Plate-type integrated check gate remote control system and method based on wireless sensor

Abstract

The invention discloses a remote control system and method for a plate-type integrated check gate based on a wireless sensor, and the system comprises the plate-type integrated check gate which comprises a plate-type integrated prefabricated dam body, a gate lifting mechanism, a gate disposed in a sliding groove of the prefabricated dam body, and a mounting frame disposed at the top of the prefabricated dam body. The gate lifting mechanism is connected with the gate and the control module; the shearing mechanism is arranged on the gate and comprises a cutting knife and a power mechanism, the power mechanism drives the cutting knife to ascend, descend and rotationally cut aquatic plants along a sliding groove of the prefabricated dam body, and the power mechanism is connected with the control module; a plurality of opposite positions of the outer side walls of the two sliding grooves in the height direction are provided with notches for the cutting knives to enter and exit from the sliding grooves; the sensor assembly comprises a water level sensor, a distance sensor which is arranged on the power mechanism and is used for detecting the movement distance of the cutting knife, and a camera which is arranged on the mounting frame and is used for collecting aquatic plant images; and the control module is in wireless communication with the sensor assembly and a server of a control room.

Description

Technical Field

[0001] The present invention relates to the technology of remote control of gates, and particularly to a remote control system and method for a plate-type integrated check gate based on wireless sensors. Background Art

[0002] A gate is a control facility used to close and open a water discharge (release) channel and is an important part of a hydraulic structure. It can be used to intercept water flow, control water level, regulate flow rate, discharge sediment and floating objects, etc. Currently, in irrigation areas, the flow in the main river is mainly introduced into the irrigation area through channels / rivers with a depth of 3 to 4 meters to irrigate the irrigation area. In order to facilitate flow regulation for irrigation and flood control, etc., water retaining dams are installed at equal intervals near the confluence of the river and the main river and on the river.

[0003] Due to the interception of the water retaining dam, the water body in the river hardly flows. Coupled with the small width and depth of the river, aquatic plants in the river grow rapidly. Many aquatic plants will grow on the edge of the water retaining dam, and some climbing plants will even climb onto the dam, twine around the gate and the dam body beside the gate, giving a connecting force between the gate and the dam body. When the gate needs to be opened, the plants twined around the gate will also increase the resistance when the gate rises, thus increasing the load of the gate lifting mechanism and affecting the service life of the gate lifting mechanism. Too many plants twined around the gate will also affect the water passing after the gate is opened; furthermore, since our gates are mainly made of steel structures, plants climbing on the gate for a long time will, to a certain extent, affect the waterproof coating on the surface of the gate.

[0004] Aiming at the possible impact of weeds growing near the gate on the long-term stable operation of the gate lifting mechanism, it is urgent to develop a control system that can clean the weeds near the gate to ensure the service life of the gate lifting mechanism and the normal water passing when the gate is opened. Summary of the Invention

[0005] Aiming at the above deficiencies in the prior art, the remote control system and method for a plate-type integrated check gate based on wireless sensors provided by this solution can regularly clean it according to the growth situation of aquatic plants to reduce the resistance when the gate is opened.

[0006] To achieve the above invention purpose, the technical solution adopted by the present invention is as follows:

[0007] In the first aspect, a remote control system for a plate-type integrated check gate based on wireless sensors is provided, which includes:

[0008] A plate-type integrated check gate, which includes a plate-type integrated precast dam body, a gate lifting mechanism, a gate arranged in the chute of the precast dam body, and an installation frame arranged on the top of the precast dam body. The gate lifting mechanism is connected to the gate and the control module.

[0009] A shearing mechanism is provided on the gate, which includes a cutting knife and a power mechanism that drives the cutting knife to lift and rotate along the chute of the prefabricated dam to cut aquatic plants. The power mechanism is connected to the control module. At multiple relative positions in the height direction on the outer side walls of the two chutes, there are notches for the cutting knife to enter and exit the chute.

[0010] A sensor assembly, which includes a water level sensor, a distance sensor arranged on the power mechanism to detect the movement distance of the cutting knife, and a camera arranged on the mounting frame to collect images of aquatic plants.

[0011] A control module communicates wirelessly with the sensor assembly and the server in the control room. The control module is used to control the gate lifting mechanism to open and close the gate, and when the gate is opened and / or at every preset time interval, and the river water level is higher than the preset threshold, it starts the camera to collect images of the aquatic plants in front of the gate, enhances the images and then inputs them into the convolutional neural network to obtain the height and density of the aquatic plants. When any one of the height and density values meets its corresponding threshold, it starts the power mechanism to rotate according to the preset conditions, so that the cutting knife makes a rotational movement at the notch and a linear lifting movement at the non-notch section of the chute to remove the aquatic plants, and then resets the cutting knife.

