Underwater laser echo signal processing model training method and water depth measurement method

Abstract

The application discloses a water laser echo signal processing model training method and a water depth measuring method, and belongs to the field of laser radar water depth detection. The method comprises the following steps: obtaining multiple groups of underwater laser echo signals and preprocessing to obtain an underwater laser echo signal dataset; the preprocessing comprises the following steps: after normalizing each group of underwater laser echo signals, labeling the data points located on the rising edge of the water surface waveform as water surface points, labeling the data located on the falling edge of the water bottom waveform as water bottom points, and labeling the remaining data points as noise points; building a classification model for predicting the types of the data points in the underwater laser echo signal and training the classification model by using the underwater laser echo signal dataset to obtain an underwater laser echo signal processing model. The application can accurately identify the data points located on the rising edge of the water surface waveform and the falling edge of the water bottom waveform in a complex environment, and provides a reliable basis for calculating the time difference between the water bottom points and the water surface points, thereby improving the measuring accuracy of the water depth.

Description

Technical Field

[0001] This invention belongs to the field of lidar water depth detection, and more specifically, relates to a method for training an underwater laser echo signal processing model and a method for measuring water depth. Background Technology

[0002] The ocean covers a vast area, approximately 71% of the Earth's surface. It provides the necessary physical environment for the sustainability of ecosystems and biodiversity, and is also crucial for human survival and well-being. Water depth measurement is fundamental to marine resource exploration. High-resolution depth sounding data not only contributes to research on the ecological environment and climate change, but also plays a vital role in environmental monitoring and protection, as well as the rational utilization of resources. Traditional ocean depth sounding methods mainly include acoustic sounding and electromagnetic sounding. Acoustic sounding measures water depth by emitting sound waves and receiving reflected signals, while electromagnetic sounding utilizes the reflected information of electromagnetic waves. While these methods can obtain accurate seabed topographic data, they suffer from limitations such as limited measurement range, slow response time, and susceptibility to environmental factors.

[0003] In contrast, lidar depth sounding, with its rapid response and adaptability to various environments, has been widely used in marine exploration. Lidar systems can acquire complete waveform information and extract ocean depth information based on the echo signals, enabling high-precision, large-scale three-dimensional seabed topographic mapping. However, laser pulse energy decays exponentially with transmission depth, and the attenuation rate is positively correlated with the turbidity of the water column. This means that lidar systems can only obtain clear surface and seabed echo signals at appropriate depths and in areas with high seawater clarity. In deeper or more turbid waters, only seabed echo signals with low signal-to-noise ratios are obtained. Traditional laser echo sounding techniques cannot effectively detect these weak seabed echo signals because more information about the signal waveform is needed for seabed data detection. Furthermore, in shallower waters, surface and seabed echoes tend to overlap due to the short time interval between them. This overlap increases the difficulty of separating and extracting surface and seabed waveforms, reducing the accuracy and reliability of depth measurements. Therefore, addressing the low signal-to-noise ratio and waveform overlap becomes a key issue in waveform processing.

[0004] To address the challenges encountered in waveform extraction, researchers have studied laser echo characteristics and water properties, proposing various waveform processing methods. For example, Saylam et al. used a moving average filtering algorithm to smooth laser waveforms; Wang et al., through experimental comparison, concluded that the Richardson-Lucy deconvolution (RLD) method is better at extracting waveforms from weak echo signals, while the average variance function (ASDF) can better handle noise. Pan et al. proposed a continuous wavelet transform (CWT) algorithm to solve the problem of overlapping surface and seabed echoes in shallow water areas. These methods offer good solutions to different problems in waveform extraction; however, this also leads to a lack of uniformity in laser echo processing methods. For echo signals in a specific area, manual classification of the echo signals is necessary, followed by selection of different algorithms for depth extraction, increasing time and manpower costs. Compared to traditional methods, neural networks in deep learning can automatically acquire waveform feature information and correlation information between adjacent waveforms, demonstrating great potential in laser echo signal processing. Hu et al. constructed a multi-layer fully connected neural network (FCN) and a one-dimensional convolutional neural network (CNN) to classify laser echo signals into land and ocean categories with an accuracy of 95.6%. Xu et al. also used deep learning methods to classify laser echo signals into five categories, and then processed the waveform signals according to these categories. However, these methods focus on using deep learning models to classify the overall waveform of laser echo signals, and their adaptability to complex aquatic environments is limited, thus compromising the accuracy of water depth measurement. Summary of the Invention

[0005] To address the shortcomings and improvement needs of existing technologies, this invention provides a training method for an underwater laser echo signal processing model and a method for measuring water depth. The aim is to propose a novel deep learning model for underwater laser echo signal processing, based on the characteristic that water depth measurement depends on the time difference between the bottom and surface points. This model is used to accurately identify data points located on the rising edge of the surface waveform and the falling edge of the bottom waveform in complex environments, providing a reliable basis for calculating the time difference between the bottom and surface points, thereby improving the accuracy of water depth measurement.

