A method for measuring gravel coverage on the surface of coarse-grained soil during rainfall

Through multi-camera acquisition and CGAN network processing, combined with the maximum inter-class variance method and bucket-shaped shell device, the problems of uneven distribution of rainwater in rainfall and overheating of equipment are solved, and the accurate determination of the gravel coverage rate on the surface of coarse-grain soil is achieved and Continuous work of the equipment.

CN114677358BInactive Publication Date: 2024-11-08HUAZHONG AGRI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210359164.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-07
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the process of simulating rainfall, the camera equipment hinders the rainwater drop, resulting in uneven distribution of rainwater, affecting the accuracy of measurement, and the equipment is prone to overheating and difficult to work continuously for a long time.

Method used

Multiple cameras are used to collect soil surface pictures, and the rain fog treatment and lens correction are performed through the CGAN network. The maximum inter-class variance method is used to separate gravel and soil, calculate coverage, and design a device including bucket-shaped shell and blowing components to prevent Rain is dead corners and overheating, ensuring continuous shooting.

Benefits of technology

The precise determination of the gravel coverage rate of coarse-grained soil surface during simulated rainfall is achieved, avoiding the dead end of rainwater, ensuring the continuous working state of the equipment, and the simulated rainfall effect is close to real rainfall.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114677358B_ABST
    Figure CN114677358B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of soil simulation technology, and in particular to a method for measuring the gravel coverage rate on the surface of coarse-grained soil during rainfall. A device for measuring the gravel coverage rate on the surface of coarse-grained soil during rainfall is also proposed. The device of the present invention can adaptively measure the gravel coverage rate on the surface of coarse-grained soil during rainfall. In the process of simulating rainfall, it can prevent the occurrence of rainfall dead corners during rain simulation, ensure comprehensive coverage of rainwater, and the normal operation of the camera equipment will not be affected by overheating during continuous shooting, thereby ensuring the continuous progress of the measurement.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of soil simulation, and in particular to a method for determining gravel coverage on the surface of coarse-grained soil during rainfall. Background Art

[0002] Soil erosion is a dynamic and complex process. During the erosion process, soil or other ground components are denuded, destroyed, separated, transported and deposited under the action of external forces. Coarse-grained soil itself has a rough texture and a high gravel content. Under the action of rainfall and runoff, fine soil particles are transported, and quartz particles in the slope soil are gradually exposed and retained on the surface. The soil surface covered with high gravel is like a suit of armor. The gravel particle layer will alleviate the erosion of the soil itself by raindrops and surface runoff, thereby protecting the surface soil. Based on the acquisition of soil surface coverage at different stages, dynamic observation of coarse-grained soil erosion can be achieved, and then the purpose of exploring the relationship between gravel coverage and soil erosion rate can be achieved. This observation method is of great significance for studying the mechanism of soil erosion and the evolution process of soil erosion.

[0003] In the existing observation technology, after the camera is used to take pictures, the image processing is performed to simulate rainfall. However, during the simulated rain falling process, the camera will always block the rain, so that a relatively small area of ​​rain is formed under the camera, which affects the accuracy of the simulation. The camera needs to take a large number of pictures during the process, and the continuous operation of the equipment also generates a lot of heat. Because the equipment is placed in the rain, it is often waterproofed, which affects the heat dissipation of the equipment and makes it difficult to maintain a good working condition for a long time. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the background technology and propose a method for measuring the gravel coverage rate on the surface of coarse-grained soil during rainfall, comprising the following steps:

[0005] S1. Install N cameras above the soil slope to be monitored and use the cameras to collect multiple images of the soil surface below;

[0006] S2, turning on the artificial simulated rainmaker to start rainfall erosion of the coarse-grained soil, collecting soil pictures within the corresponding time period at regular intervals during the rainfall period, and naming the pictures according to time;

[0007] S3, input the collected images into the trained CGAN network model to remove rain and fog from the images;

[0008] S4, performing lens correction and spatial geometry correction processing on the image from which rain and fog have been removed;

[0009] S5, graying the corrected image;

[0010] S6. The maximum inter-class variance method is used to divide the image into two parts: foreground and background, so as to separate the target (gravel) and the background (soil). The grayscale image is divided and converted into white and black pixels respectively. The black and white pixels are counted separately to obtain the representative pixels of the soil and the surface quartz gravel. The gravel particle coverage rate of the slope of the coarse-grained soil in the corresponding period is calculated.

[0011] Preferably, the rain and mist removal process in step S3 is performed by the following steps:

[0012] A1. Take and collect several pictures of the soil surface under conditions of no rainfall and rainfall;

[0013] A2. The collected pictures without rainfall are stored in the first folder, and the pictures with rainfall corresponding to the pictures without rainfall are stored in the second folder, and the pictures without rainfall and the pictures with rainfall are named the same to form a correspondence;

[0014] A3. Adjust the images in the first and second folders to a uniform size as the pre-collected data set;

[0015] A4. Classify the collected data set into a training set and a validation set. Then, input the training set into the deep learning CGAN network for training, and automatically adjust the parameters of the generator and discriminator in the deep learning CGAN network. The training set is required to train the weight parameters that the deep learning can modify autonomously.