[0012] Further, the power mechanism includes a rotary motor fixed on the gate. The output shaft of the rotary motor is fixedly connected to a lead screw extending to the bottom edge of the gate. There is a lead screw nut on the lead screw, and the lead screw nut is detachably connected to the cutting knife. The distance sensor is installed on the end face of the rotary motor on the side where the output shaft is located and is directly above the lead screw nut.

[0013] Further, it is set that the cutting knife is initially located at the uppermost end of the right chute. The rotation of the rotary motor towards the outer side wall of the right chute is defined as forward rotation, and the cutting knife moves downward during forward rotation. At this time, when starting the power mechanism to rotate according to the preset conditions, the method for removing aquatic plants includes:

[0014] A1. Control the rotary motor to rotate forward, and the lead screw nut drives the cutting blade to descend along the right chute.

[0015] A2. When the cutting knife moves to the notch, it slides out of the notch and makes a rotational movement to cut the aquatic plants. After the cutting knife rotates into the left chute, it descends along the left chute.

[0016] A3. When the distance sensor detects that the cutting knife moves to the notch, control the rotary motor to rotate in reverse, so that the cutting knife slides out of the notch and makes a rotational movement to cut the aquatic plants.

[0017] A4. When it moves into the notch of the right chute, control the rotary motor to rotate forward, and the cutting knife rotates back into the left chute and descends along the left chute.

[0018] A5. Repeat steps A3 and A4 until the cutting knife moves to the lowest end of the chute, completing the removal of aquatic plants in front of the gate.

[0019] A6. Control the rotating motor to bring the cutting knife into the right chute, and then the rotating motor reverses to drive the cutting knife to rise and return to the initial position.

[0020] Further, the cutting knife includes a rotating shaft, on which a first blade perpendicular to the gate, a second blade parallel to the gate, and a third blade are installed; one end of the rotating shaft is detachably connected to the screw nut, and the other end is installed with a pulley that slidably cooperates with the chute through a connecting member; when the first blade rotates at different heights, it cuts the aquatic plants into sections to reduce the height of the aquatic plants; when the second blade descends, it cuts the aquatic plants adjacent to the gate; when the third blade ascends, it cuts the aquatic plants adjacent to the gate.

[0021] Further, pressure sensors that communicate with the control module wirelessly are arranged on both end faces of the pulley; when the shearing mechanism is started, when the control module detects that the change amount of the pressure collected by the pressure sensor is greater than the preset pressure, it indicates that the pulley has entered the chute through the notch; when the control module detects that the river water level collected by the water level sensor is greater than the water level safety height warning line, the gate is opened to release water until the river water level is lower than the water level safety height warning line.

[0022] Further, the expression for enhancing the aquatic plant image is:

[0023]

[0024] T(u, v) = e -βd(u,v)

[0025] Among them, is the enhanced aquatic plant image; u and v are the frequencies in the x and y directions in the image space respectively; H(u, v) is the degradation function; H * (u, v) is the conjugate complex number of H(u, v); is the local variance of the unenhanced aquatic plant image P; α is the adaptive adjustment parameter; β is the turbidity coefficient; W f (u, v) is the wavelet transform coefficient; λ is the regularization parameter; R(u, v) is the gradient of P; γ is the reflection intensity coefficient; G(u, v) is the Fourier transform of P; T(u, v) is the turbidity compensation factor; e is the natural logarithm; d(u, v) is the light propagation distance; |·| is to take the absolute value.

[0026] Further, the expression of the degradation function is:

[0027]

[0028] Among them, both τ and a are fuzzy parameters obtained through calibration tests; j is an imaginary number.

[0029] Furthermore, the convolutional neural network is a ResNet-50 pre-trained model, which includes a density prediction branch and a height prediction branch. Both prediction branches include a 3×3 convolutional layer, a 1×1 convolutional layer, and an output layer. The output layer is a fully connected layer + softmax layer. The training method of the convolutional neural network includes:

[0030] B1. Take the height from the camera on the mounting bracket to the lowest end of the precast dam as the installation height Ha of the camera during dataset collection.

[0031] B2. Set up a camera at a height of Ha from the bottom of the test water tank, and then plant various aquatic plants in the water tank below it.

[0032] B3. At different planting densities, adjust the height of the aquatic plants from the water surface and above the water surface, and use the camera to collect a number of aquatic plant images under different light intensities and different water turbidities.

[0033] B4. At each planting density, plant aquatic plants with different heights, and use the camera to collect a number of aquatic plant images under different light intensities and different water turbidities.

[0034] B5. Perform image enhancement processing on all the collected aquatic plant images, and then use all the images after image enhancement as the dataset, and divide them into a training set and a validation set according to a ratio of 7:3.

[0035] B6. Use the training set and the validation set to train the ResNet-50 pre-trained model until the model converges to obtain the trained convolutional neural network.