[0006] To achieve the above objectives, according to one aspect of the present invention, a method for training an underwater laser echo signal processing model is provided, comprising:

[0007] Multiple sets of underwater laser echo signals were acquired and preprocessed to obtain an underwater laser echo signal dataset. Each set of underwater laser echo signals included multiple data points, each of which included time and echo intensity. The preprocessing included: normalizing each set of underwater laser echo signals, labeling data points located on the rising edge of the surface waveform as surface points, data points located on the falling edge of the bottom waveform as bottom points, and the remaining data points as noise points.

[0008] A classification model based on deep learning was built to predict the type of each data point in the underwater laser echo signal. The classification model was trained using the underwater laser echo signal dataset. After training, the underwater laser echo signal processing model was obtained.

[0009] Furthermore, the classification model is a CNN-LSTM deep learning model, which includes a convolutional neural network, a long short-term memory network, a fully connected network, and a softmax function layer connected in sequence;

[0010] Convolutional neural networks are used to extract local features of each data point in underwater laser echo signals;

[0011] Long Short-Term Memory (LSTM) networks are used to capture the correlation between data points in underwater laser echo signals and obtain the semantic features of each data point.

[0012] Fully connected networks are used to map the semantic features of each data point in an underwater laser echo signal to the output category features of each data point;

[0013] The Softmax function layer is used to convert the output class features of each data point into the probability that the data point belongs to each class.

[0014] Furthermore, the convolutional neural network includes five one-dimensional convolutional layers, each consisting of a convolution operation, batch normalization, and the ReLU activation function;

[0015] Long Short-Term Memory (LSTM) networks consist of an LSTM layer, a batch normalization layer, and the ReLU activation function.

[0016] According to another aspect of the present invention, an underwater laser echo signal processing model product is provided, including an underwater laser echo signal processing model; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method provided by the present invention.

[0017] According to another aspect of the present invention, a method for measuring water depth based on underwater laser echo signals is provided, comprising:

[0018] Acquire underwater laser echo signals of the water body to be measured, normalize them, and input them into the underwater laser echo signal processing model to obtain the type of each data point;

[0019] The peak echo intensity and corresponding time T1 at the water surface point and the peak echo intensity and corresponding time T2 at the bottom point are obtained respectively. The depth H of the water body to be measured is calculated according to H=v*(T2-T1).

[0020] Where v represents the propagation speed of the laser in the water body to be tested; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method provided by the present invention.

[0021] According to another aspect of the present invention, a computer program product is provided, comprising a computer program; when the computer program is executed by a processor, it implements the underwater laser echo signal processing model training method provided by the present invention, and / or the water depth measurement method based on underwater laser echo signals provided by the present invention.

[0022] According to another aspect of the present invention, a computer-readable storage medium is provided, including a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the underwater echo signal processing model training method provided by the present invention, and / or the water depth measurement method based on underwater echo signals provided by the present invention.

[0023] According to another aspect of the present invention, an electronic device is provided, comprising:

[0024] A computer-readable storage medium for storing computer programs;

[0025] And a processor for reading a computer program in a computer-readable storage medium and executing the underwater laser echo signal processing model training method provided by the present invention, and / or the water depth measurement method based on underwater laser echo signals provided by the present invention.

[0026] According to another aspect of the present invention, a water depth measurement system is provided, comprising: a lidar and a measurement module;

[0027] LiDAR is used to acquire underwater laser echo signals of the water body being measured.

[0028] The measurement module is used to normalize the underwater laser echo signal and input it into the underwater laser echo signal processing model to obtain the type of each data point; it obtains the peak echo intensity and corresponding time T1 of the water surface point and the peak echo intensity and corresponding time T2 of the water bottom point, and calculates the depth H of the water body to be measured according to H=v*(T2-T1);

[0029] Where v represents the propagation speed of the laser in the water body to be tested; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method provided by the present invention.

