1 <P 2 <P 3 ;

[0037] If n = 1, then select the water volume [P 1 , P 3 ;

[0038] If n = 2, then select the water volume [P i , P j , where i < j; j ∈ [2, 3];

[0039] If n = 3, then select the water volume [P 3 ;

[0040] S312: The spraying height level is represented by h, which are low, medium, and high respectively; the pitch angles corresponding to the three heights are θ 1 , θ 2 , θ 3 ; where θ 1 < θ 2 < θ 3 ;

[0041] If h = 1, then select the pitch angle as [θ 1 , θ 3 ;

[0042] If h = 2, then select the pitch angle as [θ i , θ j , where i < j; j ∈ [2, 3];

[0043] If h = 3, then select the pitch angle as [θ 3 ;

[0044] S32: Compare the enhanced image obtained in step S2 with the preset clean area image features. The image features include color, texture, sharpness, and contrast. If the enhanced image has low sharpness, weak contrast, and a dull color, it is determined that there is a lot of dust, and a dust reduction operation is performed;

[0045] S33: Perform object recognition on the enhanced image obtained in step S2, and further analyze its position, shape, and size features; for the transport vehicle, determine its position coordinates, body contour, and driving direction in the image; for the ore pile, analyze its shape, volume, and distribution range. At the same time, combine the driving trajectory and speed of the mining water truck to predict the relative movement trend between the object and the water truck, and perform sprinkler path planning;

[0046] S4: Based on the judgment result transmitted by the image recognition device, the CPU sends a control command to the water cannon switching control module, and the water cannon switching control module sends the corresponding control command to the water cannon unit;

[0047] S5: During the operation, when the sprinkler monitoring system detects an abnormal situation, where the abnormal situation includes that the pressure sensor detects that the pressure exceeds the threshold and the liquid level sensor detects that the liquid level is lower than the threshold, an alarm signal is sent to the CPU, and the CPU transmits the alarm signal to the alarm for alarm prompt; meanwhile, the image recognition device continuously monitors the spraying condition;

[0048] S6: After the cleaning operation is completed, images of the same area are taken again; by comparing the color, texture changes of the dirty areas in the images before and after cleaning, as well as the reduction degree of the dust concentration, the cleaning effect is evaluated.

[0049] As a further preference, the loss function for quantitatively evaluating the enhanced image in step S23 is the object recognition loss function or the object localization loss function, and the formulas are as follows respectively:

[0050]

[0051] where p i is the probability generated by the network, represents the label of the sample; represents the label of the sample; is the regression target obtained by the network, is its true value.

[0052] As a further preference, the cleaning effect in step S6 is evaluated by one or a combination of the mean squared error loss function, the structural similarity loss function, and the Euclidean loss function.

[0053] The formula of the mean squared error loss function is as follows:

[0054]

[0055] y i is the actual cleanliness of a certain area in the image, is the predicted cleanliness;

[0056] The formula of the structural similarity loss function is as follows:

[0057] l(x,y) = (2 * μx * μy + C1) / (μx^2 + μy^2 + C1)

[0058] c(x,y) = (2 * σxy + C2) / (σx^2 + σy^2 + C2)

[0059] s(x,y)=(σxy+C3) / (σx * σy+C3)

[0060] SSIM(x,y)=l(x,y)*c(x,y)*s(x,y)

[0061] Where: SSIM(x,y) is the structural similarity loss function; l(x,y) is the brightness similarity function; s(x,y) is the structural similarity function; c(x,y) is the contrast similarity function; x, y represent two images, μx, μy represent the average values ​​of the brightness of x, y; σxy represents the covariance of the brightness of x, y, σx, σy represent the variance of the brightness of x, y; C1, C2, C3 represent constants;

[0062] The Euclidean loss function formula is as follows:

[0063]

[0064] in represents the prediction target of the feature points obtained from the network, Represents its true value.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. In terms of operational safety, the remote control water cannon system for a mining sprinkler truck of the present invention allows the operator to perform remote control at a location far away from dangerous working areas, effectively avoiding exposure of the operator to potential hazards such as dust, ore splashing, high temperature, high noise, and toxic and harmful gases in the harsh environment of the mine, thereby greatly ensuring the personal safety of the operator.

[0067] 2. The remote control water cannon system for a mining sprinkler truck of the present invention has high accuracy and efficiency in operation. The remote control operation can quickly and accurately adjust the vertical pitch angle and horizontal rotation angle of the water cannon according to different operation requirements with the help of the powerful information processing capability of the control unit, and can precisely control parameters such as the flow rate of the water cannon at the same time. The coordinated work of the image recognition device further improves the accuracy of the operation, and the use of the image recognition algorithm can make the recognition more accurate.