[0036] Furthermore, the expression of the loss function of the convolutional neural network is:

[0037] L z =α1L d +β1L h ,

[0038]

[0039] Among them, L z is the loss function of the convolutional neural network; L d is the density loss; L h is the height loss; both α1 and β1 are uncertainty adjustment coefficients; σ d and σ h are both learnable noise parameters; N is the number of enhanced images in the training batch. is the predicted density of the i-th enhanced image; is the true density of the i-th enhanced image; TV(·) is the total variation regularization term; is the predicted height of the i-th enhanced image; is the true height of the i-th enhanced image; SmoothL1(·) is the smooth L1 loss.

[0040] In a second aspect, a control method applied to a plate-type integrated sluice remote control system based on wireless sensors is provided. When the plate-type integrated sluice remote control system is started, it includes the steps:

[0041] S1. Receive the water level information uploaded by the water level sensor, and determine whether the current water level is greater than the water level safety height warning line. If so, go to step S8; otherwise, go to step S2.

[0042] S2. Determine whether the current water level is greater than the preset threshold and the current time is an integer multiple of the preset duration. If so, go to step S3; otherwise, return to step S1.

[0043] S3. Start the camera to collect the image of aquatic plants in front of the sluice, and perform image enhancement processing on the image of aquatic plants.

[0044] S4. Input the image after image enhancement processing into the convolutional neural network to predict the height and density of the aquatic plants.

[0045] S5. Determine whether any value of the height and density of the aquatic plants meets its corresponding threshold. If so, go to step S6; otherwise, return to step S1.

[0046] S6. Start the power mechanism to rotate according to the preset conditions, so that the cutting knife makes a rotational motion at the notch and a linear lifting motion at the non-notch section of the chute to remove the aquatic plants.

[0047] S7. Reset the cutting knife and return to step S1.

[0048] S8. After performing steps S3 and S4, determine whether any value of the height and density of the aquatic plants meets its corresponding threshold. If so, perform step S6, reset the cutting knife, and go to step S9; otherwise, directly go to step S9.

[0049] S9. Open the sluice to release water until the river water level is lower than the water level safety height warning line, and then return to step S1.

[0050] The beneficial effects of the present invention are as follows: By setting up a shearing mechanism and a camera, this solution can regularly photograph the aquatic plants in the river channel. When the plants grow too lush or too tall, the shearing mechanism is used to clean the aquatic plants, so as to ensure that the arc area in front of the gate is in a state of having no or only a small amount of aquatic plants, keeping the aquatic plants away from the gate, and avoiding the situation where the aquatic plants entangle the gate and increase the resistance when the gate is opened later, thus ensuring the normal operation of the gate opening.

[0051] The prefabricated water retaining gate of this solution can be prefabricated in a factory using steel structures and directly transported to the river channel. It can be installed by short-term river interception. Compared with the prior art where the river is cut off on-site and a dam is formed by concrete pouring, the construction period can be significantly shortened, and the normal operation of the river channel is less affected.

[0052] The control module communicates with the server in the control room. When water needs to be dispatched for irrigation downstream, relevant personnel can remotely control through the APP on the mobile phone and the server, improving the convenience of gate operation; the camera of this solution can also, as needed and under the control of the control module, only collect images of aquatic plants when collecting aquatic plants, and when not collecting, it can collect images around the gate to play a monitoring role for the purpose of monitoring the safety of the gate. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic block diagram of a remote control system for a plate-type integrated check gate based on a wireless sensor.

[0054] Figure 2 It is a front view of the prefabricated water retaining gate after installing the shearing mechanism.

[0055] Figure 3 It is a front view of the shearing mechanism installed on the gate.

[0056] Figure 4 It is Figure 3 An enlarged view of part A in

[0057] Figure 5 It is a side view formed by the projection of the cutting knife from the side of the screw nut to the side of the sliding groove.

[0058] Figure 6 It is a front view of the cutting knife.

[0059] Figure 7 It is a flow chart of a remote control method for a gate based on a wireless sensor network.

[0060] Among them, 1. Plate - type integrated check gate; 11. Prefabricated dam body; 111. Slide groove; 1111. Notch; 12. Gate lifting mechanism; 13. Gate; 14. Installation frame; 2. Shearing mechanism; 21. Cutting knife; 211. Rotating shaft; 212. First blade; 213. Second blade; 214. Third blade; 215. Pulley; 22. Power mechanism; 221. Rotating motor; 222. Lead screw; 223. Lead screw nut; 3. Sensor assembly; 31. Distance sensor; 32. Camera; 4. Control module. Specific implementation mode

[0061] The specific implementation mode of the present invention will be described below to facilitate those skilled in the art of this technical field to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation mode. For those ordinary skilled in the art of this technical field, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are within the scope of protection.