[0030] In summary, the above-described technical solutions conceived in this invention can achieve the following beneficial effects:

[0031] (1) This invention makes full use of the advantages of deep learning methods. It uses a deep learning model to classify the data points in the underwater laser echo data. In the dataset used for model training, only the data points at the rising edge of the water surface waveform are marked as surface points and the data points at the falling edge of the bottom waveform are marked as bottom points. All other data points are marked as noise points. Thus, when the model processes underwater lasers, it can accurately identify the data points at the rising edge of the water surface waveform and the data points at the falling edge of the bottom waveform. It can also accurately determine the time corresponding to the bottom point and the water surface point in complex environments such as turbid water and overlapping water surface and bottom echoes, thereby improving the measurement accuracy of water depth.

[0032] (2) This invention uses the CNN-LSTM model as a classification model. This model can extract the feature information of data points and capture the correlation information between data points, effectively improving the accuracy of data point classification. Attached Figure Description

[0033] Figure 1 A schematic diagram of existing underwater laser echo signal data;

[0034] Figure 2 A flowchart of the underwater laser echo signal processing model training method provided in an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram of the classification model provided in an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the underwater laser echo signal processing results provided in an embodiment of the present invention; wherein (a), (b) and (c) represent the processing results of underwater laser echo signals in three different scenarios, respectively;

[0037] Figure 5 A schematic diagram of a water depth measurement method based on underwater laser echo signals provided in an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0039] In this invention, the terms "first," "second," etc. (if present) in the invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0040] By illuminating the water surface with a lidar sensor and collecting the corresponding echo signals, underwater laser echo signals can be obtained. Clear underwater laser echo signals can be acquired at appropriate depths and in areas with high seawater clarity. Figure 1 The diagram shows a waveform of an underwater laser echo signal. The surface waveform and the bottom waveform can be separated. The peak values ​​of the echo intensity correspond to the data points where the laser reaches the surface and the bottom, respectively. The time difference between these two data points represents the laser's propagation time in the water. By obtaining this time difference and combining it with the laser's propagation speed, the water depth can be calculated. However, in deeper or more turbid water, only seabed echo signals with a low signal-to-noise ratio can be obtained. In shallower water, the surface and seabed echoes are prone to overlap due to the short time interval between them. These complexities all affect the accuracy and reliability of bottom depth measurements.

[0041] To address the aforementioned problems, this invention provides a training method for an underwater laser echo signal processing model and a method for measuring water depth. The overall concept is based on the characteristic that water depth measurement depends on the time difference between the bottom and surface points. A novel deep learning model for underwater laser echo signal processing is proposed, enabling it to accurately identify data points located on the rising edge of the surface waveform and the falling edge of the bottom waveform in complex environments. This provides a reliable basis for calculating the time difference between the bottom and surface points, thereby improving the accuracy of water depth measurement.

[0042] The following is an example.

[0043] Example 1:

[0044] A training method for an underwater laser echo signal processing model, such as Figure 2 As shown, this includes steps S1-S2. The specific steps are as follows:

[0045] S1: Acquire multiple sets of underwater laser echo signals and preprocess them to obtain an underwater laser echo signal dataset;

[0046] It is easy to understand that during the scanning process, LiDAR divides the space into multiple small regions, each corresponding to a bin value. These bin values ​​contain information such as the distance and intensity measured by the LiDAR within that region. In this embodiment, the underwater laser echo signal consists of several bin values, each bin value corresponding to a data point, and each set of underwater laser echo signals can be viewed as a time series.

[0047] Considering that the data environment sampled at different times is not consistent, resulting in significant differences in signal strength due to complex noise distribution, the preprocessing of the underwater laser echo signal in this embodiment includes: normalizing the data points that make up the underwater laser echo signal to reduce its data range to the range of 0 to 1, in preparation for subsequent training.

[0048] To address the issue of overlapping surface and bottom waveforms, this embodiment further includes preprocessing by labeling data points located on the rising edge of the surface waveform as surface points, data points located on the falling edge of the bottom waveform as bottom points, and the remaining data points as noise points. Based on this data point labeling method, the trained model can accurately identify data points on the rising edge of the surface waveform and data points on the falling edge of the bottom waveform in the underwater laser echo signal.

[0049] S2: A classification model based on deep learning is built to predict the type of each data point in the underwater laser echo signal. The classification model is trained using the underwater laser echo signal dataset. After training, the underwater laser echo signal processing model is obtained.

