A method and system for analyzing the size and concentration of mountain torrent particles

Abstract

The application discloses a mountain torrent particle size and concentration analysis method and system, and the method comprises the following steps: obtaining a surface image of a mountain torrent; inputting the surface image into a preset YOLOv12-MD network, and obtaining high pixel values and wide pixel values of all mountain torrent particles in the surface image through the YOLOv12-MD network; determining the particle size of each mountain torrent particle by using a preset mountain torrent particle size estimation strategy according to water level data, flow rate data and the high pixel values and the wide pixel values of the mountain torrent particles, and obtaining the particle size of each mountain torrent particle; and calculating the concentration of the mountain torrent particles in the mountain torrent according to the particle size of each mountain torrent particle and a preset surface volume empirical relationship model. The estimation accuracy of the particle size and the concentration of the mountain torrent particles in a complex environment is improved, and the complex environment conditions that cannot be covered by a single parameter are compensated.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of mountain torrent particle size and concentration analysis, and particularly relates to a mountain torrent particle size and concentration analysis method and system. BACKGROUND

[0002] Most mountain torrents occur at night, and their suddenness and strong destructive power will cause great harm if they cannot be discovered in time and protective measures are not taken. Currently, the products on the market for detecting mountain torrent particle size and concentration are all contact type, and when used, mountain torrent sampling (sampling offline detection) needs to be performed, or the sensor needs to be placed in the mountain torrent (such as placing a pipeline in the river for online sampling). The above-mentioned methods have three main problems: 1. Sampling detection is an offline method, which cannot be monitored in real time and online to achieve mountain torrent early warning and timely prevention purposes; 2. Placing the sensor in the mountain torrent is easy to wash away the measuring equipment, causing equipment loss and losing the monitoring capability at the same time; 3. Offline detection equipment is generally large in size, and the online sampling method of placing a pipeline in the river has high engineering difficulty and cost. In view of this, the present application can greatly reduce the property loss and the risk of human life safety threatened by mountain torrents, and has great practical value. SUMMARY

[0003] The present application provides a mountain torrent particle size and concentration analysis method, system and readable storage medium, which are used to solve the technical problem of difficult accurate analysis of mountain torrent particle concentration.

[0004] In a first aspect, the present application provides a mountain torrent particle size and concentration analysis method, comprising:

[0005] obtaining a surface image, water level data and flow rate data of a mountain torrent, wherein the surface image contains mountain torrent particles;

[0006] inputting the surface image into a preset YOLOv12-MD network, wherein the YOLOv12-MD network outputs high pixel values and wide pixel values of all mountain torrent particles in the surface image;

[0007] determining the particle size of each mountain torrent particle by using a preset mountain torrent particle size estimation strategy according to the water level data, the flow rate data and the high pixel values and the wide pixel values of each mountain torrent particle, to obtain the particle size of each mountain torrent particle;

[0008] calculating the mountain torrent particle concentration in the mountain torrent according to the particle size of each mountain torrent particle and a preset surface volume empirical relationship model.

[0009] In a second aspect, the present application provides a mountain torrent particle size and concentration analysis system, comprising:

[0010] The acquisition module is configured to acquire a surface image of the mountain torrent, water level data and flow velocity data, wherein the surface image contains mountain torrent particles;

[0011] The output module is configured to input the surface image into a preset YOLOv12-MD network, and the YOLOv12-MD network outputs high pixel values and wide pixel values of all mountain torrent particles in the surface image.

[0012] The determination module is configured to determine the particle size of each mountain torrent particle according to the water level data, the flow velocity data and the high pixel values and the wide pixel values of each mountain torrent particle, and obtain the particle size of each mountain torrent particle by using a preset mountain torrent particle size estimation strategy.

[0013] The calculation module is configured to calculate the concentration of the mountain torrent particles in the mountain torrent according to the particle size of each mountain torrent particle and a preset surface volume empirical relationship model.

[0014] In a third aspect, an electronic device is provided, which includes at least one processor and a memory connected to the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the mountain torrent particle size and concentration analysis method of any embodiment of the present application.

[0015] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and the program instructions are executed by a processor to enable the processor to perform the steps of the mountain torrent particle size and concentration analysis method of any embodiment of the present application.