[0062] The remote control system of the gate 13 in this solution is mainly applied to small - sized river channels in irrigation areas with a width of only a few meters. Since the depth of such river channels is often not large, the water flow is not very rapid, and in addition, because they are located in irrigation areas, fertilizers and the like will enter the river channels along with groundwater, and the nutrients in the river channels are relatively rich, and it is extremely easy to have a situation of overgrown weeds in the river channels.

[0063] As Figure 1 shown, the remote control system of the gate 13 based on the wireless sensor network provided by this solution includes a prefabricated water - retaining gate 1, a shearing mechanism 2, a sensor assembly 3, and a control module 4. The detailed structure of each component is described as follows:

[0064] The plate - type integrated check gate 1 includes a plate - type integrated prefabricated dam body 11, a gate lifting mechanism 12, a gate 13 arranged in the slide groove 111 of the prefabricated dam body 11, and an installation frame 14 arranged on the top of the prefabricated dam body 11. The installation frame 14 includes a guardrail structure and a frame structure for installing the gate lifting mechanism 12. The gate lifting mechanism 12 is connected to the gate 13 and the control module 4. The gate lifting mechanism 12 in this solution is preferably a lead - screw nut 223 structure. A plurality of lead - screw nuts 223 are fixed on the gate 13 to ensure the stability of the gate 13 during lifting.

[0065] After the plate - type integrated check gate 1 in this solution is produced in the factory, it is transported to the site for installation, which can achieve batch production and has a short production cycle. When installing, by intercepting the river channel for a short time and digging a foundation pit, installation can be carried out. The construction cycle is short, and it often takes less than one day to complete the installation, and the impact on the water flow passage of the river channel is also very small.

[0066] The shearing mechanism 2 is arranged on the gate 13 and includes a cutting knife 21 and a power mechanism 22 that drives the cutting knife 21 to lift and rotate along the chute 111 of the precast dam body 11 to cut aquatic plants. The power mechanism 22 is connected to the control module 4. Notches 1111 for the cutting knife 21 to enter and exit the chute 111 are arranged at multiple relative positions in the height direction on the outer side walls of the two chutes 111.

[0067] The sensor assembly 3 includes a water level sensor, a distance sensor 31 arranged on the power mechanism 22 for detecting the movement distance of the cutting knife 21, and a camera 3 arranged on the mounting frame 14 for collecting images of aquatic plants. The camera 3 can also be replaced by a rotatable camera 3, so that plant images can be collected when needed, and when not needed, images around the gate 13 can be collected to play a monitoring role for the purpose of monitoring the safety of the gate 13.

[0068] The control module 4 communicates wirelessly with the sensor assembly 3 and the server in the control room. The control module 4 is used to control the gate lifting mechanism 12 to open and close the gate 13, and when the gate 13 is opened and / or at every preset time interval, and when the river water level is higher than the preset threshold, it starts the camera 3 to collect images of aquatic plants in front of the gate 13, enhances the images and then inputs them into the convolutional neural network to obtain the height and density of the aquatic plants.

[0069] During implementation, the preferred expression for enhancing the aquatic plant image in this solution is:

[0070]

[0071] T(u,v) = e -βd(u,v)

[0072] where is the enhanced aquatic plant image; u and v are the frequencies in the x and y directions in the image space respectively; H(u,v) is the degradation function; H * (u,v) is the conjugate complex number of H(u,v); is the local variance of the unenhanced aquatic plant image P; α is the adaptive adjustment parameter; β is the turbidity coefficient, obtained by water sampling or preset empirical value (such as β≈0.1m for clear water -1 , β>1m for turbid water -1 ); W f (u,v) is the wavelet transform coefficient; λ is the regularization parameter; R(u,v) is the gradient of P; γ is the reflection intensity coefficient; G(u,v) is the Fourier transform of P; T(u,v) is the turbidity compensation factor; e is the natural logarithm; d(u,v) is the light propagation distance, calculated according to the installation height and viewing angle of the camera 3; |·| is to take the absolute value.

[0073] This solution targets the characteristics of aquatic plant images in water. The improved model can more accurately restore aquatic plant images through water body reflection modeling, turbidity compensation, and illumination correction. This method is applicable to image restoration tasks in complex water body environments and can provide high-quality input data for subsequent aquatic plant recognition.

[0074] When any one of the height and density values meets its corresponding threshold, the power mechanism 22 is activated to rotate according to preset conditions, causing the cutting knife 21 to perform a rotational motion at the notch 1111 and a linear lifting motion at the non-notch 1111 section of the chute 111 to remove aquatic plants, and then resetting the cutting knife 21.

[0075] Before collecting aquatic plant images in this solution, the water level of the river channel is considered because when the water in the river channel is too shallow or there is no water, there is basically no water discharge in this case, the water volume is small, and the growth of aquatic plants is not very lush. Therefore, the growth situation of plants does not need to be considered temporarily, that is, there is no need to collect aquatic plant images.