[0050] Since underwater laser echo signals have certain temporal information, in order to improve the accuracy of data point classification, as a preferred implementation method, in this embodiment, the classification model is specifically a CNN-LSTM deep learning model, which includes a convolutional neural network, a long short-term memory network, a fully connected network, and a softmax function layer connected in sequence.

[0051] Convolutional Neural Networks (CNNs) are deep learning models specifically designed to process data with a grid-like structure, such as images. Through specialized layer structures like convolutional and pooling layers, CNNs effectively capture local features in images and achieve high-level representations and classifications of them.

[0052] Building upon CNNs' ability to capture spatial features of data, LSTMs can further capture temporal dependencies. CNNs excel at processing spatiotemporal data, effectively extracting local features through convolutional and pooling layers, and exhibiting translation invariance to adapt to features at different locations. LSTMs, through recurrent connections and gating mechanisms, capture long-term dependencies in data, enabling them to handle temporal information in sequential data. Secondly, the CNN-LSTM model can learn multi-level abstract features from the data. In waveform datasets, each point's value contains information about noise points, surface points, and bottom points, which often require different levels of feature representation to be accurately distinguished and classified. The multi-layered structure of the CNN-LSTM model can progressively extract higher-level features, from local to global features, allowing the model to better understand the data's inherent structure and semantic information. Furthermore, the CNN-LSTM model offers advantages in parameter sharing and parallel computation. Parameter sharing means that within the convolutional layers of a CNN, the same convolutional kernel can extract similar features at different locations, reducing the number of parameters and improving model efficiency. Parallel computing fully leverages the parallel computing capabilities of modern computing devices, accelerating model training and inference processes to meet the demands of large-scale data and real-time applications. Finally, CNN-LSTM models have achieved significant results in many sequential data domains, such as speech recognition, natural language processing, and action recognition. Data in these domains exhibits temporal and spatial correlations; therefore, CNN-LSTM models possess a priori advantage and superior performance in processing similar data. The primary reason for selecting the CNN-LSTM model as the backbone structure in this embodiment is its ability to effectively process sequential data with spatiotemporal relationships and to learn multi-level abstract feature representations. By combining the advantages of CNN and LSTM, this model can capture the spatial features and temporal dependencies of the data, adapting to the characteristics of the dataset and thus achieving accurate classification tasks.

[0053] like Figure 3 As shown, this embodiment uses five one-dimensional convolutional layers to extract features from the input data. To represent the local semantic information of the data after convolutional processing, the feature sequence after passing through a convolutional neural network (CNN) is input into an LSTM. The LSTM layer obtains the correlation between each data point and other data points, thereby better predicting the data point category. The features output by the LSTM are mapped to the final output category through a fully connected layer, and the classification probability is calculated using the Softmax function. Finally, each data point contains three feature information, representing the probability of the data point being a surface point, a bottom point, or a noise point. The category corresponding to the highest probability is the category of the data point, thus realizing the classification of laser echo data points.

[0054] In other words, the main structure of a deep learning model is a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network. The CNN uses five one-dimensional convolutional layers, each consisting of a convolution operation, batch normalization, and the ReLU activation function. Batch normalization normalizes the data for each batch, accelerating the network training process and providing regularization, thus improving the model's classification performance. Additionally, the ReLU activation function is used after each convolutional layer. The ReLU function sets the negative values ​​to zero while keeping the positive values ​​unchanged. Its purpose is to introduce non-linearity, increasing the model's expressive power. A CNN has one input channel and 32 output channels. After passing through the CNN, the input data transforms from one-dimensional intensity information into 32-dimensional feature information, thereby achieving feature extraction from the input data.

[0055] To represent the local semantic information of the data after convolution, the feature sequence after convolution and batch normalization is input into an LSTM. LSTM is a recurrent neural network suitable for processing sequential data. It has memory units and gating mechanisms, which can effectively capture and utilize long-term dependencies in sequential data. The correlation between each data point and other data points is obtained through the LSTM layer, thereby better predicting the data point category. The LSTM has 32 input channels and 64 output channels. Each data point in the output data contains not only its own feature information but also correlation information with other data points. The data then passes through a fully connected layer, which maps the features output by the LSTM to the final output category. The role of the fully connected layer is to linearly map and transform the high-dimensional features to obtain the final prediction result. The fully connected layer has 64 input channels and 3 output channels. The three pieces of information contained in each data point in the output data represent the weights of the three categories: surface point, bottom point, and noise point. After the fully connected layer, the Softmax function is used to calculate the classification probability. The Softmax function transforms the output of the fully connected layer into a probability distribution, representing the predicted probability of each category. Each data point contains three pieces of information representing the probability that the data point belongs to one of three categories: surface point, bottom point, or noise point. The category corresponding to the highest probability is the category of the data point, thus realizing the classification of underwater laser echo data points.