[0016] The parameters of the generator and discriminator are adjusted by the following steps:

[0017] B1. Constructing a conditional generative deep learning CGAN network: The deep learning CGAN network includes two models: the generator and the discriminator. Let G be the generator and D be the discriminator.

[0018] B2, read the training set (various image files in the first folder and the second folder): let the clear image (soil surface image without rainfall) in the first folder be y, the image in the second folder with rainfall matching the clear image be x, and the random noise matrix be z; input the rain image or fog image x and the random noise matrix z in the training set into the generator G, and generate a new image;

[0019] B3. Compare the new image z* generated by the generator G with the clear image y through the discriminator D, and calculate the loss function based on the comparison results:

[0020] Loss(D,G)=Ey~Pdata(y)[log D(y|x)]+Ez~p_z(z)[log(1-D(G(z*|x)))]

[0021] Where E represents the mean of the image set in each batch-size, D(y|x) represents the probability of distinguishing the clear image y as a rain-free fog image when the fog image or rain image is used as the label; D(G(z*|x)) represents the probability of distinguishing the generated new image z* as a rain-free fog image when the fog image or rain image is used as the label;

[0022] B4. Calculate the loss value according to the loss function, automatically train and continue to adjust the parameters of the generator and discriminator; repeat steps 2 to 4 until the value of the loss function reaches the relative minimum value or the epoch reaches the preset threshold. The specific process is as follows:

[0023] First, adjust the parameters of the generator G so that the new image z* generated in the generator G is closer to the picture without rain and will not be misjudged, that is, the smaller G(z|x) is; then adjust the parameters of the discriminator D so that the larger the discrimination probability D(y|x) of the clear image y in the loss function, the larger the discrimination probability D(G(z*|x)) of the generated new image z* is; thus, the parameters of the two networks are continuously adjusted to meet the minimum probability of G(z I x) and the maximum probability of D(y|x) and D(G(z*|x)), and finally the relative minimum value MinMax of the loss function is obtained:

[0024]

[0025] B5. A CGAN-based rain and fog removal network model G is obtained, which can take the image to be processed (the image with rain and fog) as input and finally generate a high-definition image without rain and fog;

[0026] A5. Observe the output results of the deep learning CGAN network during the training process through the validation set to determine whether it is necessary to manually modify the hyperparameters and initialization values; if necessary, modify the hyperparameters and initialization values; if not, directly output the network model.

[0027] Preferably, the specific steps of calculation in step S6 are:

[0028] C1. Processing the image including image lens correction and space geometry correction;

[0029] C2. Image grayscale processing: The maximum value of the three-component brightness in the color image is used as the grayscale value of the grayscale image, using the following formula:

[0030] Gray(i,j)=max{R(i,j),G(i,j),B(i,j)}

[0031] Get the grayscale histogram of the grayscale value statistical features of the image. The average brightness value of each grayscale channel on the grayscale image is set to 255, the brightness value of the darkest color level is set to 0, and the remaining brightness values ​​change linearly between 0 and 255.

[0032] C3. Thresholding: The maximum between-class variance method (OTSU) is used as the threshold segmentation method, and the gravel coverage is defined as:

[0033]

[0034] Where: n is the total pixels; nw is the white pixel; nb is the black pixel.

[0035] A device for measuring gravel coverage on the surface of coarse-grained soil during rainfall, comprising a bucket-shaped shell, a camera device is fixedly arranged on the inner side of the bucket-shaped shell, and in the method of carrying out the method, the bucket-shaped shell is movably installed through a bracket structure, a soil container is arranged on the inner bottom surface of the bracket structure, a cylindrical end is extended from the lower end port of the bucket-shaped shell, a lens of the camera device is cylindrical in shape and extends from the lower end port of the bucket-shaped shell, and a blowing assembly is rotatably sleeved on the outer surface of the lens, a connecting cylinder is movably inserted on the lower end port of the cylindrical end, a sealing plate is fixedly installed on the upper end of the bucket-shaped shell, a ventilation and waterproof heat dissipation plate is arranged on the upper end surface of the sealing plate, and a plurality of drainage rods are convexly arranged on the outer surface of the bucket-shaped shell, wherein:

[0036] The blowing assembly blows air downward to form an air flow area, and water forms a downstream portion through the surface of the ventilated waterproof heat dissipation plate. The water in the downstream portion flows to the lower end of the drainage rod and drips to form a dripping portion. The water droplets around the air flow area float in the direction of approaching.

[0037] Preferably, the blowing assembly includes a rotating sleeve which is rotatably mounted on the outer surface of the lens on the camera device, a plurality of guide vanes are fixedly connected to the outer surface of the rotating sleeve in an annular array, a transmission gear ring is fixedly connected to the upper end port of the rotating sleeve, and the outer side of the transmission gear ring is provided with no less than three limit seats, the end of the limit seat is fixedly connected to the inner wall of the cylindrical end, and the transmission gear ring is driven to rotate by a driving mechanism arranged on the inner side of the bucket-shaped shell.

[0038] Preferably, the driving mechanism includes a driving motor fixedly connected to the upper end surface of the camera device, the output shaft of the driving motor is fixedly connected to the driving shaft, the driving shaft extends downward and is fixedly connected to a transmission gear at the lower end, and the transmission gear is meshed with a transmission gear ring.