[0076] In addition, due to the influence of illumination and the turbidity of the water in the river channel during photographing, the clarity of plants in the captured photos will be affected, which has a great impact on the accurate recognition of the subsequent neural network. To address this, this solution can improve the recognition accuracy of the model through image enhancement.

[0077] Such as Figure 2 and Figure 3 As shown, the power mechanism 22 includes a rotary motor 221 fixed to the gate 13. The output shaft of the rotary motor 221 is fixedly connected to a lead screw 222 extending to the bottom edge of the gate 13. A lead screw nut 223 is arranged on the lead screw 222, and the lead screw nut 223 is detachably connected to the cutting knife 21. The distance sensor 31 is installed on the end face of the rotary motor 221 on the side where the output shaft is located and is directly above the lead screw nut 223.

[0078] This solution uses a lead screw nut structure as the power part. When the rotary motor 221 rotates, if the cutting knife 21 is blocked by the structure, the lead screw nut 223 can drive the cutting knife 21 to perform a vertical lifting motion. If the constraint of this structure is lost, the lead screw nut 223 will rotate with the lead screw 222. Based on this principle, a notch 1111 is provided on the chute 111, and different motion modes are presented by the non-notch 1111 section and the notch 1111 section of the chute 111, so that a set of power systems can both lift and rotate the cutting knife 21 to cut aquatic plants at different heights and gradually reduce the height of aquatic plants.

[0079] In an embodiment of the present invention, to facilitate understanding of how the shearing mechanism 2 realizes the shearing of aquatic plants, this solution gives a detailed implementation process of the movement of the shearing mechanism 2:

[0080] First, set the cutting knife 21 at the uppermost end of the right chute 111 at the initial moment. The rotation motor 221 rotates in the direction of the outer wall of the right chute 111, and the forward rotation makes the cutting knife 21 move downward; at this time, start the power mechanism 22 to rotate according to preset conditions. The method for removing aquatic plants includes:

[0081] A1. Control the forward rotation of the rotation motor 221, and the screw nut 223 drives the cutting blade to descend along the right chute 111;

[0082] A2. When the cutting knife 21 moves to the notch 1111, it slides out of the notch 1111 to make a rotational movement to cut the aquatic plants. After the cutting knife 21 rotates into the left chute 111, it descends along the left chute 111;

[0083] A3. When the distance sensor 31 detects that the cutting knife 21 moves to the notch 1111, control the reverse rotation of the rotation motor 221 to make the cutting knife 21 slide out of the notch 1111 and make a rotational movement to cut the aquatic plants;

[0084] A4. When it moves into the notch 1111 of the right chute 111, control the forward rotation of the rotation motor 221, and the cutting knife 21 rotates back into the left chute 111 and descends along the left chute 111;

[0085] A5. Repeat steps A3 and A4 until the cutting knife 21 moves to the lowest end of the chute 111, completing the removal of the aquatic plants in front of the gate 13;

[0086] A6. Control the rotation motor 221 to drive the cutting knife 21 into the right chute 111, and then the rotation motor 221 rotates in reverse to drive the cutting knife 21 to rise and return to the initial position.

[0087] The distance sensor 31 can pre-store the distance between the distance sensor 31 and the screw nut at each notch in the control module by detecting the distance between it and the screw nut. When the distance detected by the distance sensor 31 is the information stored in the control module, it indicates that the cutting knife 21 moves to the notch at this time.

[0088] Such as Figures 4 to 6As shown in the figure, the cutting knife 21 provided by this solution includes a rotating shaft 211. A first blade 212 perpendicular to the gate 13, a second blade 213 parallel to the gate 13, and a third blade 214 are installed on the rotating shaft 211. One end of the rotating shaft 211 is detachably connected to the screw nut 223, and a pulley 215 that is slidably engaged with the chute 111 is installed at the other end through a connecting member. When the first blade 212 rotates at different heights, it cuts the aquatic plants into sections to reduce the height of the aquatic plants. When the second blade 213 descends, it cuts the aquatic plants adjacent to the gate 13. When the third blade 214 ascends, it cuts the aquatic plants adjacent to the gate 13.

[0089] After the cutting knife 21 of this solution is provided with 3 blades, it can remove the aquatic plants attached to or adjacent to the surface of the gate 13 when ascending and descending, and can remove the aquatic plants in the arc area in front of the gate 13 when rotating, avoiding the growth of these plants towards the gate 13. One end of the rotating shaft 211 of this solution is detachably connected to the screw nut 223, which can facilitate the replacement of the cutting knife 21 after it is damaged.

[0090] In an embodiment of the present invention, pressure sensors that communicate with the control module 4 wirelessly are provided on both end faces of the pulley 215. After the shearing mechanism 2 is started, when the control module 4 detects that the change amount of the pressure collected by the pressure sensor is greater than the preset pressure, it indicates that the pulley 215 has entered the chute 111 through the notch 1111. When the control module 4 detects that the river water level collected by the water level sensor is greater than the water level safety height warning line, the gate 13 is opened to release water until the river water level is lower than the water level safety height warning line.