[0068] 3. The present invention can more efficiently complete complex and changeable tasks such as dust reduction and fire extinguishing, significantly improving the work efficiency and quality. The automatic / manual switching module of the present invention gives the system a high degree of flexibility, which can perform manual precise operations when manual intervention is required, and can also achieve automatic operation under routine operations or preset conditions. It adapts to the diverse needs of different operation scenes and working conditions in the mine, reduces the labor intensity and tediousness of manual operations, and improves the overall intelligent level of operations.

[0069] 4. The image recognition device provided in the present invention monitors the spraying status in real time during the operation. When it is found that some areas do not achieve the expected spraying effect, this information is timely fed back to the CPU. Based on this, the CPU readjusts the spraying parameters of the water cannon to ensure that the entire operation area can be evenly and effectively sprayed, thereby further guaranteeing the quality and effect of the operation, and making the water cannon system of the present invention have stronger adaptability and practicability in the mine operation.

[0070] 5. The monitoring system of the sprinkler truck designed in the present invention can monitor multiple key parameters of the sprinkler truck in real time and store these data in the information memory, which is convenient for subsequent data analysis and vehicle maintenance management, helps to optimize the operation plan and resource allocation of the sprinkler truck, and further improves the economy and scientificity of the mine operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a schematic diagram of the overall structure of the water cannon launch tube of the present invention.

[0072] Figure 2 It is a schematic diagram of the water cannon launch tube facing downwards of the present invention.

[0073] Figure 3 It is a schematic diagram of the structure in which the water cannon of the present invention is rotatably connected to the base.

[0074] Figure 4 It is a schematic diagram of the horizontal rotation range of the water cannon of the present invention.

[0075] Figure 5 It is a block diagram of the water cannon system of the present invention.

[0076] Labels in the figure:

[0077] 1. Launch tube; 2. Pitching mechanism; 3. Water supply pipe; 4. Hydraulic cylinder; 5. Telescopic rod; 6. First fixing frame; 7. Second fixing frame; 8. Platform; 9. Base; 10. Rotating shaft; 11. First gear; 12. Motor; 13. Second gear. DETAILED DESCRIPTION OF THE INVENTION

[0078] The advantages and features of the present invention will be illustrated and explained by the following non-restrictive description of the preferred embodiments, which are given only as examples with reference to the accompanying drawings.

[0079] Such as Figures 1 to 3As shown in the figure, the present invention provides an intelligent remote control water cannon system for a mine sprinkler truck, which includes a water cannon unit and a control unit for controlling the firing of the water cannon unit. The water cannon unit includes a base 9 disposed inside the sprinkler truck, and the base 9 penetrates the vehicle body of the sprinkler truck and is fixedly connected to the sprinkler truck. A platform 8 is disposed above the base 9, and a motor 12 is disposed inside the base 9. The output end of the motor 12 vertically faces the platform 8, and a first gear 11 is fixedly connected to the output end of the motor 12. A rotating shaft 10 is disposed at the center of the bottom of the platform 8, and a second gear 13 is fixedly connected to the rotating shaft 10. The second gear 13 meshes with the first gear 11. The rotating shaft 10 is parallel to the output end of the motor 12. Driven by the motor 12, the platform 8 can rotate relative to the base 9. The water cannon unit further includes a water supply pipe 3. One end of the water supply pipe 3 penetrates the platform 8 and the base 9 and is finally connected to the water tank inside the sprinkler truck, and the other end is rotatably connected to a pitching mechanism 2. A firing tube 1 is fixedly connected to the pitching mechanism 2. A second fixing frame 7 is fixedly connected above the platform 8, and a hydraulic cylinder 4 is disposed on the second fixing frame 7. The output end of the hydraulic cylinder 4 is fixedly connected to a telescopic rod 5, and the other end of the telescopic rod 5 is hinged to the firing tube 1. Preferably, the hinge point between the telescopic rod 5 and the firing tube 1 is far from the pitching mechanism 2. A first fixing frame 6 can be disposed between the telescopic rod 5 and the firing tube 1 to facilitate fixing the telescopic rod 5. The pitching mechanism 2 is horizontally disposed and can realize the pitching action of the firing tube 1 when rotating relative to the water supply pipe 3. The power of the pitching action comes from the hydraulic cylinder 4. The extension and contraction action of the output end of the hydraulic cylinder 4 can drive the telescopic rod 5 to extend and contract, thereby driving the relative rotation between the pitching mechanism 2 and the water supply pipe 3, and finally realizing the pitching action of the firing tube 1.