[0016] The mountain torrent particle size and concentration analysis method and system of the present application avoid the non-real-time or the damageability of the monitoring equipment of the traditional contact measurement through the heterogeneous fusion of the bimodal optical imaging and the hydrodynamic parameters, improve the real-time, stability and safety of the monitoring, and combine the hydrodynamic parameters (water level and flow velocity) to construct a nonlinear particle size and concentration estimation model, thereby improving the estimation accuracy of the mountain torrent particle size and concentration in complex environments and making up for the complex environmental conditions that cannot be covered by a single parameter. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0018] Figure 1A flow chart of a mountain torrent particle size and concentration analysis method provided by an embodiment of the present application is shown in

[0019] Figure 2 A structural block diagram of a mountain torrent particle size and concentration analysis system provided by an embodiment of the present application is shown in

[0020] Figure 3 A structural schematic diagram of an electronic device provided by an embodiment of the present application is shown in DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0022] Reference is made to Figure 1 , which shows a flow chart of a mountain torrent particle size and concentration analysis method.

[0023] As shown in Figure 1 , the mountain torrent particle size and concentration analysis method specifically includes the following steps:

[0024] In step S101, a surface image, water level data and flow rate data of a mountain torrent are obtained, wherein the surface image contains mountain torrent particles.

[0025] In this step, the surface image is an image obtained via dual-mode optical imaging.

[0026] In step S102, each region image, the water level data and the flow rate data are input into a preset YOLOv12-MD network, and the YOLOv12-MD network outputs high pixel values and wide pixel values of all mountain torrent particles in each region image.

[0027] In this step, the YOLOv12-MD network is divided into a backbone network, a neck network and a detection head, wherein the backbone network outputs multi-scale feature maps of the mountain torrent particles for multi-level feature extraction, the neck network fuses the dual-mode and multi-scale feature maps output by the backbone network, and the detection head outputs the mountain torrent particle class, quantity and each target pixel size detection result.

[0028] It should be noted that the backbone network includes a shallow backbone network, a middle backbone network and a deep backbone network.

[0029] Shallow backbone network: output detailed feature map, mainly used for granular edge, texture and other low-level feature extraction. Improved super-resolution feature extraction module (MAB) is introduced in the shallow layer to enhance the image detail recovery ability and solve the problem of blurred collected images.

[0030] Middle layer backbone network: output local feature map, used for extracting local region features of granules. Continue to use MAB module in the middle layer to further improve the feature extraction effect.

[0031] Deep layer backbone network: output global semantic feature map, used for capturing global semantic information of granules. Parameterized Mamba block is introduced in the deep layer backbone network to establish a dynamic mapping relationship between the state transition matrix and the environmental parameters, optimize the long-distance dependence relationship, improve the target contour clarity, and optimize the calculation efficiency.

[0032] Specifically, a dynamic deformable convolution compensation mechanism (DDFConv) is introduced in the shallow layer backbone network and the middle layer backbone network, and the offset and the spatial attention weight are generated by the following expression:

[0033] ,

[0034] In the formula, is the output feature map of the dynamic deformable convolution, is the 7x7 convolution operation of the offset matrix, is the offset matrix, is the 3x3 convolution of the spatial attention weight matrix, is the spatial attention weight matrix, is the 3x3 convolution of the spatial attention weight matrix, and the Softmax function is used for normalization, is the Hadamard product operation, is the input feature, H is the height pixel number, W is the width pixel number, and C is the channel number;

[0035] The expression of the dynamic mapping relationship is:

[0036] ,

[0037] ,

[0038] In the formula, is the state transition matrix adjusted dynamically, is the initial state transition matrix, is, is the water level influence weight matrix, is the real-time water level height, is the flow rate influence weight matrix, is the real-time flow rate, This is the weighting matrix for the interaction between water level and flow velocity. For Kronecker product.

[0039] The multimodal fusion strategy dynamically weights the forward hidden state and the backward hidden state using gating weights:

[0040] ,

[0041] In the formula, For the fusion of multimodal time series data, the hidden state, For the Sigmoid function, The number of multimodal data points is [number], and the multimodal data points are surface images, water level data, and flow velocity data. For learnable gated matrices, For Hadamard product operation, For the first The hidden states output by the modal feedforward Mamba block. For the first The hidden state output by the modal reverse Mamba block.