[0091] During implementation, the expression of the degradation function preferably used in this solution is:

[0092]

[0093] Among them, both τ and a are fuzzy parameters obtained through calibration tests; j is an imaginary number.

[0094] In image restoration, the degradation function H(u, v) is used to describe the physical process of image degradation (such as blur, noise, light attenuation, etc.). In this solution, for the weed images in the water environment, the degradation function additionally considers the influence of water body characteristics (such as turbidity, reflection, refraction), which can further improve the effect of image enhancement.

[0095] In an embodiment of the present invention, the convolutional neural network is a ResNet-50 pre-trained model, which includes a density prediction branch and a height prediction branch. Both prediction branches include a 3×3 convolutional layer, a 1×1 convolutional layer, and an output layer. The output layer is a fully connected layer + softmax layer. The training method of the convolutional neural network includes:

[0096] B1. Take the height from the camera 3 on the mounting bracket 14 to the lowest end of the precast dam body 11 as the installation height Ha of the camera 3 during dataset collection.

[0097] B2. Install the camera 3 at a height of Ha from the bottom of the test water tank, and then plant various aquatic plants in the water tank below it.

[0098] B3. At different planting densities, adjust the height of the aquatic plants from the water surface and above the water surface, and use the camera 3 to collect a number of aquatic plant images under different light intensities and different water turbidities.

[0099] B4. At each planting density, plant aquatic plants of different heights, and use the camera 3 to collect a number of aquatic plant images under different light intensities and different water turbidities.

[0100] B5. Perform image enhancement processing on all the collected aquatic plant images, and then use all the images after image enhancement as the dataset, and divide them into a training set and a validation set according to a ratio of 7:3.

[0101] B6. Use the training set and the validation set to train the ResNet-50 pre-trained model until the model converges to obtain a trained convolutional neural network.

[0102] In this solution, when collecting the dataset, the fixed position of the camera 3 is the same as the layout height in actual application, which can make the resolution and clarity of the images used in training close to the actual situation. In this way, when the trained model recognizes the images of the actual working conditions, the recognition accuracy is higher.

[0103] During implementation, the expression of the loss function of the convolutional neural network in this solution is preferably:

[0104] L z = α1L d + β1L h ,

[0105]

[0106] where L z is the loss function of the convolutional neural network; L d is the density loss; L h is the height loss; both α1 and β1 are uncertainty adjustment coefficients; σ d and σ h are both learnable noise parameters; N is the number of enhanced images in the training batch; is the predicted density of the i-th enhanced image; is the true density of the i-th enhanced image; TV(·) is the total variation regularization term, which is used to suppress noise and enhance spatial smoothness; is the predicted height of the i-th enhanced image; is the true height of the i-th enhanced image; SmoothL1(·) is the smooth L1 loss, which is used to balance the influence of outliers.

[0107] The density loss of this scheme can constrain the distribution consistency between the density map and the true value, and the height loss can constrain the accuracy of height prediction while suppressing outliers; by designing the above adaptive multi-task loss function, combining the characteristics of density and height tasks, and introducing regularization constraints, the model accuracy is improved, and the accuracy of model recognition is further improved.

[0108] As Figure 7 shown, this scheme also provides a control method applied to the remote control system of the gate 13 based on the wireless sensor network. When the remote control system of the gate 13 is started, this method S includes steps S1 to S9.

[0109] In step S1, the water level information uploaded by the water level sensor is received, and it is judged whether the current water level is greater than the water level safety height warning line. If so, go to step S8; otherwise, go to step S2;

[0110] In step S2, it is judged whether the current water level is greater than the preset threshold and the current time is an integer multiple of the preset duration. If so, go to step S3; otherwise, return to step S1;

[0111] In step S3, the camera 3 is started to collect the image of aquatic plants in front of the gate 13, and the image of aquatic plants is subjected to image enhancement processing;

[0112] In step S4, the image after image enhancement processing is input into the convolutional neural network, and the height and density of the aquatic plants are predicted;

[0113] In step S5, it is judged whether any value of the height and density of the aquatic plants satisfies its corresponding threshold. If so, go to step S6; otherwise, return to step S1;

[0114] In step S6, the power mechanism 22 is started to rotate according to the preset conditions, so that the cutting knife 21 makes a rotational motion at the notch 1111 and makes a linear lifting motion at the non-notch 1111 section of the chute 111 to remove the aquatic plants;

[0115] In step S7, the cutting knife 21 is reset and returns to step S1;

[0116] In step S8, after performing steps S3 and S4, it is determined whether any value of the height and density of the aquatic plants meets its corresponding threshold. If so, after performing step S6, the cutting knife 21 is reset and step S9 is entered; otherwise, step S9 is directly entered.