[0080] The main body of the control unit is disposed in the cab of the sprinkler truck. The control unit includes a CPU, an on-vehicle camera, a control panel, an operating component, an image recognition device, a sprinkler truck monitoring system, a mobile signal analog conversion control module, an alarm, a water cannon switching control module, and an automatic / manual switching module.

[0081] The CPU is signal-connected to the control panel, the sprinkler truck monitoring system, the image recognition device, the mobile signal analog conversion control module, the alarm, the water cannon switching control module, the automatic / manual switching module, and the on-vehicle camera, and is used for receiving the image data sent by the on-vehicle camera, the monitoring data sent by the sprinkler truck monitoring system, the image recognition result sent by the image recognition device, and the control signals sent by the control panel, the mobile signal analog conversion control module, and the automatic / manual switching module, and sending the control signals to the alarm, the mobile signal analog conversion control module, the image recognition device, the sprinkler truck monitoring system, and the water cannon switching control module. The CPU provides computing power support for the entire system and conveys control commands.

[0082] The sprinkler monitoring system consists of a pressure sensor, a flow sensor, a liquid level sensor, a mileage sensor, and a water filling times counter, all of which are signal-connected to the CPU. The liquid level sensor is arranged in the water tank of the sprinkler to measure the liquid level height in the water tank of the sprinkler and transmit the liquid level signal to the CPU. The CPU judges whether the current water volume can meet the requirements of this sprinkler operation according to the preset liquid level threshold, so as to ensure the smooth completion of the sprinkler operation. The mileage sensor is arranged in the cab. After the sprinkler starts and travels, the mileage sensor monitors the vehicle driving mileage in real time and continuously transmits the data to the CPU. The pressure sensor is arranged in the emission pipe 1 to dynamically monitor the water pressure in the emission pipe 1. When the water pressure deviates from the preset standard working range, it immediately sends a signal to the CPU. The CPU judges whether the sprinkler system is operating normally according to this, and issues an alarm in time in case of a failure to ensure the uniformity of the sprinkler operation. The flow sensor is arranged in the water supply pipe 3 to synchronously monitor the water flow data and feedback it to the CPU in real time. The water filling times counter is arranged at the water inlet of the water tank of the sprinkler. After the sprinkler operation is completed, if the water source needs to be replenished again, the water filling times counter will update and continue the operation.

[0083] The alarm is connected to the CPU. When an accident occurs to the sprinkler, the CPU sends an alarm signal to the alarm.

[0084] The operating component is signal-connected to the mobile signal analog conversion control module and the automatic / manual switching module. The operating component can be a mobile operating terminal such as a tablet computer or a remote control, and is used to manually perform operations such as pitching and rotating actions on the water cannon unit.

[0085] The input end of the automatic / manual switching module is signal-connected to the operating component, and the sending end of the automatic / manual switching module is signal-connected to the CPU. After receiving the signal from the operating component, the automatic / manual switching module sends it to the CPU, and the CPU sends a signal to the water cannon switching control module to control the pitching and rotating actions of the water cannon unit.

[0086] The mobile signal analog conversion control module is also signal-connected to the CPU and is used to receive the control signal of the operating component and transmit it to the CPU.

[0087] The control panel is arranged in the cab of the sprinkler and is used to receive the images and monitoring data transmitted by the CPU and send the operation instructions of the driver to the CPU for execution.

[0088] The water cannon switching control module is signal-connected to the CPU and is used to receive the action control command of the water cannon unit from the CPU and send the corresponding control command to the water cannon unit.

[0089] The vehicle-mounted camera is used to capture real-time images of the work site and send them to the CPU. The camera has auto-focus and image stabilization functions to ensure that clear and stable images are captured, and can adapt to the complex light environment of the mine.

[0090] The image recognition device is used to receive image data transmitted by the CPU for image recognition, determine the pollution situation of the work site, and send the determination result to the CPU. The image recognition device can also be used for path planning, and transmit the planned path to the CPU, and the CPU sends the planned path to the control panel in the form of a picture, and the control panel displays the planned path to the driver.

