Dredging engineering multi-beam data rapid processing method based on AI

Abstract

The invention provides a dredging engineering multi-beam data rapid processing method based on AI, and belongs to the field of dredging engineering. According to the method, when a feature matrix Xalpha is extracted, data information of beams of the same angle is directly extracted to form the feature matrix Xalpha, and a plurality of feature matrixes Xalpha are formed through data information of beams of different angles; training the RNN neural network through a plurality of feature matrixes X alpha, and outputting a result, wherein each matrix can intuitively reflect water depth and geological information of reflection points at the same angle; the difference between independent variables is increased through matrixes of different angles, the RNN training speed is increased, and an RNN model can be trained more quickly; and meanwhile, setting of each element parameter in the output matrix facilitates batch mapping of reflection points to a GPS in the later period, and a submarine geological distribution map is made through three-dimensional software.

Description

Technical Field

[0001] The present invention relates to the field of dredging engineering, and in particular to an AI-based fast processing method for multi-beam data of dredging engineering. Background Art

[0002] Dredging engineering refers to the artificial excavation, silt removal, widening and other operations on water bodies such as rivers, lakes, reservoirs and ports, with the aim of improving the smoothness of water flow, increasing navigation capacity, preventing floods, improving water quality, etc. With the acceleration of global urbanization and the increase in transportation demand, dredging engineering plays an increasingly important role in the construction of waterways, ports and water conservancy facilities. At present, dredging technology at home and abroad is constantly being updated and developed, and the equipment and methods used are becoming more and more advanced, such as hydraulic dredging, mechanical dredging and combined dredging technology. However, no matter what method is used, it is necessary to detect the terrain conditions of the seabed.

[0003] At present, the technologies for seabed detection mainly include: sonar detection technology, laser scanning technology and remote sensing technology; among them, sonar detection technology is the main technology for seabed detection of dredging projects. Sonar detection technology includes single-beam detection and multi-beam detection. Since single-beam detection is inefficient when it is necessary to detect the terrain distribution of a larger seabed area, multi-beam detection is usually used to detect the seabed terrain; multi-beam detection uses a transducer array to form hundreds of fan-shaped sound waves to detect the seabed terrain and analyze the depth and geological distribution of the seabed; but the data obtained usually need to be processed manually, which is very labor-intensive and time-consuming; therefore, how to improve the processing efficiency and accuracy of multi-beam detection data is an important task for dredging projects. Summary of the invention

[0004] The purpose of the present invention is to overcome the above problems existing in the prior art and greatly improve its technical effect on the basis of the original technology; the present invention provides a method for fast processing of multi-beam data of dredging engineering based on AI, the method comprising: First, obtain the feature set: perform multiple multi-beam measurements on the target dredging area through the multi-beam bathymetry system. The multiple multi-beam measurements refer to scanning the entire target dredging area multiple times according to different fan-shaped angle ranges formed by the transducer array; after each scan of the entire target dredging area is completed, extract the data information of the same angle beam received by the receiving end to form a feature matrix X α =[X1, X2, …, X i , …, X n ]; where X α Refers to the matrix composed of beams with an angle of α after one scan, X i is the vertical vector, X iThe elements in include: the angle of the corresponding beam, reflection speed, reflection intensity and reflection time; and according to the obtained data, the water depth when the corresponding beam is reflected, the horizontal distance to the launch point and the geological distribution are obtained to form a matrix X α The corresponding feature matrix Y α =[Y1, Y2, ..., Y i , …, Y n ]; among them, Y i is the vertical vector, Y i The elements in include: water depth, horizontal distance to the launch point and geological classification; All feature matrices X α As input, the corresponding Y α The recurrent neural network RNN ​​is trained as output to obtain a trained RNN model; When the target dredging area needs to be detected, the target detection area is scanned by forming a suitable angle range through the transducer array to obtain the scanned data; according to the received data, the characteristic matrix X of the same angle beam is extracted respectively. β , X β and the matrix X α have the same matrix form; X β Input into the trained RNN model to get the output result Y β ; According to the output result Y β The water depth measured by the corresponding beam, the horizontal distance to the launch point and the geological classification are combined with the position of the launch source when the corresponding beam is launched to calculate and map the matrix Y β The position information corresponding to each vertical vector, the entire matrix Y β That is, the position scatter points of all vertical vectors are formed; all matrices Y are obtained β The position scatter points, according to the matrix Y corresponding to the position scatter points β The water depth and geological information in the seabed are used to produce a seabed geological distribution map.