[0117] In step S9, the gate 13 is opened to release water until the river water level is lower than the water level safety height warning line, and then the process returns to step S1.

[0118] When this control method is implemented, first, the water level of the river is confirmed through the water level sensor. When the water in the river surges, after evaluating the situation of the aquatic plants, the gate 13 is automatically opened to achieve the purpose of flood discharge; by comparing the water level with the preset threshold, it can be evaluated whether the water level in the river is too shallow. When it is too shallow, there is no need for flood discharge and there is no water available for scheduling. When the water volume is relatively small, the growth of the aquatic plants will also be greatly affected. At this time, the gate 13 will not be opened temporarily, so there is no need to collect the graphics of the aquatic plants in the river.

[0119] After the remote control system of the gate 13 in this solution adopts the above control method, in the case of system startup, it can automatically clean the water level of the river and the water plants in front of the gate 13, and can be remotely controlled through the server when water use scheduling is required.

Claims

1. A remote control system for a plate integrated sluice based on wireless sensors, characterized in that Comprising: A plate - type integrated sluice gate, which includes a plate - type integrated precast dam body, a gate lifting mechanism, a gate arranged in the chute of the precast dam body, and a mounting frame arranged at the top of the precast dam body. The gate lifting mechanism is connected to the gate and the control module; A shearing mechanism, which is arranged on the gate, includes a cutting knife and a power mechanism that drives the cutting knife to lift and rotate along the chute of the precast dam body to cut aquatic plants. The power mechanism is connected to the control module; Notches for the cutting knife to enter and exit the chute are arranged at multiple relative positions in the height direction on the outer side walls of the two chutes; A sensor assembly, which includes a water - level sensor, a distance sensor arranged on the power mechanism for detecting the movement distance of the cutting knife, and a camera arranged on the mounting frame for collecting images of aquatic plants; A control module, which communicates wirelessly with the sensor assembly and the server in the control room; The control module is used to control the gate lifting mechanism to open and close the gate, and when the gate is opened and / or at every preset time interval, and when the river water level is higher than a preset threshold, start the camera to collect images of aquatic plants in front of the sluice gate, perform image enhancement and then input them into a convolutional neural network to obtain the height and density of the aquatic plants. When any one of the height and density values meets its corresponding threshold, start the power mechanism to rotate according to preset conditions, so that the cutting knife makes a rotational movement at the notch and a linear lifting movement in the non - notch section of the chute to remove the aquatic plants, and then reset the cutting knife.

2. The plate-type integrated check gate remote control system according to claim 1, characterized in that The power mechanism includes a rotary motor fixed to the gate. The output shaft of the rotary motor is fixedly connected to a lead screw extending to the bottom edge of the gate. A lead - screw nut is arranged on the lead screw, and the lead - screw nut is detachably connected to the cutting knife; The distance sensor is installed on the end face of the rotary motor on the side where the output shaft is located and is directly above the lead - screw nut.

3. The remote control system for the plate integrated check gate according to claim 2, characterized in that, The cutting knife is initially located at the uppermost end of the right chute. The rotation of the rotary motor towards the outer side wall of the right chute is defined as forward rotation, and the cutting knife moves downward during forward rotation; At this time, when starting the power mechanism to rotate according to preset conditions, the method for removing aquatic plants includes: A1. Control the rotary motor to rotate forward, and the lead - screw nut drives the cutting blade to descend along the right chute; A2. When the cutting knife moves to the notch, it slides out of the notch and makes a rotational movement to cut the aquatic plants. After the cutting knife rotates into the left chute, it descends along the left chute; A3. When the distance sensor detects that the cutting knife moves to the notch, control the rotary motor to rotate in reverse, so that the cutting knife slides out of the notch and makes a rotational movement to cut the aquatic plants; A4. When moving into the notch of the right chute, control the rotary motor to rotate forward, and the cutting knife rotates back into the left chute and descends along the left chute; A5. Repeat steps A3 and A4 until the cutting knife moves to the lowest end of the chute, completing the removal of aquatic plants in front of the sluice gate; A6. Control the rotary motor to drive the cutting knife into the right chute, and then the rotary motor rotates in reverse to drive the cutting knife to rise and return to the initial position.

4. The remote control system for the plate integrated check gate according to claim 2, characterized in that, The cutting knife includes a rotating shaft, on which a first blade perpendicular to the gate, a second blade parallel to the gate, and a third blade are installed; one end of the rotating shaft is detachably connected to a screw nut, and the other end is installed with a pulley slidably engaged with a chute through a connecting member; when the first blade rotates at different heights, it cuts the aquatic plants into sections to reduce the height of the aquatic plants; when the second blade descends, it cuts the aquatic plants adjacent to the gate; when the third blade ascends, it cuts the aquatic plants adjacent to the gate.