[0091] The automatic-manual switching module is operated through the control panel. If the manual mode is selected, the vertical pitch and horizontal rotation of the water cannon unit are directly controlled through the operating components. The signal of the operating components is transmitted to the CPU via the mobile signal analog conversion control module, and the CPU controls the water cannon switching control module to drive the corresponding action of the water cannon. If the automatic mode is selected, the image sensor in the image recognition device first identifies the difficulty of cleaning the area to be sprayed, the shape and size of the spray point and other data, and feeds these monitoring data back to the CPU. At the same time, the CPU automatically controls the vertical pitch and horizontal rotation of the water cannon unit after comprehensive analysis and processing based on the data obtained by the sprinkler truck monitoring system, including the pressure sensor monitoring the pressure, the liquid level sensor monitoring the liquid level, the mileage sensor monitoring the watering mileage, and the water adding times counter monitoring the water adding times, so that the spray parameters of the water cannon can be intelligently optimized and adjusted according to the actual situation of the spraying area and the vehicle's own state to achieve the best spraying effect.

[0092] In order to facilitate subsequent data analysis and vehicle maintenance management, help optimize the sprinkler truck's operation plan and resource allocation, and further improve the economy and scientificity of mine operations, an information storage device can be added to the control unit. The information storage device is connected to the CPU through signals and is used to store all data received and sent by the CPU, including data transmitted to the CPU by pressure sensors, flow sensors, liquid level sensors, mileage sensors, and image recognition devices, and also includes control commands sent to each module after the CPU processes the transmitted data.

[0093] like Figure 4 As shown, the rotation angle of the platform 8 is ±95°.

[0094] like Figure 5 The figure is a block diagram of the intelligent remote control water cannon system for a mine sprinkler truck provided by the present invention. The present invention also provides a method for using the intelligent remote control water cannon system for a mine sprinkler truck applicable to the present invention, comprising the following steps:

[0095] Step 1: During the driving of the mine sprinkler truck, use on-vehicle cameras to collect images of the mine working area from different angles, including scenes such as transportation roads, ore stacking areas, and mining operation areas.

[0096] Step 2: Transmit the images captured by the on-vehicle cameras to the image recognition device through the CPU. The image recognition device uses deep learning algorithms to learn and process the images captured by the on-vehicle cameras, and enhance the images captured by the on-vehicle cameras to obtain enhanced images.

[0097] Step 21: For the input image, that is, the image captured by the on-vehicle camera, use a matrix as the convolution kernel, use the values of the matrix as the weight parameters of the neural network, and move the convolution kernel from the upper left corner of the input image from left to right and from top to bottom with a step size of 1 to perform convolution operations to obtain a number of convolution feature matrices. The formula for the value of each element in a single convolution feature matrix is as follows:

[0098] h=f(W (1) X small +b (1) )

[0099] where X small represents the small image extracted from the large image by the convolution kernel, W (1) and b (1) represent the weight matrix and bias term corresponding to the first convolution kernel, and h is the element value at the position corresponding to X small in the first convolution feature matrix, where the function f(*) is the activation function.

[0100] In this step, for the image captured by the on-vehicle camera, use a matrix as the convolution kernel, use the values of the matrix as the weight parameters of the neural network, then move the convolution kernel from the upper left corner of the original image from left to right and from top to bottom with a step size of 1, and perform convolution operations. Finally, a convolution feature matrix can be obtained, which can correspond to the value of a neuron in the hidden layer of the fully connected structure. During the convolution operation, each convolution feature matrix corresponds to a convolution kernel, and there is weight sharing in this process. Most of the time, there are many convolution kernels in the convolution layer and multiple convolution feature matrices are output. The value of each element in a single convolution feature matrix can be expressed as: h=f(W (1) X small +b (1) ).

[0101] Therefore, for a given image of size r×c, k convolution kernels of size a×b perform convolution operations on it with a step size of 1. Finally, the convolution layer outputs k convolution features of size (r - a + 1)×(c - b + 1), where the size of the step can be changed according to the application requirements, and in some cases, operations such as edge padding are required.

[0102] It should be noted that the object to be collected, that is, the image captured by the vehicle-mounted camera, needs to be blurred, which is achieved by combining the spatial domain function and the range domain kernel function during the convolution process. A typical kernel function is the Gaussian distribution function, as shown below:

[0103]

[0104] where is for the purpose of normalization.

[0105] σ s is the standard deviation of the spatial domain Gaussian function, σ r is the standard deviation of the range domain Gaussian function, and Ω represents the domain of convolution.

[0106] Step 22: Construct the final recognition model based on the weakly supervised learning data augmentation system, the attention network, and the Mobilenet-v2 network, and obtain the augmented image for the input image. The formula is as follows:

[0107]

[0108] where f(*) is the convolution function, and A i represents a component of the recognition object.