[0042] Step S103: Based on the water level data, the flow velocity data, and the high pixel value and wide pixel value of each flash flood particle, the particle size of each flash flood particle is determined using a preset flash flood particle size estimation strategy, and the particle size of each flash flood particle is obtained.

[0043] In this step, the expression for the flash flood particle size estimation strategy is:

[0044] ,

[0045] ,

[0046] ,

[0047] ,

[0048] In the formula, For the first The particle size of a flash flood particle, For the first The pixel size of each flash flood particle. This is the initial refraction correction factor. Gain for flow rate correction This is an estimated value of the band attenuation coefficient under bright field conditions. This is an estimate of the band attenuation coefficient under dark conditions. For the first The high pixel value of each flash flood particle For the first The width pixel value of each flash flood particle. These are the calibration coefficients for the experiment, determined using experimental data and a regression algorithm. It is an exponentially decaying function. These are learnable parameters under bright field conditions. For real-time flow rate, This is the real-time water level height. This is the initial band attenuation coefficient.

[0049] Step S104: Calculate the concentration of flash flood particles in the flash flood based on the particle size of each flash flood particle and a preset empirical relationship model of surface volume.

[0050] In this step, the mass of a single flash flood particle is calculated using the following expression:

[0051] ,

[0052] In the formula, For the first The particle mass of each flash flood particle, For the first The particle size of a flash flood particle, For the density of sediment particles, take 2650. ;

[0053] The concentration of flash flood particles in the flash flood is calculated based on the empirical relationship model between the particle mass of a single flash flood particle and the surface volume. The expression for calculating the concentration of flash flood particles is as follows:

[0054] ,

[0055] ,

[0056] ,

[0057] In the formula, The volume of the flash flood area within the surface image. The surface area of ​​the flash flood region within the surface image. , All are experimental calibration coefficients, determined using experimental data and a regression algorithm. The effect of flow velocity on suspension, For real-time flow rate, The real-time water level is represented by L, and the total number of particles is represented by L. This represents the median particle size.

[0058] In summary, the method of this application can achieve the following technical effects:

[0059] Non-contact mountain flood particle size and concentration monitoring: Through the isomorphic fusion of bimodal optical imaging and hydrodynamic parameters, the non-real-time or monitoring equipment damage problem of traditional contact measurement is avoided, and the real-time, stability and safety of monitoring are improved.

[0060] Image quality enhancement: The MAB module introduces a dynamic deformable convolution compensation mechanism in the shallow and middle layers of the backbone network, effectively solving the image blur problem caused by field environment (such as dust, water mist), and significantly improving the detail recovery ability of super-resolution images.

[0061] Long-distance dependence capture: The Mamba block optimizes the capture ability of long-distance dependence relationship through the dynamic mapping of state transition matrix and hydrodynamic parameters in the deep network, eliminates the modal misalignment caused by wave disturbance, and improves the clarity and recognition accuracy of target outline.

[0062] Multi-modal fusion optimization: The BiMamba fusion strategy and multi-source confidence weighting mechanism are used to realize the dynamic fusion of bright field / dark field images and hydrodynamic parameters, avoid the dimension explosion problem caused by splicing operation, and enhance the complementarity of multi-modal data and the environmental adaptability of the model.

[0063] Nonlinear dynamic modeling: Combined with hydrodynamic parameters (water level, flow rate), a nonlinear particle size and concentration estimation model is constructed to improve the estimation accuracy of mountain flood particle size and concentration in complex environments, and to make up for the complex environment situation that cannot be covered by a single parameter.

[0064] Overall, the method is superior to the prior art in image enhancement, feature extraction, long-distance dependence capture, and multi-modal fusion, and is suitable for mountain flood particle size identification and detection in complex field environments, improving the reliability and accuracy of engineering applications.

[0065] Please refer to Figure 2 , which shows a structural block diagram of a mountain flood particle size and concentration analysis system of the present application.

[0066] As Figure 2 shown, the mountain flood particle size and concentration analysis system 200 includes an acquisition module 210, an output module 220, a determination module 230, and a calculation module 240.