5. The plate-type integrated check gate remote control system according to claim 4, characterized in that, Pressure sensors that communicate wirelessly with the control module are provided on both end faces of the pulley; when the shearing mechanism is started, when the control module detects that the change amount of the pressure collected by the pressure sensors is greater than the preset pressure, it indicates that the pulley has entered the chute through the notch; when the control module detects that the water level of the river channel collected by the water level sensor is greater than the water level safety height warning line, the gate is opened to release water until the water level of the river channel is lower than the water level safety height warning line.

6. The plate-type integrated check gate remote control system according to claim 1, characterized in that, The expression for enhancing the aquatic plant image is: T(u, v) = e -βd(u,v) Among them, is the enhanced aquatic plant image; u and v are the frequencies in the x and y directions in the image space; H(u, v) is the degradation function; H * (u, v) is the conjugate complex number of H(u, v); is the local variance of the unenhanced aquatic plant image P; α is the adaptive adjustment parameter; β is the turbidity coefficient; W f (u, v) is the wavelet transform coefficient; λ is the regularization parameter; R(u, v) is the gradient of P; γ is the reflection intensity coefficient; G(u, v) is the Fourier transform of P; T(u, v) is the turbidity compensation factor; e is the natural logarithm; d(u, v) is the light propagation distance; |·| is to take the absolute value.

7. The remote control system for the plate-type integrated check gate according to claim 6, wherein, The expression for the degradation function is: Among them, both τ and a are fuzzy parameters obtained through calibration tests; j is an imaginary number.

8. The remote control system for the plate integrated check gate according to claim 1 or 6, characterized in that, The convolutional neural network is a ResNet-50 pre-trained model, which includes a density prediction branch and a height prediction branch. Both prediction branches include a 3×3 convolutional layer, a 1×1 convolutional layer, and an output layer. The output layer is a fully connected layer + softmax layer; the training method of the convolutional neural network includes: B1. Take the height from the camera on the mounting frame to the lowest end of the precast dam body as the mounting height Ha of the camera during dataset collection. B2. Set a camera at a height of Ha from the bottom of the test water tank, and then plant various aquatic plants in the water tank below it. B3. At different planting densities, adjust the height of the aquatic plants above and below the water surface, and use the camera to collect a number of aquatic plant images under different light intensities and different water turbidities. B4. At each planting density, plant aquatic plants with different heights, and use the camera to collect a number of aquatic plant images under different light intensities and different water turbidities. B5. Perform image enhancement processing on all the collected aquatic plant images, and then use all the images after image enhancement as the dataset, and divide them into a training set and a validation set according to a ratio of 7:

3. B6. Use the training set and the validation set to train the ResNet-50 pre-trained model until the model converges to obtain the trained convolutional neural network.

9. The remote control system for the plate integrated check gate according to claim 7, characterized in that, The expression for the loss function of the convolutional neural network is: L z = α1L d + β1L h , Among them, L z is the loss function of the convolutional neural network; L d is the density loss; L h is the height loss; α1 and β1 are both uncertainty adjustment coefficients; σ d and σ h are both learnable noise parameters; N is the number of augmented images in the training batch; is the predicted density of the i-th augmented image; is the true density of the i-th augmented image; TV(·) is the total variation regularization term; is the predicted height of the i-th augmented image; is the true height of the i-th augmented image; SmoothL1(·) is the smooth L1 loss.

10. A control method applied to the remote control system of the plate-type integrated sluice based on wireless sensors according to any one of claims 1-9, characterized in that, When the remote control system of the plate-type integrated check gate is started, it includes the steps: S1. Receive the water level information uploaded by the water level sensor, and judge whether the current water level is greater than the water level safety height warning line. If so, enter step S8, otherwise enter step S2. S2. Judge whether the current water level is greater than the preset threshold and the current time is an integer multiple of the preset duration. If so, enter step S3, otherwise return to step S1. S3. Start the camera to collect the aquatic plant images in front of the gate, and perform image enhancement processing on the aquatic plant images. S4. Input the image after image enhancement processing into a convolutional neural network to predict the height and density of aquatic plants; S5. Determine whether any value in the height and density of aquatic plants meets its corresponding threshold. If so, proceed to step S6; otherwise, return to step S1; S6. Start the power mechanism to rotate according to preset conditions, so that the cutting knife makes a rotational motion at the notch and a linear lifting motion at the non-notch section of the chute to remove aquatic plants; S7. Reset the cutting knife and return to step S1; S8. After executing steps S3 and S4, determine whether any value in the height and density of aquatic plants meets its corresponding threshold. If so, execute step S6, reset the cutting knife, and proceed to step S9; otherwise, directly proceed to step S9; S9. Open the gate to release water until the river water level is lower than the warning line of the safe water level height, and then return to step S1.

Images