[0005] In addition, the multi-beam bathymetry system includes: the multi-beam bathymetry system can simultaneously emit multiple sound waves through the transducer array, and the emitted multi-beam sound waves are arranged in a fan shape. The fan-shaped plane is perpendicular to the navigation direction of the ship carrying the multi-beam bathymetry system, and the emitted multi-channel sound waves are arranged at different angles to the vertical downward direction, and the angle size is ≤ half of the fan-shaped angle; multiple multi-beam measurements refer to each setting of a fan-shaped arrangement of the transducer array, and the corresponding fan-shaped arrangement is required to complete multiple scans of the target dredging area. As the number of completions increases, the acquired feature set will also increase, and the trained AI model will be more reliable.

[0006] In addition, the feature matrix X α Includes: feature matrix X αThe vector in is a vertical vector. The elements of the vertical vector include: angle, reflection speed, reflection intensity and reflection time information. The expression of the vertical vector is: Among them, X i is the feature matrix X α The i-th vertical vector in i For X i The corresponding sound wave reflection angle, v i For X i The corresponding sound wave reflection velocity, σ i For X i The corresponding sound wave reflection intensity, △t i is the time difference between the transmission and reception of the sound wave; X α All sound waves in the same reflection angle, so α1 = α2 = ... α i …=α n ; In addition, the water depth corresponding to the beam reflection, the horizontal distance to the launch point and the geological distribution are obtained based on the obtained data, including: The formula for obtaining the water depth corresponding to the beam reflection is: ; The formula for obtaining the horizontal distance of the corresponding beam reflection point and the emission point is: ; The method for obtaining the geological distribution is: determining the geological classification of the corresponding beam by the empirical threshold classification method; the characteristic matrix Y α Y i The expression is: Among them, H i Represents Sound Wave X i Corresponding water depth information, L i is the horizontal distance from the beam reflection point to the emission point, C represents the sound wave X i The geological classification at the time of reflection, the classification of C is determined by the empirical threshold classification method.

[0007] In addition, the target detection area is scanned by forming a suitable angle range through the transducer array, including: the suitable angle range refers to the angle of the fan-shaped light beam formed by the transducer array; the suitable angle range is determined according to the shape and area of ​​the target detection area, and the angle of the fan-shaped light beam formed by the transducer array is determined by analyzing the shape and area of ​​the target detection area without affecting the detection accuracy, and the determined angle is maintained to scan the entire target detection area.

[0008] In addition, the matrix Y is calculated and mapped βThe position information corresponding to each vertical vector includes: β is the angle between the light beam and the vertical downward direction; the position information corresponding to each vertical vector refers to the position information of the light beam reflection point; the steps for determining the corresponding position information are: first, obtain the position information of the light beam emission point, and map the emission point position information to the GPS map; then, through the matrix Y β Obtain the vertical vector corresponding to the light beam, and obtain the water depth of the corresponding light beam reflection position and the horizontal distance to the launch point through the vertical vector; finally, according to the position information of the launch point, combined with the triangular relationship formed by the water depth of the corresponding light beam reflection position and the horizontal distance to the launch point, determine the position information of the light beam launch point, map the position information to the GPS map, and mark the water depth information.

[0009] In addition, get all matrices Y β The position scattering includes: Get the matrix Y of all beams β , and find each matrix Y β The position scatter points of all vertical vectors in ; according to the matrix Y corresponding to the position scatter points β The water depth and geological information are collected and three-dimensional software is used to create a seabed geological distribution map.

[0010] The beneficial effects of the present invention are:

[0011] The present invention provides a method for fast processing of multi-beam data of dredging engineering based on AI, which has the following advantages: 1. This method transforms the feature matrix X α and the feature matrix Y α The recurrent neural network RNN ​​is trained as input and output. After the trained model is obtained, the feature matrix X of the target dredging area is obtained in real time. β Input the trained model; directly obtain the water depth of the sound wave reflection point in the target dredging area, the horizontal distance to the launch point and the geological distribution according to the output of the model; greatly increase the efficiency of multi-beam data processing.