[0039] Preferably, an insertion tube extends upward from the inner edge of the connecting tube, and the insertion tube is inserted into the inner side of the cylindrical end. Guide rods extend from the outer surfaces of both sides of the insertion tube respectively, and the guide rods slide through the outer surface of the cylindrical end. A reset spring is sleeved on the outer side of the guide rod, and the reset spring is located in the gap between the cylindrical end and the insertion tube. A diverter block is protruding from the lower end port of the connecting tube, and the number of the diverter blocks is set to be several and arranged in a circular array.

[0040] Preferably, the ventilated waterproof heat dissipation panel includes a heat dissipation tube fixedly embedded in the sealing plate, a plurality of heat dissipation plates arranged in a circular array are fixedly connected to the inner wall of the heat dissipation tube, the lower end of the heat dissipation plate is fixedly connected to the upper side of the camera device, the upper end of the heat dissipation plate is fixedly connected to a heat dissipation cover, the heat dissipation cover covers the top of the heat dissipation tube, and the lower end port of the heat dissipation cover is fixedly connected to a waterproof edge.

[0041] Preferably, the support structure includes an outer frame, and stepless telescopic mechanisms are arranged at the four corners of the outer frame, and a top cross frame is fixedly connected between the upper ends of the two stepless telescopic mechanisms close to one side, and a track mechanism is arranged between the top cross frames on both sides, and the track mechanism controls the horizontal movement of the bucket-shaped shell, and the soil container is arranged on the inner bottom surface of the outer frame, and rollers are rotatably installed at the lower end corners of the outer frame, and the track mechanism includes mounting seats respectively fixedly installed on the upper ends of the top cross frames, a scale rod fixedly connected between the mounting seats on both sides, and a screw rod rotatably connected between the two mounting seats, one end of the screw rod is fixedly connected to the output shaft of the pitch-adjusting motor, and the pitch-adjusting motor is fixedly installed on the side surface of the mounting seat, and a sliding block is fixedly connected to the upper end surface of the bucket-shaped shell, and the screw rod is threadedly matched with the sliding block, and the sliding block is slidably matched with the scale rod.

[0042] Preferably, the stepless telescopic mechanism includes a lower connecting block fixedly connected to the corner of the outer frame, and an upper connecting block fixedly connected to the end of the top cross frame, the upper end of the lower connecting block is fixedly connected to a guide rail, a rack is inserted into the inner side of the guide rail, the lower end of the rack is fixedly connected to a snap-in block, the snap-in block slides with the inner side of the guide rail, the upper end of the rack is fixedly connected to the lower end of the upper connecting block, a driving gear is rotatably installed on the inner side of the guide rail near the upper end, the driving gear is interlocked with the rack, the driving gear is fixedly connected to the output shaft of a servo motor, and the servo motor is fixedly installed on the side of the guide rail.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The device of the present invention can adaptively measure the gravel coverage rate on the surface of coarse-grained soil during rainfall. In the process of simulating rainfall, it can prevent the occurrence of rainfall dead corners during rain simulation and ensure full coverage of rain. During the continuous shooting process, the camera equipment will not be affected by overheating and affect normal work, thus ensuring the continuous measurement.

[0045] 2. The simulated soil is placed in a soil container, and water is dropped on the top of the outer frame to simulate rainfall operation. The rainwater above the camera device passes through the surface of the heat dissipation cover to form a downstream portion, and then slides down along the surface of the bucket-shaped shell. After being acted upon by the drainage rod, it becomes water droplets and falls down to form a drop portion. During this process, the driving motor drives the driving shaft to rotate, and the rotating sleeve rotates under the cooperation of the transmission gear and the transmission gear ring. Under the action of the drainage blades, wind is blown downward to form an air flow area. Due to the fast air flow rate at this position, a negative pressure area is formed, so that the water droplets fall toward the bottom of the camera device, and then the fallen raindrops can be fully distributed on the surface of the simulated soil, so that the final speed of the raindrops reaching the soil surface is close to the final speed of real rainfall, avoiding the formation of dead angles.

[0046] 3. When the rotating sleeve rotates, the flow of air causes the insertion tube to swing, that is, it swings in accordance with the guidance of the guide rod. Under the action of the reset spring, the conversion between kinetic energy and elastic potential energy increases the duration and number of swings. When the remaining water flows to the lower end of the connecting tube, the water drips under the action of the diverter block due to the swinging action, and the formation of water droplets is promoted to avoid the formation of water columns, which more closely simulates the rainfall.