[0056] In step S2 of this embodiment, the classification model is trained using the underwater laser echo signal dataset, specifically including:

[0057] The underwater laser echo signal dataset was divided into a training set, a test set, and a validation set.

[0058] First, the CNN-LSTM model is trained using the training set. For each training set of data, the loss between the output data and the label is calculated using the loss function. After calculating the loss between the two, the model's internal parameters are updated by backpropagation through the loss, thereby reducing the loss.

[0059] After each training iteration, the updated deep learning model is tested on the test set to monitor changes in the model's evaluation metrics. The batch range of the dataset for the next training iteration is adjusted based on the accuracy on the test set. Optionally, if the accuracy does not exceed a certain threshold (e.g., 90%), the entire training set is used for the next training iteration. If the accuracy exceeds the threshold, the batch range of the training set data is adjusted accordingly. The training model then undergoes a test on the test set after each training iteration to observe changes in the accuracy of the classification model on the test set and thus discover the optimal solution.

[0060] Ultimately, the CNN-LSTM deep learning model with the best test results was selected as the echo data classification model for this study and validated using a validation set.

[0061] Figure 4 This paper demonstrates the processing results of the underwater laser echo signal processing model trained in this embodiment on underwater laser echo signals under different scenarios. (a), (b), and (c) correspond to three different scenarios, respectively. The left half of the figure represents the original underwater laser echo signal, and the right half represents the corresponding processing result. Figure 4 The processing results show that the underwater laser echo signal processing model trained in this embodiment can accurately identify data points located on the rising edge of the surface waveform and data points located on the falling edge of the bottom echo in the underwater laser echo signal, especially based on... Figure 4 As shown in (b) and (c), even when the surface and bottom waveforms overlap, this embodiment can accurately classify data points in the underwater laser echo signal. Experimental results show that the accuracy of data point classification in this embodiment reaches 97.62%. Compared with real water depth data, the root mean square error of the water depth value obtained by the CNN-LSTM model is 0.46m, achieving excellent water depth measurement results.

[0062] Overall, this embodiment can accurately identify data points located on the rising edge of the water surface waveform and data points located on the falling edge of the bottom echo in underwater laser echo signals in complex water environments. It finds the peak echo intensity at the water surface point and the bottom point respectively, and uses them as the water surface peak point and the bottom peak point. The water depth information can be obtained based on the time difference between the water surface peak point and the bottom peak point, thereby accurately improving the water depth measurement accuracy in complex water environments.

[0063] Example 2:

[0064] An underwater laser echo signal processing model product includes an underwater laser echo signal processing model; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method provided in Embodiment 1 above.

[0065] Example 3:

[0066] A method for measuring water depth based on underwater laser echo signals, such as Figure 5 ,include:

[0067] Acquire underwater laser echo signals of the water body to be measured, normalize them, and input them into the underwater laser echo signal processing model to obtain the type of each data point;

[0068] The peak echo intensity and corresponding time T1 at the water surface point and the peak echo intensity and corresponding time T2 at the bottom point are obtained respectively. The depth H of the water body to be measured is calculated according to H=v*(T2-T1).

[0069] Where v represents the propagation speed of the laser in the water body to be tested; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method provided in Example 1 above.

[0070] Example 4:

[0071] A computer program product includes a computer program; when the computer program is executed by a processor, it implements the underwater laser echo signal processing model training method provided in Embodiment 1 above, and / or the water depth measurement method based on underwater laser echo signals provided in Embodiment 3 above.

[0072] Example 5:

[0073] A computer-readable storage medium includes a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to perform the underwater echo signal processing model training method provided in Embodiment 1 above, and / or the water depth measurement method based on underwater echo signals provided in Embodiment 3 above.

[0074] Example 6:

[0075] An electronic device, comprising:

[0076] A computer-readable storage medium for storing computer programs;

[0077] And a processor for reading a computer program in a computer-readable storage medium and executing the underwater laser echo signal processing model training method provided in Embodiment 1 above, and / or the water depth measurement method based on underwater laser echo signals provided in Embodiment 3 above.