[0067] The acquisition module 210 is configured to acquire a surface image of a mountain torrent, water level data and flow rate data, and identify the surface image according to a target detection algorithm to obtain at least one region image, wherein each region image contains mountain torrent particles.

[0068] It should be understood that Figure 2 the modules described in the above Figure 1 correspond to the steps in the methods described in the above Figure 2 The operations and features described above for the methods also apply to the modules in the , and will not be described here again.

[0069] In some other embodiments, the present application also provides a computer readable storage medium having a computer program stored thereon, wherein the program instructions are executed by a processor to cause the processor to perform the mountain torrent particle size and concentration analysis method in any of the above method embodiments.

[0070] As an implementation form, the computer readable storage medium of the present application stores computer executable instructions, and the computer executable instructions are configured to:

[0071] acquire a surface image of a mountain torrent, water level data and flow rate data, and identify the surface image according to a target detection algorithm to obtain at least one region image, wherein each region image contains mountain torrent particles.

[0072] input each region image, the water level data and the flow rate data into a preset YOLOv12-MD network, and the YOLOv12-MD network outputs high pixel values and wide pixel values of all mountain torrent particles in each region image.

[0073] determine the particle size of each mountain torrent particle according to the water level data, the flow rate data and the high pixel value and the wide pixel value of each mountain torrent particle by using a preset mountain torrent particle size estimation strategy, and obtain the particle size of each mountain torrent particle.

[0074] The concentration of flash flood particles in the flash flood is calculated based on the particle size of each flash flood particle and a preset empirical relationship model of surface volume.

[0075] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the flash flood particle size and concentration analysis system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely configured relative to a processor, which can be connected to the flash flood particle size and concentration analysis system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0076] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the flash flood particle size and concentration analysis method described in the above embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the flash flood particle size and concentration analysis system. The output device 340 may include a display screen or other display device.

[0077] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0078] In one implementation, the above-described electronic device is used in a flash flood particle size and concentration analysis system for a client application, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0079] Obtain a surface image of the mountain torrent, water level data and flow rate data, and identify the surface image according to a target detection algorithm to obtain at least one region image, wherein each region image contains mountain torrent particles;

[0080] Input each region image, the water level data and the flow rate data into a preset YOLOv12-MD network, and the YOLOv12-MD network outputs high pixel values and wide pixel values of all mountain torrent particles in each region image;

[0081] According to the water level data, the flow rate data and the high pixel values and the wide pixel values of each mountain torrent particle, a preset mountain torrent particle size estimation strategy is used to determine the particle size of each mountain torrent particle to obtain the particle size of each mountain torrent particle.

[0082] According to the particle size of each mountain torrent particle and a preset surface volume empirical relationship model, the concentration of the mountain torrent particles in the mountain torrent is calculated.

[0083] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and the necessary general hardware platform, and of course, it can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method of each embodiment or some parts of the embodiment.

[0084] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A flash flood particle size and concentration analysis method, characterized by, The method comprises: acquiring surface images, water level data and flow rate data of the mountain torrent, wherein the surface images contain mountain torrent particles; inputting the surface images into a preset YOLOv12-MD network, wherein the YOLOv12-MD network outputs high pixel values and wide pixel values of all the mountain torrent particles in the surface images, and the YOLOv12-MD network contains a shallow backbone network, a middle backbone network and a deep backbone network; wherein the shallow backbone network and the middle backbone network both contain a dynamic deformable convolution compensation mechanism, and the expression is as follows: ， In the formula, is the output feature map of the dynamic deformable convolution, is a 7x7 convolution operation on the offset matrix, is an offset matrix, is a 3x3 convolution on the spatial attention weight matrix, is a spatial attention weight matrix, is a 3x3 convolution on the spatial attention weight matrix, and then normalized using a Softmax function, is a Hadamard product operation, is an input feature, H is the number of height pixels, W is the number of width pixels, and C is the number of channels. the deep backbone network contains a dynamic mapping relationship between a state transition matrix and environmental parameters, wherein the environmental parameters include water level and flow rate, and the expression of the dynamic mapping relationship is as follows: ， ， wherein, is the state transition matrix after dynamic adjustment, is the initial state transition matrix, is the water level influence weight matrix, is the real-time water level height, is the flow rate influence weight matrix, is the real-time flow rate, is the water level-flow rate interaction weight matrix, is the Kronecker product; determining the particle sizes of the mountain torrent particles according to the water level data, the flow rate data and the high pixel values and the wide pixel values of the mountain torrent particles by using a preset mountain torrent particle size estimation strategy, to obtain the particle sizes of the mountain torrent particles; calculating the concentration of the mountain torrent particles in the mountain torrent according to the particle sizes of the mountain torrent particles and a preset surface volume empirical relationship model.