[0012] 2. This method extracts the feature matrix X α When the data information of the same angle beam is directly extracted to form the feature matrix X α , then the data information of beams at different angles constitutes multiple feature matrices X α ; Through multiple feature matrices X αThe RNN neural network is trained. When the trained network detects the target area, each matrix of the output results can intuitively reflect the water depth and geological information of the reflection points at the same angle. Matrices with different angles increase the differences between independent variables, speed up the training of the RNN network, and improve the efficiency of data processing. At the same time, the setting of the parameters of each element in the output matrix facilitates the subsequent mapping of reflection points to GPS in batches, as well as the production of seabed geological distribution maps through three-dimensional software. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flowchart of a method for rapidly processing multi-beam data of dredging projects based on AI. DETAILED DESCRIPTION

[0014] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings; it should be understood that the specific embodiments given here are only used to illustrate and explain the present invention and cannot be used to limit the present invention.

[0015] It should be noted that many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention may also have other implementations and variations thereof. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0016] like Figure 1 As shown, a flowchart of a method for fast processing multi-beam data of dredging engineering based on AI according to an embodiment of the present invention is shown; the flowchart comprises: step S1, first, obtaining a feature set: performing multiple multi-beam measurements on the target dredging area through a multi-beam bathymetric system, wherein the multiple multi-beam measurements refer to scanning the entire target dredging area multiple times according to different fan-shaped angle ranges formed by the transducer array; after each scan of the entire target dredging area is completed, extracting data information of the same angle beam received by the receiving end to form a feature matrix X α =[X1, X2, …, X i , …, X n ]; where X α Refers to the matrix composed of beams with an angle of α after one scan, X i is the vertical vector, X i The elements in include: the angle of the corresponding beam, reflection speed, reflection intensity and reflection time; and according to the obtained data, the water depth when the corresponding beam is reflected, the horizontal distance to the launch point and the geological distribution are obtained to form a matrix X α The corresponding feature matrix Y α =[Y1, Y2, ..., Y i , …, Y n ]; among them, Y i is the vertical vector, Y iThe elements in include: water depth, horizontal distance to the launch point and geological classification; Step S2, all the feature matrices X α As input, the corresponding Y α As output, the recurrent neural network RNN ​​is trained to obtain the trained RNN model; Step S3, when the target dredging area needs to be detected, the target detection area is scanned by forming a suitable angle range through the transducer array to obtain the scanned data; according to the received data, the feature matrix X of the same angle beam is extracted respectively β , X β and the matrix X α have the same matrix form; X β Input into the trained RNN model to get the output result Y β ; Step S4, according to the output result Y β The water depth measured by the corresponding beam, the horizontal distance to the launch point and the geological classification are combined with the position of the launch source when the corresponding beam is launched to calculate and map the matrix Y β The position information corresponding to each vertical vector, the entire matrix Y β That is, the position scatter points of all vertical vectors are formed; all matrices Y are obtained β The position scatter points, according to the matrix Y corresponding to the position scatter points β The water depth and geological information in the seabed are used to produce a seabed geological distribution map.

[0017] Specifically, in step S1, the multi-beam echo sounder system can simultaneously emit multiple sound waves through the transducer array. The emitted multi-beam sound waves are arranged in a fan shape. The fan plane is perpendicular to the navigation direction of the ship carrying the multi-beam echo sounder system. The emitted multi-channel sound waves are arranged at different angles to the vertical downward direction, and the angle is ≤ half of the fan angle; multiple multi-beam measurements refer to each setting of a fan arrangement of the transducer array, and the corresponding fan arrangement is required to complete multiple scans of the target dredging area. As the number of completions increases, the acquired feature set will also increase, and the trained AI model will be more reliable.

[0018] Feature matrix X α The vector in is a vertical vector. The elements of the vertical vector include: angle, reflection speed, reflection intensity and reflection time information. The expression of the vertical vector is: Among them, X i is the feature matrix X α The i-th vertical vector in i For X i The corresponding sound wave reflection angle, v i For X i The corresponding sound wave reflection velocity, σ i For X i The corresponding sound wave reflection intensity, △ti is the time difference between the transmission and reception of the sound wave; X α All sound waves in the same reflection angle, so α1 = α2 = ... α i …=α n .