[0047] 4. The heat sink will conduct the heat generated by the camera equipment to the surface of the heat sink cover. When it comes into contact with water, it effectively promotes heat absorption and improves the heat dissipation effect. When the rotating sleeve rotates, the air above is drawn in through the gap between the heat sink cover and the heat sink, promoting the flow of air inside, thereby assisting in heat dissipation. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic structural diagram of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0049] Figure 2 It is a schematic diagram of a bucket-shaped shell of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0050] Figure 3 It is a cross-sectional view of the bucket-shaped shell of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0051] Figure 4It is a partial cross-sectional view of a bucket-shaped shell of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0052] Figure 5 This is an enlarged view of the A in the soil 4 of the device for measuring the gravel coverage rate on the surface of coarse-grained soil during rainfall according to the present invention;

[0053] Figure 6 It is a schematic diagram of the structure of a ventilation and waterproof heat dissipation plate of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0054] Figure 7 A cross-sectional view of a ventilation and waterproof heat dissipation plate of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0055] Figure 8 It is a schematic diagram of a blowing component of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0056] Fig. 9 It is a schematic diagram of a distance adjustment mechanism of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0057] Fig.10 It is a schematic diagram of a stepless telescopic mechanism of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention;

[0058] Fig.11 It is a partial schematic diagram of a device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention during simulated rainfall;

[0059] Fig.12 This is an image processing flow chart of a method for determining gravel coverage on the surface of coarse-grained soil during rainfall according to the present invention.

[0060] 1. bucket shell; 2. drainage rod; 3. camera device; 4. cylindrical end; 5. rotating sleeve; 6. drainage blade; 7. transmission gear ring; 8. limit seat; 9. transmission gear; 10. driving shaft; 11. driving motor; 12. connecting tube; 13. inserting tube; 14. guide rod; 15. reset spring; 16. diverter block; 17. sealing plate; 18. heat dissipation tube; 19. heat dissipation plate; 20. heat dissipation cover; 21. anti Water edge; 22. Sliding block; 23. Mounting seat; 24. Screw rod; 25. Scale rod; 26. Pitch-adjusting motor; 27. Top cross frame; 28. Outer frame; 29. ​​Lower connecting block; 30. Upper connecting block; 31. Guide rail; 32. Rack; 33. Snap-in block; 34. Servo motor; 35. Driving gear; 36. Roller; 37. Soil container; 38. Downstream part; 39. Dropping part; 40. Air flow area. DETAILED DESCRIPTION

[0061] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0062] A method for determining the gravel coverage on the surface of coarse-grained soil during rainfall is proposed, which includes the following steps:

[0063] S1. Install N cameras above the soil slope to be monitored and use the cameras to collect multiple images of the soil surface below;

[0064] S2, turning on the artificial simulated rainfall device to start rainfall erosion of the coarse-grained soil, collecting soil pictures within the corresponding time period at regular intervals during the rainfall period, and naming the pictures according to time;

[0065] S3, input the collected images into the trained CGAN network model to remove rain and fog from the images;

[0066] S4, performing lens correction and spatial geometry correction processing on the image from which rain and fog have been removed;

[0067] S5, graying the corrected image;

[0068] S6. The maximum inter-class variance method is used to divide the image into two parts: foreground and background, so as to separate the target (gravel) and the background (soil). The grayscale image is divided and converted into white and black pixels respectively. The black and white pixels are counted separately to obtain the representative pixels of the soil and the surface quartz gravel. The gravel particle coverage rate of the slope of the coarse-grained soil in the corresponding period is calculated.

[0069] The rain and fog removal process in step S3 is performed by the following steps:

[0070] A1. Take and collect several pictures of the soil surface under conditions of no rainfall and rainfall;

[0071] A2. The collected pictures without rainfall are stored in the first folder, and the pictures with rainfall corresponding to the pictures without rainfall are stored in the second folder, and the pictures without rainfall and the pictures with rainfall are named the same to form a correspondence;

[0072] A3. Adjust the images in the first and second folders to a uniform size as the pre-collected data set;

[0073] A4. Classify the collected data set into a training set and a validation set. Then, input the training set into the deep learning CGAN network for training, and automatically adjust the parameters of the generator and discriminator in the deep learning CGAN network. The training set is required to train the weight parameters that the deep learning can modify autonomously.

[0074] The parameters of the generator and discriminator are adjusted by the following steps:

[0075] B1. Constructing a conditional generative deep learning CGAN network: The deep learning CGAN network includes two models: the generator and the discriminator. Let G be the generator and D be the discriminator.

[0076] B2, read the training set (various image files in the first folder and the second folder): let the clear image (soil surface image without rainfall) in the first folder be y, the image in the second folder with rainfall matching the clear image be x, and the random noise matrix be z; input the rain image or fog image x and the random noise matrix z in the training set into the generator G, and generate a new image;

[0077] B3. Compare the new image z* generated by the generator G with the clear image y through the discriminator D, and calculate the loss function based on the comparison results:

[0078] Loss(D,G)=Ey~Pdata(y)[log D(y|x)]+Ez~p_z(z)[log(1-D(G(z*|x)))]

[0079] Where E represents the mean of the image set in each batch-size, D(y|x) represents the probability of distinguishing the clear image y as a rain-free fog image when the fog image or rain image is used as the label; D(G(z*|x)) represents the probability of distinguishing the generated new image z* as a rain-free fog image when the fog image or rain image is used as the label;

[0080] B4. Calculate the loss value according to the loss function, automatically train and continue to adjust the parameters of the generator and discriminator; repeat steps 2 to 4 until the value of the loss function reaches the relative minimum value or the epoch reaches the preset threshold. The specific process is as follows:

[0081] First, adjust the parameters of the generator G so that the new image z* generated in the generator G is closer to the picture without rain and will not be misjudged, that is, the smaller G(z|x) is; then adjust the parameters of the discriminator D so that the larger the discrimination probability D(y|x) of the clear image y in the loss function, the larger the discrimination probability D(G(z*|x)) of the generated new image z* is; thus, the parameters of the two networks are continuously adjusted to meet the minimum probability of G(z I x) and the maximum probability of D(y|x) and D(G(z*|x)), and finally the relative minimum value MinMax of the loss function is obtained:

[0082]

[0083] B5. A CGAN-based rain and fog removal network model G is obtained, which can take the image to be processed (the image with rain and fog) as input and finally generate a high-definition image without rain and fog;

[0084] A5. Observe the output results of the deep learning CGAN network during the training process through the validation set to determine whether it is necessary to manually modify the hyperparameters and initialization values; if necessary, modify the hyperparameters and initialization values; if not, directly output the network model.

[0085] The specific calculation steps in step S6 are:

[0086] C1. Processing the image including image lens correction and space geometry correction;

[0087] C2. Image grayscale processing: The maximum value of the three-component brightness in the color image is used as the grayscale value of the grayscale image, using the following formula:

[0088] Gray(i,j)=max{R(i,j),G(i,j),B(i,j)}

[0089] Get the grayscale histogram of the grayscale value statistical features of the image. The average brightness value of each grayscale channel on the grayscale image is set to 255, the brightness value of the darkest color level is set to 0, and the remaining brightness values ​​change linearly between 0 and 255.

[0090] C3. Thresholding: The maximum between-class variance method (OTSU) is used as the threshold segmentation method, and the gravel coverage is defined as:

[0091]

[0092] Where: n is the total pixels; nw is the white pixel; nb is the black pixel.

[0093] like Fig.12 The figure shows the image processing flow chart during the measurement process.

[0094] like Figure 1-Figure 11The device shown is used for measuring gravel coverage rate on the surface of coarse-grained soil during rainfall, comprising a bucket-shaped shell 1, a camera device 3 is fixedly arranged on the inner side of the bucket-shaped shell 1, the bucket-shaped shell 1 is movably installed through a bracket structure, a soil container 37 is arranged on the inner bottom surface of the bracket structure, a cylindrical end 4 extends from the lower end of the bucket-shaped shell 1, a lens of the camera device 3 is cylindrical in shape and extends from the lower end of the bucket-shaped shell 1, and a blowing assembly is rotatably sleeved on the outer surface of the lens, a connecting tube 12 is movably inserted at the lower end of the cylindrical end 4, a sealing plate 17 is fixedly installed on the upper end of the bucket-shaped shell 1, a ventilation and waterproof heat dissipation plate is arranged on the upper end surface of the sealing plate 17, and a plurality of drainage rods 2 are convexly arranged on the outer surface of the bucket-shaped shell 1, wherein:

[0095] The blowing assembly blows the air downward to form an air flow area 40, and water forms a downstream portion 38 through the surface of the ventilated waterproof heat dissipation plate. The water in the downstream portion 38 flows to the lower end of the drainage rod 2 and drips to form a dripping portion 39. The water droplets around the air flow area 40 float in a direction close to it.

[0096] The device of the present invention can adaptively measure the gravel coverage rate on the surface of coarse-grained soil during rainfall. In the process of simulating rainfall, it can prevent the occurrence of rainfall blind spots during rain simulation and ensure comprehensive coverage of rain. During the continuous shooting process, the camera device 3 will not affect normal operation due to overheating, thereby ensuring the continuous measurement.

[0097] The blowing assembly includes a rotating sleeve 5 rotatably mounted on the outer surface of the lens of the camera device 3, a plurality of guide vanes 6 are fixedly connected to the outer surface of the rotating sleeve 5 in an annular array, a transmission gear ring 7 is fixedly connected to the upper end port of the rotating sleeve 5, and the outer side of the transmission gear ring 7 is provided with no less than three limit seats 8, and the end of the limit seat 8 is fixedly connected to the inner wall of the cylindrical end 4. The transmission gear ring 7 is driven to rotate by a driving mechanism arranged on the inner side of the bucket-shaped shell 1, and the driving mechanism includes a driving motor 11 fixedly connected to the upper end surface of the camera device 3, and the output shaft of the driving motor 11 is fixedly connected to a driving shaft 10, and the driving shaft 10 extends downward and is fixedly connected to a transmission gear 9 at the lower end, and the transmission gear 9 is meshed with the transmission gear ring 7 to provide power for the rotation of the rotating sleeve 5.

[0098] An insertion tube 13 extends upward from the inner edge of the connecting tube 12, and the insertion tube 13 is inserted into the inner side of the cylindrical end 4. Guide rods 14 extend from the outer surfaces of both sides of the insertion tube 13 respectively, and the guide rods 14 slide through the outer surface of the cylindrical end 4. A return spring 15 is sleeved on the outer side of the guide rod 14, and the return spring 15 is located in the gap between the cylindrical end 4 and the insertion tube 13. A diverter block 16 is protruded from the lower end port of the connecting tube 12. The number of the diverter blocks 16 is set to be several and arranged in a circular array. The diverter blocks 16 promote the formation of water droplets.