[0109] It should be noted that in deep learning, a sufficient number of samples is a common requirement. Increasing the sample size usually can improve the training effect and generalization ability of the model. However, in actual scenarios, problems such as a lack of sample quantity or poor quality are often faced. To solve this problem, the present invention introduces a weakly supervised learning data augmentation system, which is implemented using the attention network and combines the Mobilenet-v2 network to construct the final recognition model.

[0110] Assign the feature map to the attention layer, set a threshold, and select the most important region according to the attention weights of each region. After size adjustment, it becomes a new image and is input into the network for training again. To further improve the robustness of the network, select a sub-important region and repeat the above process, and data augmentation is performed through the above process.

[0111] Step 23: Use the loss function to quantitatively evaluate the augmented image.

[0112] In some embodiments, the object recognition loss function can be used. The formula is as follows:

[0113]

[0114] where p i is the probability generated by the network, represents the label of the sample.

[0115] The target recognition loss function is used to measure the difference between the prediction result and the true label of the target recognition model when distinguishing different targets (ore piles, vehicles, pedestrians, etc.). By minimizing this loss function, the model can learn better target recognition features and improve the recognition accuracy of different targets.

[0116] In some embodiments, a target localization loss function can be used:

[0117] where is the regression target obtained by the network, is its true value.

[0118] This target localization loss function is used to measure the difference between the target position predicted by the model and the true target position. During training, by minimizing this function, the predicted target position can be made closer to the true position. For the localization of ore piles, if the deviation between the predicted position and the actual position is large, the loss value will increase, and the model will adjust the network parameters accordingly to improve the accuracy of target localization, thereby providing a more accurate basis for subsequent water cannon control and ensuring that the water cannon can accurately operate on the target.

[0119] Step 3: Identify and judge the enhanced image obtained in Step 2, determine the decontamination, dust reduction, and avoidance operations, and transmit the judgment result to the CPU.

[0120] Step 31: Compare the enhanced image obtained in Step 2 with the preset clean area image features, which include color, texture, clarity, and contrast. If there is a color deviation from the preset clean area and the texture is rough in the enhanced image, it is determined as a dirty area and decontamination operations are performed.

[0121] Step 311: Represent the dirt level with n, which is divided into three categories: n = 1 for mild dirt, n = 2 for moderate dirt, and n = 3 for severe dirt. The corresponding water volumes for the three levels are P 1 , P 2 , P 3 , where P 1 <P 2 <P 3 ;

[0122] If n = 1, then select the water volume [P 1 , P 3 ;

[0123] If n = 2, then select the water volume [P i , P j , where i < j; j ∈ [2, 3]

[0124] If n = 3, then select the water volume [P 3 .

[0125] Step 312: The spraying height level is represented by h, which are low, medium, and high respectively; the pitch angles corresponding to the three heights are θ 1 , θ 2 , θ 3 . Among them, θ 1 < θ 2 < θ 3 .

[0126] If h = 1, then select the pitch angle as [θ 1 , θ 3 ;

[0127] If h = 2, then select the pitch angle as [θ i , θ j , where i < j; j ∈ [2, 3]

[0128] If h = 3, then select the pitch angle as [θ 3 .

[0129] Step 32: Compare the enhanced image obtained in Step 2 with the preset clean area image features. The image features include color, texture, sharpness, and contrast. If the enhanced image has low sharpness, weak contrast, and a dull color, it is determined that there is a lot of dust, and a dust reduction operation is performed.

[0130] The present invention generally divides pictures into two categories, one is for decontamination and the other is for dust reduction.

[0131] By comparing the preset clean area image features (color, texture, etc.), it is determined which parts of the image have obvious stains, dirt accumulation, etc. that need to be cleaned. If the color of a certain area significantly deviates from the normal road surface color and the texture is rough, it can be determined as a dirty area.

[0132] Based on the dust concentration distribution characteristics in the image, it is determined which parts need dust reduction. When the sharpness of the image in a certain area is low, the contrast is weakened, and the color is dull, combined with the specific blurred pattern presented by the dust in the image, it can be determined that there is a lot of dust in that area and dust reduction treatment is required.

[0133] Step 33: Perform object recognition on the enhanced image obtained in Step 2, and further analyze its position, shape, and size features; for a transport vehicle, determine its position coordinates, body contour, and driving direction in the image; for an ore pile, analyze its shape, volume, and distribution range. At the same time, combined with the driving trajectory and speed of the mining sprinkler truck, predict the relative motion trend between the object and the sprinkler truck, and plan the sprinkling path in advance to avoid collision accidents.