[0019] Specifically, the water depth at the time of corresponding beam reflection, the horizontal distance to the transmitting point and the geological distribution include: The formula for obtaining the water depth at the time of corresponding beam reflection is: ; The formula for obtaining the horizontal distance of the corresponding beam reflection point and the emission point is: ; The method for obtaining the geological distribution is: determine the geological classification of the corresponding beam through the empirical threshold classification method; the characteristic matrix Y α Y i The expression is: Among them, H i Represents Sound Wave X i Corresponding water depth information, L i is the horizontal distance from the beam reflection point to the emission point, C represents the sound wave X i The geological classification at the time of reflection, the classification of C is determined by the empirical threshold classification method.

[0020] In step S2, the recurrent neural network RNN ​​is a neural network model with memory characteristics, which is suitable for processing sequence data and can process data of feature matrix; therefore, the feature matrix X is processed by RNN. α Train and output Y α As a standard, the weights during RNN training are adjusted to obtain the trained RNN model.

[0021] In step S3, the appropriate angle range refers to the angle of the fan-shaped beam formed by the transducer array; the appropriate angle range is determined based on the shape and area of ​​the target detection area. Without affecting the detection accuracy, the angle of the fan-shaped beam formed by the transducer array is determined by analyzing the shape and area of ​​the target detection area, and the determined angle is maintained to scan the entire target detection area.

[0022] In step S4, the matrix Y is calculated and mapped β The position information corresponding to each vertical vector includes: β is the angle between the light beam and the vertical downward direction; the position information corresponding to each vertical vector refers to the position information of the light beam reflection point; the steps for determining the corresponding position information are: first, obtain the position information of the light beam emission point, and map the emission point position information to the GPS map; then, through the matrix Y βObtain the vertical vector corresponding to the light beam, and obtain the water depth of the corresponding light beam reflection position and the horizontal distance to the launch point through the vertical vector; finally, according to the position information of the launch point, through the trigonometric relationship: combine the horizontal distance from the corresponding light beam reflection position to the launch point to determine the plane position information of the reflection point, and then determine the position information of the light beam reflection point through the plane position information combined with the water depth information; and map the position information to the GPS map and mark the water depth information.

[0023] Specifically, the method of obtaining all matrices Y β The position scattering includes: obtaining the matrix Y of all beams at different angles β , and find each matrix Y β The position scatter points of all vertical vectors in ; according to the matrix Y corresponding to the position scatter points β The water depth and geological information are collected and three-dimensional software is used to create a seabed geological distribution map.

Claims

1. A fast processing method for multi-beam data of dredging engineering based on AI, characterized in that: The method comprises: First, obtain the feature set: perform multiple multi-beam measurements on the target dredging area through the multi-beam bathymetry system. The multiple multi-beam measurements refer to scanning the entire target dredging area multiple times according to different fan-shaped angle ranges formed by the transducer array; after each scan of the entire target dredging area is completed, extract the data information of the same angle beam received by the receiving end to form a feature matrix X α =[X1, X2, …, X i , …, X n ]; where X α Refers to the matrix composed of beams with an angle of α, X i is the vertical vector, X i The elements in include: the angle of the corresponding beam, the reflection speed, the reflection intensity and the reflection time; and according to X i Obtain the water depth corresponding to the beam reflection point, the horizontal distance to the launch point, and the geological classification, which is composed of X α The corresponding feature matrix Y α =[Y1, Y2, ..., Y i , …, Y n ]; among them, Y i is the vertical vector, Y i The elements in refer to water depth, horizontal distance to the launch point, and geological classification; All feature matrices X α and Y α Train the recurrent neural network RNN ​​as input and output to obtain a trained RNN model; The target dredging area is detected in real time, and the target detection area is scanned by forming a suitable angle through the transducer array; according to the received data, the characteristic matrix X of the same angle beam is extracted respectively. β , β is the angle between the light beam and the vertical downward direction, X β and the matrix X α have the same matrix form; X β Input into the trained RNN model to get the output result Y β ; According to the output result Y β The water depth of the corresponding beam, the horizontal distance to the launch point and the geological classification are combined with the position of the launch source when the corresponding beam is launched to calculate and map the matrix Y β The position information corresponding to each vertical vector, the entire matrix Y β That is, the position scatter points of all vertical vectors are formed; all matrices Y are obtained β The position scatter points, according to the matrix Y corresponding to the position scatter points β The water depth and geological information in the seabed are used to produce a seabed geological distribution map.