[0099] The ventilated and waterproof heat dissipation panel includes a heat dissipation tube 18 fixedly embedded in the upper side of the sealing plate 17, and a plurality of heat dissipation plates 19 arranged in a circular array are fixedly connected to the inner wall of the heat dissipation tube 18, and the lower end of the heat dissipation plate 19 is fixedly connected to the upper side of the camera device 3, and the upper end of the heat dissipation plate 19 is fixedly connected to a heat dissipation cover 20, and the heat dissipation cover 20 covers the top of the heat dissipation tube 18, and the lower end port of the heat dissipation cover 20 is fixedly connected to a waterproof edge 21 to prevent water from flowing into the inner side of the heat dissipation tube 18.

[0100] The support structure includes an outer frame 28, and the four corners of the outer frame 28 are provided with infinite telescopic mechanisms. The upper ends of the two infinite telescopic mechanisms near one side are fixedly connected with a top cross frame 27. A track mechanism is provided between the top cross frames 27 on both sides. The track mechanism controls the horizontal movement of the bucket-shaped shell 1. The soil container 37 is provided on the inner bottom surface of the outer frame 28. The lower corner of the outer frame 28 is rotatably provided with a roller 36. The track mechanism includes a mounting seat 23 fixedly installed on the upper end of the top cross frame 27, and a mounting seat 23 fixedly connected between the mounting seats 23 on both sides. The scale rod 25 is a screw rod 24 rotatably connected between the two mounting seats 23. One end of the screw rod 24 is fixedly connected to the output shaft of the pitch-adjusting motor 26. The pitch-adjusting motor 26 is fixedly mounted on the side surface of the mounting seat 23. The upper end surface of the bucket-shaped shell 1 is fixedly connected with a sliding block 22. The screw rod 24 is threadedly matched with the sliding block 22. The sliding block 22 is slidably matched with the scale rod 25. The pitch-adjusting motor 26 drives the screw rod 24 to rotate, thereby causing the sliding block 22 to move and adjust the shooting position. The scale rod 25 can provide accurate readings for the adjusted distance and precise control.

[0101] The stepless telescopic mechanism includes a lower connecting block 29 fixedly connected to the corner of the outer frame 28, and an upper connecting block 30 fixedly connected to the end of the top cross frame 27. The upper end of the lower connecting block 29 is fixedly connected to a guide rail 31, and a rack 32 is inserted into the inner side of the guide rail 31. The lower end of the rack 32 is fixedly connected to a snap-in block 33, and the snap-in block 33 slides with the inner side of the guide rail 31. The upper end of the rack 32 is fixedly connected to the lower end of the upper connecting block 30, and a driving gear 35 is rotatably installed on the inner side of the guide rail 31 near the upper end. The driving gear 35 and the rack 32 are mutually cored, and the driving gear 35 is fixedly connected to the output shaft of the servo motor 34. The servo motor 34 is fixedly installed on the side of the guide rail 31. When the servo motor 34 runs, the driving gear 35 rotates, and then the rack 32 rotates, thereby prompting the rack 32 to move up and down, and then the top cross frame 27 moves up and down, and the four stepless telescopic mechanisms keep synchronous operation.

[0102] During the measurement, simulated soil is placed in a soil container 37, and water is dropped on the top of the outer frame 28 to simulate rainfall operation. The rainwater above the camera device 3 passes through the surface of the heat dissipation cover 20 to form a downstream portion 38, and then slides down along the surface of the bucket-shaped shell 1, and after being acted upon by the drainage rod 2, it becomes water droplets and falls down to form a drop portion 39. During this process, the drive motor 11 drives the drive shaft 10 to rotate, and the rotating sleeve 5 rotates under the cooperation of the transmission gear 9 and the transmission gear ring 7. Under the action of the drainage blades 6, wind is blown downward to form an air flow area 40. Since the air flow rate at this position is fast, a negative pressure area is formed, so that the water droplets fall toward the bottom of the camera device 3, and then the fallen raindrops can be fully distributed on the surface of the simulated soil to avoid the formation of dead corners. When the rotating sleeve 5 rotates, the flow of air causes the insertion tube 13 to swing, that is, it swings in accordance with the guidance of the guide rod 14. Under the action of the return spring 15, the conversion between kinetic energy and elastic potential energy increases the duration and number of swings of the swing, so that when the remaining water flows to the lower end of the connecting tube 12, the water drips under the action of the diverter block 16 due to the swinging action, and promotes the formation of water droplets, avoids the formation of water columns, and more accurately simulates the rainfall; the heat sink 19 will conduct the heat generated by the camera device 3 to the surface of the heat sink cover 20. When it comes into contact with water, it effectively promotes heat absorption and improves the heat dissipation effect. When the rotating sleeve 5 rotates, the air above is drawn in along the gap between the heat sink cover 20 and the heat sink 18, promoting the flow of air inside, thereby assisting in heat dissipation.