[0134] Step 4: Based on the judgment result transmitted by the image recognition device, the CPU sends a control command to the water cannon switching control module, and the water cannon switching control module sends a corresponding control command to the water cannon unit.

[0135] When it is detected that the transport vehicle is driving in a specific area, adjust the angle of the water cannon to avoid spraying water on the vehicle; when dust suppression is required for the ore pile, control the spraying range of the water cannon to cover the surface of the ore pile and improve the dust suppression effect.

[0136] Step 5: During the operation, when the sprinkler monitoring system detects an abnormal situation, the abnormal situation includes that the pressure sensor detects that the pressure exceeds the threshold and the liquid level sensor detects that the liquid level is lower than the threshold, an alarm signal is sent to the CPU, and the CPU transmits the alarm signal to the alarm for alarm prompt; at the same time, the image recognition device continuously monitors the spraying condition. If it is found that some areas do not reach the expected spraying effect, such as there are missed sprays, uneven spraying, etc., the image recognition device will feedback this information to the CPU in real time, and the CPU will adjust the spraying parameters of the water cannon again accordingly to ensure that the entire operation area can be evenly and effectively sprayed.

[0137] Step 6: After the cleaning operation is completed, take images of the same area again. By comparing the color, texture changes of the dirty areas in the images before and after cleaning, and the reduction degree of the dust concentration, visually evaluate the cleaning effect.

[0138] In some embodiments, the mean squared error (MSE) loss function is used to evaluate the difference between the predicted cleanliness and the actual cleanliness. The mean squared error loss function is:

[0139]

[0140] y i is the actual cleanliness of a certain area in the image (quantified by features such as pixel color, texture, etc.), is the predicted cleanliness.

[0141] In some embodiments, the structural similarity (SSIM) loss function is used: to measure the structural similarity of the images before and after cleaning, and comprehensively evaluate from three aspects: brightness, contrast, and structure.

[0142] Among them, the brightness similarity function is: l(x,y) = (2 * μx * μy + C1) / (μx^2 + μy^2 + C1);

[0143] Among them, the contrast similarity function is: c(x,y) = (2 * σxy + C2) / (σx^2 + σy^2 + C2);

[0144] Among them, the structural similarity function is: s(x, y) = (σxy + C3) / (σx * σy + C3);

[0145] The SSIM calculation formula is: SSIM(x, y) = l(x, y) * c(x, y) * s(x, y).

[0146] Among them: x and y represent two images, μx and μy represent the average values of the brightness of x and y; σxy represents the covariance of the brightness of x and y, σx and σy represent the variances of the brightness of x and y; C1, C2, and C3 represent constants.

[0147] The closer the SSIM value is to 1, the better the cleaning effect.

[0148] In some embodiments, the Euclidean loss function is used:

[0149] Among them represents the prediction target of the feature points obtained from the network, represents its true value.

[0150] This function is used to evaluate the dust suppression effect during the operation of the sprinkler truck. By comparing the changes in multiple feature values related to the dust suppression effect in the images before and after sprinkling, and combining the corresponding weights, a comprehensive dust suppression effect evaluation value is calculated. If the dust concentration feature value decreases significantly after sprinkling and the object surface humidity feature value increases to an appropriate range, and the value calculated according to the weights is small, it indicates that the dust suppression effect meets the expectations; if the value is large, operations such as increasing the spraying amount or adjusting the spraying angle may be required to improve the dust suppression effect.

[0151] The present invention can more comprehensively and accurately evaluate the cleaning effect by comprehensively considering the values of these loss functions, providing a basis for subsequent parameter adjustment and optimization. It is also possible to separately consider the value of one of the loss functions to evaluate the cleaning effect.

[0152] In addition to the above embodiments, the present invention may also have other implementation manners. Any technical solutions formed by equivalent replacement or equivalent transformation fall within the protection scope required by the present invention.