2. The method according to claim 1, wherein, The expression of the mountain torrent particle size estimation strategy is as follows: ， ， ， ， wherein, is the particle size of the th flash flood particle, is the pixel size of the th flash flood particle, is the initial refraction correction factor, is the flow rate correction gain, is the band attenuation coefficient estimate under bright field, is the band attenuation coefficient estimate under dark field, is the high pixel value of the th flash flood particle, is the wide pixel value of the th flash flood particle, is the experimental calibration coefficient, is the exponential decay function, is the learnable parameter under bright field, is the real-time flow rate, is the real-time water level height, is the initial band attenuation coefficient.

3. The method according to claim 1, wherein, The calculation of the concentration of the mountain torrent particles in the mountain torrent according to the particle sizes of the mountain torrent particles and the preset surface volume empirical relationship model comprises: calculating the particle mass of a single mountain torrent particle, and the expression is as follows: ， wherein is the mass of the nth flood particle, is the diameter of the nth flood particle, is the density of the sediment particles, taken as 2650 ;​​ calculating the concentration of the mountain torrent particles in the mountain torrent according to the particle mass of the single mountain torrent particle and the surface volume empirical relationship model, and the expression for calculating the concentration of the mountain torrent particles is as follows: ， ， ， wherein, is the volume of the flash flood region within the surface image, is the surface area of the flash flood region within the surface image, , are experimental calibration coefficients, is the flow rate impact factor on the suspended, is the real-time flow rate, is the real-time water level height, L is the total number of particles, is the median particle size.

4. A flash flood particle size and concentration analysis system, characterized by, The method comprises: an acquisition module configured to acquire surface images, water level data and flow rate data of the mountain torrent, wherein the surface images contain mountain torrent particles; an output module configured to input the surface images into a preset YOLOv12-MD network, wherein the YOLOv12-MD network outputs high pixel values and wide pixel values of all the mountain torrent particles in the surface images, and the YOLOv12-MD network contains a shallow backbone network, a middle backbone network and a deep backbone network; wherein the shallow backbone network and the middle backbone network both contain a dynamic deformable convolution compensation mechanism, and the expression is as follows: ， In the formula, is the output feature map of the dynamic deformable convolution, is a 7x7 convolution operation on the offset matrix, is an offset matrix, is a 3x3 convolution on the spatial attention weight matrix, is a spatial attention weight matrix, is a 3x3 convolution on the spatial attention weight matrix, and then normalized using a Softmax function, is a Hadamard product operation, is an input feature, H is the number of height pixels, W is the number of width pixels, and C is the number of channels. the deep backbone network contains a dynamic mapping relationship between a state transition matrix and environmental parameters, wherein the environmental parameters include water level and flow rate, and the expression of the dynamic mapping relationship is as follows: ， ， wherein, is the state transition matrix after dynamic adjustment, is the initial state transition matrix, is the water level influence weight matrix, is the real-time water level height, is the flow rate influence weight matrix, is the real-time flow rate, is the water level-flow rate interaction weight matrix, is the Kronecker product; a determination module configured to determine the particle sizes of the mountain torrent particles according to the water level data, the flow rate data and the high pixel values and the wide pixel values of the mountain torrent particles by using a preset mountain torrent particle size estimation strategy, to obtain the particle sizes of the mountain torrent particles; a calculation module configured to calculate the concentration of the mountain torrent particles in the mountain torrent according to the particle sizes of the mountain torrent particles and a preset surface volume empirical relationship model.

5. An electronic device, comprising: The method comprises: at least one processor, and a memory connected with the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 3.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the method of any one of claims 1 to 3. The program, when executed by a processor, implements the method of any one of claims 1 to 3.

Images