2. According to the AI-based fast processing method for multi-beam data of dredging engineering according to claim 1, it is characterized in that: The multi-beam bathymetry system includes: the multi-beam bathymetry system can emit multiple sound waves simultaneously through a transducer array, the emitted multiple sound waves are arranged in a fan shape, the fan plane is perpendicular to the navigation direction of the ship carrying the multi-beam bathymetry system, and the emitted multiple sound waves are arranged at different angles to the vertical downward direction, and the angle is ≤ half of the fan angle; multiple multi-beam measurements refer to each setting of a fan arrangement of the transducer array, and the corresponding fan arrangement is required to complete multiple scans of the target dredging area. As the number of completions increases, the acquired feature set will also increase, and the trained AI model will be more reliable.

3. According to the AI-based fast processing method for multi-beam data of dredging engineering according to claim 1, it is characterized in that: The feature matrix X α Includes: feature matrix X α The vector in is a vertical vector. The elements of the vertical vector include: angle, reflection speed, reflection intensity and reflection time information. The expression of the vertical vector is: Among them, X i is the feature matrix X α The i-th vertical vector in i For X i The corresponding sound wave reflection angle, v i For X i The corresponding sound wave reflection velocity, σ i For X i The corresponding sound wave reflection intensity, △t i X is the time difference between the transmission and reception of sound waves; α All sound waves in the same reflection angle, so α1 = α2 = ... α i …=α n ; Said and according to X i The water depth at the corresponding beam reflection point, the horizontal distance to the launch point, and the geological classification are obtained: The formula for obtaining the water depth at the corresponding beam reflection point is: ; The formula for obtaining the horizontal distance from the corresponding beam reflection point to the transmitting point is: ; The method for obtaining the geological distribution is: determining the geological classification of the corresponding beam by the empirical threshold classification method; the characteristic matrix Y α Y i The expression is: Among them, H i Represents Sound Wave X i Corresponding water depth information, L i is the horizontal distance from the beam reflection point to the emission point, C represents the sound wave X i The geological classification at the time of reflection, the classification of C is determined by the empirical threshold classification method.

4. According to the AI-based fast processing method for multi-beam data of dredging engineering in claim 1, it is characterized in that: The method of scanning the target detection area by forming a suitable angle through the transducer array includes: the suitable angle refers to the angle of the fan-shaped light beam formed by the transducer array; the suitable angle is determined according to the shape and area of ​​the target detection area, and the angle of the fan-shaped light beam formed by the transducer array is determined by analyzing the shape and area of ​​the target detection area without affecting the detection accuracy, and the determined angle is maintained to scan the entire target detection area.

5. According to the AI-based fast processing method for multi-beam data of dredging engineering in claim 1, it is characterized in that: The calculation and mapping of the matrix Y β The position information corresponding to each vertical vector includes: β is the angle between the light beam and the vertical downward direction; the position information corresponding to each vertical vector refers to the position information of the light beam reflection point; the steps for determining the corresponding position information are: first, obtain the position information of the light beam emission point, and map the emission point position information to the GPS map; then, through the matrix Y β Obtain the vertical vector corresponding to the light beam, and obtain the water depth of the corresponding light beam reflection position and the horizontal distance to the launch point through the vertical vector; finally, according to the position information of the launch point, combined with the triangular relationship formed by the water depth of the corresponding light beam reflection position and the horizontal distance to the launch point, determine the position information of the light beam launch point, map the position information to the GPS map, and mark the water depth information.

6. The AI-based fast processing method for multi-beam data of dredging engineering according to claim 1 is characterized in that: Get all matrices Y β The position scattering includes: Get the matrix Y of all beams β , and find each matrix Y β The position scatter points of all vertical vectors in ; according to the matrix Y corresponding to the position scatter points β The water depth and geological information are collected and three-dimensional software is used to create a seabed geological distribution map.

Images