Claims

1. An intelligent remote-controlled water cannon system for a mining sprinkler truck, comprising a water cannon unit and a control unit for controlling the firing of the water cannon unit, characterized in that: The water cannon unit comprises a base (9) arranged inside the water truck, a platform (8) arranged above the base (9), a water supply pipe (3) penetrating the base (9) and the platform (8) and connected to a water tank inside the water truck, a pitch mechanism (2) arranged horizontally and rotatably connected to the water supply pipe (3), a launch tube (1) fixedly connected to the pitch mechanism (2), and a second fixed frame (7) arranged above the platform (8) and fixedly connected to the platform (8), wherein the base (9) penetrates the body of the water truck and is fixedly connected to the water truck; a hydraulic cylinder (4) is arranged on the second fixed frame (7), an output end of the hydraulic cylinder (4) is fixedly connected to a telescopic rod (5), and the other end of the telescopic rod (5) is hinged to the launch tube (1); the control unit comprises: CPU, which provides computing power support and transmission of control commands for the entire system; The water sprinkler monitoring system is connected to the CPU through signals to transmit monitoring data; The automatic manual switching module has an input end connected to the operating component signal and a sending end connected to the CPU signal, and is used to receive the control signal of the operating component and transmit it to the CPU; The mobile signal analog conversion control module is connected with the CPU and the operating component through signals, and is used to receive the control signal of the operating component and transmit it to the CPU; An operating component used to manually perform pitch and rotation actions on the water monitor unit; The vehicle-mounted camera is used to capture real-time images of the work site and send them to the CPU; An image recognition device is used to receive image data transmitted by the CPU to perform image recognition, determine the pollution situation of the work site, and send the determination result to the CPU; The water cannon switching control module is connected to the CPU through signals, and is used to receive the CPU's action control commands for the water cannon unit and send corresponding control commands to the water cannon unit; The control panel is arranged in the driving cab of the sprinkler truck and is used for receiving the images and monitoring data transmitted by the CPU and sending the driver's operation instructions to the CPU.

2. The intelligent remote control water cannon system for a mine sprinkler truck according to claim 1 is characterized in that: The water cannon unit further comprises a motor (12) arranged inside the base, a first gear (11) fixedly connected to a fixed end of the motor (12), a rotating shaft (10) arranged at the bottom center of the platform (8) and fixedly connected, and a second gear (13) fixedly connected to the rotating shaft (10), wherein the output end of the motor (12) is perpendicular to the platform (8), the rotating shaft (10) is parallel to the output end of the motor (12), and the second gear (13) is meshed with the first gear (11).

3. The intelligent remote control water cannon system for a mining sprinkler truck according to claim 2 is characterized in that: The sprinkler truck monitoring system is composed of a pressure sensor, a flow sensor, a liquid level sensor, a mileage sensor, and a water adding times counter, and all are connected to the CPU through signals; the liquid level sensor is arranged in the water tank of the sprinkler truck, the mileage sensor is arranged in the cab, the pressure sensor is arranged in the transmitting tube (1), the flow sensor is arranged in the water supply pipe (3), and the water adding times counter is arranged at the water inlet of the water tank of the sprinkler truck.

4. The intelligent remote control water cannon system for a mining sprinkler truck according to claim 3 is characterized in that: The control unit also includes an alarm, and the alarm is connected to the CPU via a signal.

5. The intelligent remote control water cannon system for a mining sprinkler truck according to claim 4 is characterized in that: The control unit also includes an information memory, and the information memory is connected to the CPU via a signal.

6. The intelligent remote control water cannon system for a mining sprinkler truck according to claim 4 is characterized in that: The hinge point between the telescopic rod (5) and the launching tube (1) is far away from the pitch mechanism (2).

7. The intelligent remote control water cannon system for a mining sprinkler truck according to claim 4 is characterized in that: A first fixing frame (6) may be provided between the telescopic rod (5) and the launching tube (1).