[0103] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for determining the gravel coverage rate on the surface of coarse-grained soil during rainfall, characterized in that: The following steps are involved: S1. Install N cameras above the soil slope to be monitored and use the cameras to collect multiple images of the soil surface below; S2, turning on the artificial simulated rainmaker to start rainfall erosion of the coarse-grained soil, collecting soil pictures within the corresponding time period at regular intervals during the rainfall period, and naming the pictures according to time; S3, input the collected images into the trained CGAN network model to remove rain and fog from the images; S4, performing lens correction and spatial geometry correction processing on the image from which rain and fog have been removed; S5, graying the corrected image; S6. The maximum inter-class variance method is used to divide the image into two parts, the foreground and the background, so as to separate the target gravel from the background soil. The grayscale image is divided and converted into white and black pixels respectively. The black and white pixels are counted respectively to obtain the representative pixels of the soil and the surface quartz gravel. The coverage rate of the gravel particles on the slope of the coarse-grained soil in the corresponding period is calculated. The rain and fog removal process in step S3 is performed by the following steps: A1. Take and collect several pictures of the soil surface under conditions of no rainfall and rainfall; A2. The collected pictures without rainfall are stored in the first folder, and the pictures with rainfall corresponding to the pictures without rainfall are stored in the second folder, and the pictures without rainfall and the pictures with rainfall are named the same to form a correspondence; A3. Adjust the images in the first and second folders to a uniform size as the pre-collected data set; A4. Classify the collected data set into a training set and a validation set. Then, input the training set into the deep learning CGAN network for training, and automatically adjust the parameters of the generator and discriminator in the deep learning CGAN network. The training set is required to train the weight parameters that the deep learning can modify autonomously. The parameters of the generator and discriminator are adjusted by the following steps: B1. Constructing a conditional generative deep learning CGAN network: The deep learning CGAN network includes two models: the generator and the discriminator. Let G be the generator and D be the discriminator. B2, read various image files in the first folder and the second folder of the training set: let the soil surface picture of the clear image without rainfall in the first folder be y, the picture under rainfall conditions matching the clear image in the second folder be x, and the random noise matrix be z; input the rain image or fog image x and the random noise matrix z in the training set into the generator G, and generate a new image; B3. Compare the new image z* generated by the generator G with the clear image y through the discriminator D, and calculate the loss function based on the comparison results: Loss(D,G)=Ey~Pdata(y)[log D(y|x)]+Ez~p_z(z)[log(1-D(G(z*|x)))] Where E represents the mean of the image set in each batch-size, D(y|x) represents the probability of distinguishing the clear image y as a rain-free fog image when the fog image or rain image is used as the label; D(G(z*|x)) represents the probability of distinguishing the generated new image z* as a rain-free fog image when the fog image or rain image is used as the label; B4. Calculate the loss value according to the loss function, automatically train and continue to adjust the parameters of the generator and discriminator; repeat steps 2 to 4 until the value of the loss function reaches the relative minimum value or the epoch reaches the preset threshold. The specific process is as follows: First, adjust the parameters of the generator G so that the new image z* generated in the generator G is closer to the picture without rain and will not be misjudged, that is, the smaller G(z|x) is; then adjust the parameters of the discriminator D so that the larger the discrimination probability D(y|x) of the clear image y in the loss function, the larger the discrimination probability D(G(z*|x)) of the generated new image z* is; thus, the parameters of the two networks are continuously adjusted to meet the minimum probability of G(z I x) and the maximum probability of D(y|x) and D(G(z*|x)), and finally the relative minimum value MinMax of the loss function is obtained: B5. A CGAN-based rain and fog removal network model G is obtained, which can take the image with rain and fog as input and finally generate a high-definition image without rain and fog; A5. Observe the output results of the deep learning CGAN network during the training process through the validation set to determine whether it is necessary to manually modify the hyperparameters and initialization values; if necessary, modify the hyperparameters and initialization values; if not, directly output the network model.

2. The method for determining the gravel coverage rate on the surface of coarse-grained soil during rainfall according to claim 1, characterized in that: The specific calculation steps in step S6 are: C1. Processing the image including image lens correction and space geometry correction; C2. Image grayscale processing: The maximum value of the three-component brightness in the color image is used as the grayscale value of the grayscale image, using the following formula: Gray(i,j)=max{R(i,j),G(i,j),B(i,j)} Get the grayscale histogram of the grayscale value statistical features of the image. The average brightness value of each grayscale channel on the grayscale image is set to 255, the brightness value of the darkest color level is set to 0, and the remaining brightness values ​​change linearly between 0 and 255. C3. Thresholding: The maximum inter-class variance method (OTSU) is used as the threshold segmentation method, and the gravel coverage is defined as: Where: n is the total pixels; nw is the white pixel; nb is the black pixel.