[0078] Example 7:

[0079] A water depth measurement system includes: a lidar and a measurement module;

[0080] LiDAR is used to acquire underwater laser echo signals of the water body being measured.

[0081] The measurement module is used to normalize the underwater laser echo signal and input it into the underwater laser echo signal processing model to obtain the type of each data point; it obtains the peak echo intensity and corresponding time T1 of the water surface point and the peak echo intensity and corresponding time T2 of the water bottom point, and calculates the depth H of the water body to be measured according to H=v*(T2-T1);

[0082] Where v represents the propagation speed of the laser in the water body to be tested; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method provided in Example 1 above.

[0083] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for training an underwater laser echo signal processing model, characterized in that, include: Multiple sets of underwater laser echo signals were acquired and preprocessed to obtain an underwater laser echo signal dataset. Each group of underwater laser echo signals includes multiple data points, and each data point includes time and echo intensity; the preprocessing includes: after normalizing each group of underwater laser echo signals, the data points located on the rising edge of the water surface waveform in each group of underwater laser echo data are labeled as water surface points, the data points located on the falling edge of the bottom waveform are labeled as bottom points, and the remaining data points are labeled as noise points; A classification model based on a deep learning model is built to predict the type of each data point in an underwater laser echo signal. The classification model is trained using the underwater laser echo signal dataset. After training, the underwater laser echo signal processing model is obtained. The classification model is a CNN-LSTM deep learning model, which includes a convolutional neural network, a long short-term memory network, a fully connected network, and a softmax function layer connected in sequence. The convolutional neural network is used to extract local features of each data point in the underwater laser echo signal; The long short-term memory network is used to capture the correlation between data points in the underwater laser echo signal and obtain the semantic features of each data point; The fully connected network is used to map the semantic features of each data point in the underwater laser echo signal to the output category features of each data point; The Softmax function layer is used to convert the output category features of each data point into the probability that the data point belongs to each category.

2. The underwater laser echo signal processing model training method as described in claim 1, characterized in that, The convolutional neural network includes five one-dimensional convolutional layers, each of which consists of a convolution operation, batch normalization, and the ReLU activation function. The Long Short-Term Memory (LSTM) network consists of an LSTM layer, a batch normalization layer, and an activation function ReLU.

3. An underwater laser echo signal processing model product, characterized in that, Includes an underwater laser echo signal processing model; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method described in claim 1 or 2.

4. A method for measuring water depth based on underwater laser echo signals, characterized in that, include: Acquire underwater laser echo signals of the water body to be measured, normalize them, and input them into the underwater laser echo signal processing model to obtain the type of each data point; The peak echo intensity at the water surface and the corresponding time T1, and the peak echo intensity at the bottom and the corresponding time T2 are obtained respectively, according to H=v (T2-T1) Calculate the depth H of the water body to be measured; Where v represents the propagation speed of the laser in the water body to be tested; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method described in claim 1 or 2.

5. A computer program product, characterized in that, Includes a computer program; when the computer program is executed by a processor, it implements the underwater laser echo signal processing model training method of claim 1 or 2, and / or the water depth measurement method based on underwater laser echo signals of claim 4.

6. A computer-readable storage medium, characterized in that, Includes a stored computer program; when the computer program is executed by a processor, it controls the device containing the computer-readable storage medium to execute the underwater echo signal processing model training method of claim 1 or 2, and / or the water depth measurement method based on underwater echo signals of claim 4.

7. An electronic device, characterized in that, include: A computer-readable storage medium for storing computer programs; And a processor for reading a computer program in the computer-readable storage medium to execute the underwater laser echo signal processing model training method of claim 1 or 2, and / or the water depth measurement method based on underwater laser echo signals of claim 4.

8. A water depth measurement system, characterized in that, include: LiDAR and measurement module; The lidar is used to acquire underwater laser echo signals of the water body to be measured; The measurement module is used to normalize the underwater laser echo signal and input it into the underwater laser echo signal processing model to obtain the type of each data point; it also obtains the peak echo intensity and corresponding time T1 at the surface point, and the peak echo intensity and corresponding time T2 at the bottom point, according to H=v (T2-T1) Calculate the depth H of the water body to be measured; Where v represents the propagation speed of the laser in the water body to be tested; the underwater laser echo signal processing model is trained by the underwater laser echo signal processing model training method described in claim 1 or 2.

Images