8. A method for using the intelligent remote-controlled water cannon system for a mining sprinkler truck according to any one of claims 5 to 7, characterized in that: The steps include: S1: When the mine sprinkler truck is driving, the on-board camera is used to collect images of the mine working area from different angles; S2: The image captured by the vehicle camera is transmitted to the image recognition device through the CPU. The image recognition device uses a deep learning algorithm to learn and process the image captured by the vehicle camera, and enhances the image captured by the vehicle camera to obtain an enhanced image. S21: For the input image, the matrix is ​​used as the convolution kernel, and the value of the matrix is ​​used as the weight parameter of the neural network. The convolution kernel is moved from the upper left corner of the input image from left to right and from top to bottom with a step size of 1 to perform a convolution operation to obtain several convolution feature matrices. The formula for the value of each element in a single convolution feature matrix is ​​as follows: h=f(W (1) X small +b (1) ) Where X small Represents the small image extracted from the large image by the convolution kernel, W (1) and b (1) represents the weight matrix and bias term corresponding to the first convolution kernel, and h is the first convolution feature matrix corresponding to X small The element value of the position, f(*) is the activation function; In the convolution process, the spatial domain function and the range kernel function are combined to blur the image taken by the vehicle camera. The formula is as follows: in is the normalization effect; σ s is the standard deviation of the spatial Gaussian function, σ r is the standard deviation of the range Gaussian function, Ω represents the domain of convolution; S22: The final recognition model is constructed based on the weakly supervised learning data enhancement system, the attention network, and the Mobilenet-v2 network. The enhanced image is obtained for the input image. The formula is as follows: Where f(*) is the convolution function, A i Represents a component of an identified object; S23: Use the loss function to quantitatively evaluate the enhanced image. S3: Identify and judge the enhanced image obtained in step S2, determine the decontamination, dust reduction, and avoidance operations, and transmit the judgment result to the CPU; S31: Compare the enhanced image obtained in step S2 with the preset clean area image features, where the image features include color, texture, clarity and contrast. If the enhanced image has a color that deviates from the preset clean area and a rough texture, it is determined to be a dirty area and a decontamination operation is performed; S311: The dirt level is represented by n and divided into three categories: n = 1 light dirt, n = 2 moderate dirt, n = 3 heavy dirt; the water volumes corresponding to the three levels are P1, P2, and P3, where P1 <P2<P3; If n = 1, select the water volume [P1, P3]; If n = 2, then select the water volume [P i , P j ], where i <j;j∈[2,3]; If n = 3, select water volume [P3]; S312: The spraying height level is represented by h, which is low, medium and high respectively; the pitch angles corresponding to the three heights are θ1, θ2 and θ3; where θ1<θ2<θ3; If h = 1, the pitch angle is selected as [θ1, θ3]; If h = 2, then the pitch angle is [θ i ,θ j ], where i <j;j∈[2,3]; If h = 3, the pitch angle is selected as [θ3]; S32: Compare the enhanced image obtained in step S2 with the preset clean area image features, where the image features include color, texture, clarity and contrast. If the enhanced image has low clarity, weak contrast and dark color, it is determined that there is a lot of dust, and dust reduction operation is performed; S33: Perform object recognition on the enhanced image obtained in step S2, and further analyze its position, shape, and size characteristics; for transport vehicles, determine their position coordinates, body contours, and driving direction in the image; for ore piles, analyze their shape, volume, and distribution range, and at the same time, combine the driving trajectory and speed of the mining sprinkler truck to predict the relative movement trend between the object and the sprinkler truck, and plan the sprinkler path; S4: The CPU sends a control command to the water cannon switching control module according to the judgment result transmitted by the image recognition device, and the water cannon switching control module sends a corresponding control command to the water cannon unit; S5: During the operation, when the monitoring system of the sprinkler truck detects an abnormal situation, including the pressure sensor detecting that the pressure exceeds the threshold and the liquid level sensor detecting that the liquid level is lower than the threshold, an alarm signal is sent to the CPU, and the CPU transmits the alarm signal to the alarm for alarm prompt; at the same time, the image recognition device continuously monitors the spraying status; S6: After the cleaning operation is completed, an image of the same area is taken again; the color and texture changes of the dirty area in the image before and after cleaning, as well as the reduction degree of dust concentration are compared to evaluate the cleaning effect.

9. The method for using the intelligent remote control water cannon system for a mine sprinkler truck according to claim 8, characterized in that: The loss function for quantitatively evaluating the enhanced image in step S23 is a target recognition loss function or a target positioning loss function, and the formulas are as follows: where p i is the probability generated by the network, The label representing the sample; is the regression target obtained by the network, for its true value.

10. The method for using the intelligent remote control water cannon system for a mine sprinkler truck according to claim 8, characterized in that: In step S6, the cleaning effect is evaluated by one or more combinations of a mean square error loss function, a structural similarity loss function, and a Euclidean loss function. The mean square error loss function formula is as follows: y i is the actual cleanliness of a region in the image, The predicted degree of cleanliness; The formula of structural similarity loss function is as follows: l(x,y)=(2 * μx * μy+C1) / (μx^2+μy^2+C1) c(x,y)=(2 * σxy+C2) / (σx^2+σy^2+C2) s(x,y)=(σxy+C3) / (σx * σy+C3) SSIM(x,y)=l(x,y)*c(x,y)*s(x,y) Where: SSIM(x,y) is the structural similarity loss function; l(x,y) is the brightness similarity function; s(x,y) is the structural similarity function; c(x,y) is the contrast similarity function; x, y represent two images, μx, μy represent the average values ​​of the brightness of x, y; σxy represents the covariance of the brightness of x, y, σx, σy represent the variance of the brightness of x, y; C1, C2, C3 represent constants; The Euclidean loss function formula is as follows: in represents the prediction target of the feature points obtained from the network, Represents its true value.

Images