3. A device for measuring gravel coverage on the surface of coarse-grained soil during rainfall, comprising a bucket-shaped shell (1), wherein a camera device (3) is fixedly arranged on the inner side of the bucket-shaped shell (1), and is used in the method according to any one of claims 1 to 2, characterized in that: The bucket-shaped shell (1) is movably mounted via a support structure, a soil container (37) is provided on the inner bottom surface of the support structure, a cylindrical end (4) extends from the lower end of the bucket-shaped shell (1), a lens of the camera device (3) is cylindrical in shape and extends from the lower end of the bucket-shaped shell (1), and a blowing assembly is rotatably sleeved on the outer surface of the lens, a connecting tube (12) is movably inserted into the lower end of the cylindrical end (4), a sealing plate (17) is fixedly mounted on the upper end of the bucket-shaped shell (1), a ventilation and waterproof heat dissipation plate is provided on the upper end surface of the sealing plate (17), and a plurality of drainage rods (2) are convexly provided on the outer surface of the bucket-shaped shell (1), wherein: The blowing assembly blows air downward to form an air flow area (40), and water forms a downstream portion (38) via the surface of the ventilated waterproof heat dissipation plate. The water in the downstream portion (38) flows to the lower end of the drainage rod (2) and drips to form a dripping portion (39), and the water droplets around the air flow area (40) float in a direction close to it.

4. The device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to claim 3, characterized in that: The blowing assembly comprises a rotating sleeve (5) rotatably mounted on the outer surface of a lens of a camera device (3); a plurality of guide vanes (6) are fixedly connected in an annular array on the outer surface of the rotating sleeve (5); a transmission gear ring (7) is fixedly connected to the upper end of the rotating sleeve (5); a number of not less than three limit seats (8) are mounted on the outer side of the transmission gear ring (7); the end of the limit seat (8) is fixedly connected to the inner wall of the cylindrical end (4); and the transmission gear ring (7) is driven to rotate by a driving mechanism arranged on the inner side of the bucket-shaped shell (1).

5. The device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to claim 4, characterized in that: The driving mechanism comprises a driving motor (11) fixedly connected to the upper end surface of the camera device (3); the output shaft of the driving motor (11) is fixedly connected to a driving shaft (10); the driving shaft (10) extends downward and is fixedly connected to a transmission gear (9) at the lower end; the transmission gear (9) is meshed with a transmission gear ring (7).

6. The device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to claim 5, characterized in that: An insertion tube (13) extends upward from the inner edge of the connecting tube (12), and the insertion tube (13) is inserted into the inner side of the cylindrical end (4). Guide rods (14) extend from the outer surfaces of both sides of the insertion tube (13), and the guide rods (14) slide through the outer surface of the cylindrical end (4). A return spring (15) is sleeved on the outer side of the guide rod (14), and the return spring (15) is located in the gap between the cylindrical end (4) and the insertion tube (13). A diverter block (16) is protruding from the lower end port of the connecting tube (12), and the number of the diverter blocks (16) is set to be several and arranged in a ring array.

7. The device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to claim 3, characterized in that: The ventilated waterproof heat dissipation plate comprises a heat dissipation tube (18) fixedly embedded in the upper side of the sealing plate (17); a plurality of heat dissipation plates (19) arranged in a circular array are fixedly connected to the inner wall of the heat dissipation tube (18); the lower end of the heat dissipation plate (19) is fixedly connected to the upper side of the camera device (3); the upper end of the heat dissipation plate (19) is fixedly connected to a heat dissipation cover (20); the heat dissipation cover (20) covers the upper side of the heat dissipation tube (18); and the lower end port of the heat dissipation cover (20) is fixedly connected to a waterproof edge (21).

8. The device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to claim 3, characterized in that: The support structure comprises an outer frame (28), and stepless telescopic mechanisms are arranged at four corners of the outer frame (28). A top cross frame (27) is fixedly connected between the upper ends of two stepless telescopic mechanisms close to one side. A track mechanism is arranged between the top cross frames (27) on both sides. The track mechanism controls the horizontal movement of the bucket-shaped shell (1). The soil container (37) is arranged on the inner bottom surface of the outer frame (28). Rollers (36) are rotatably installed at the lower corners of the outer frame (28). The track mechanism comprises rollers (36) respectively fixedly installed on the top cross frames ( 27) a mounting seat (23) at the upper end, a scale rod (25) fixedly connected between the mounting seats (23) on both sides, and a screw rod (24) rotatably connected between the two mounting seats (23), one end of the screw rod (24) is fixedly connected to the output shaft of the pitch-adjusting motor (26), and the pitch-adjusting motor (26) is fixedly mounted on the side surface of the mounting seat (23); the upper end surface of the bucket-shaped shell (1) is fixedly connected with a sliding block (22), the screw rod (24) is threadedly matched with the sliding block (22), and the sliding block (22) is slidably matched with the scale rod (25).

9. The device for measuring gravel coverage on the surface of coarse-grained soil during rainfall according to claim 8, characterized in that: The stepless telescopic mechanism comprises a lower connecting block (29) fixedly connected to a corner of an outer frame (28), and an upper connecting block (30) fixedly connected to an end of a top cross frame (27); the upper end of the lower connecting block (29) is fixedly connected to a guide rail (31); a rack (32) is inserted into the inner side of the guide rail (31); the lower end of the rack (32) is fixedly connected to a snap-in block (33); the snap-in block (33) is slidably matched with the inner side of the guide rail (31); the upper end of the rack (32) is fixedly connected to the lower end of the upper connecting block (30); a driving gear (35) is rotatably installed on the inner side of the guide rail (31) near the upper end; the driving gear (35) is cored with the rack (32); the driving gear (35) is fixedly connected to the output shaft of a servo motor (34); and the servo motor (34) is fixedly installed on the side of the guide rail (31).