71 results about "Discrete wavelet transform" patented technology

This invention relates to the field of image restoration technology, specifically disclosing an image restoration method and device based on discrete wavelet transform. An image restoration network, called Multi-Wavelet Attention Memory Neural Network (MWA-MNN), is designed, consisting of three sub-networks. These sub-networks include Discrete Wavelet Attention Blocks (DWABs) and Coordinate Channel Attention Blocks (CCABs) to extract spatial and channel feature information, effectively separating degraded image layers such as rain streaks from the image to obtain a clean image. The invention leverages the expressive power of wavelet transform for high-frequency information, thereby improving image restoration performance. The sub-networks also include Supervised Selective Kernel Blocks (SSKBs) for filtering effectively extracted features in each sub-network. This invention also proposes a memristor-based cross-array MWA-MNN circuit implementation scheme. Memristors are a novel type of non-volatile memory with faster read speeds and lower power consumption. Experimental results show that this method has good performance and versatility in various image restoration tasks.

This invention discloses a baseline removal method for multi-channel magnetocardiogram (MCC) signals based on wavelet low-frequency reconstruction and morphological secondary smoothing. The invention includes: acquiring multi-channel MCC signal data; performing power frequency notch filtering on each channel to suppress the power frequency fundamental frequency and at least one harmonic component; performing configurable IIR low-pass filtering on the notched signal; setting a discrete wavelet transform extension mode and performing discrete wavelet decomposition on the filtered signal; selectively retaining decomposition coefficients based on the retention layer set and setting the remaining coefficients to zero, then performing inverse wavelet reconstruction to obtain an initial baseline estimate; performing morphological secondary smoothing on the median-smoothed baseline signal to obtain a corrected baseline signal; subtracting the corrected baseline signal from the filtered signal to obtain the baseline-corrected channel signals; and assembling and outputting a multi-channel baseline-corrected signal matrix. This invention can improve the stability and cross-channel consistency of baseline estimation for multi-channel MCC signals, reduce the risk of over-correction and under-correction, and is applicable to multi-channel batch processing scenarios.

The present application relates to the technical field of electric vehicle reducer fatigue analysis, and particularly relates to a method for preparing a frequency domain of a gear fatigue load spectrum of an electric vehicle reducer based on CCWOA, comprising: S1: constructing a bending stress load spectrum of a reducer gear of an electric vehicle; S2: generating an optimal wavelet base function based on discrete wavelet parameter optimization of an energy leakage criterion; S3: performing threshold optimization based on a CCWOA algorithm to obtain an optimal threshold; the CCWOA algorithm introduces a Logistic-Tent chaotic mapping mechanism and a cosine iteration strategy in the WOA algorithm; S4: performing discrete wavelet transform on the bending stress load spectrum of the reducer gear of the electric vehicle based on the optimal wavelet base function and performing load spectrum frequency domain coding through the optimal threshold; and S5: realizing fatigue analysis of the electric vehicle reducer based on a gear fatigue acceleration load spectrum of the electric vehicle reducer. The present application realizes high-precision compression of the reducer gear load spectrum and equivalent retention of fatigue damage, and provides a reliable scheme for fatigue analysis of the electric vehicle reducer.

This application relates to a method and apparatus for low-light image enhancement based on wavelet frequency domain residual quotient learning. The method includes: performing a two-dimensional discrete wavelet transform on a preprocessed low-light input image using a 2D-DWT module to obtain a low-frequency sub-band and three high-frequency detail sub-bands; inputting the low-frequency sub-bands into a low-frequency illumination prediction module to obtain a predicted illumination ratio map; obtaining enhanced low-frequency components based on the predicted illumination ratio map and the low-frequency sub-bands; inputting the three high-frequency detail sub-bands and the low-frequency sub-bands into a high-frequency suppression module to obtain corresponding enhanced high-frequency components; and inputting the enhanced low-frequency components and the three high-frequency components into an image reconstruction module to obtain an enhanced low-light image. This method embeds residual quotient learning into the wavelet domain, achieving interpretable enhancement in the frequency dimension. This method inherits the physical consistency of residual quotient learning while introducing the sparse representation capability of the wavelet domain, providing a new perspective for low-light enhancement tasks.

The invention provides a camera PRNU fingerprint extraction method and system based on high-frequency enhancement, and belongs to the technical field of digital multimedia evidence collection and image processing. The method comprises the following steps: constructing a fingerprint extraction network based on a Swin Transform; performing unsupervised training by using patch-level fingerprints to compare learning loss, and pulling close the fingerprints of the same equipment and pushing away the fingerprints of different equipment; designing high-frequency enhancement module loss, generating a high-frequency mask by using discrete wavelet transform, and guiding a network to focus a high-frequency region related to a fingerprint; introducing frequency domain fingerprint regularization loss to constrain uniform distribution of fingerprint power spectrum density; jointly optimizing the loss training network; and during reasoning, extracting normalized cross-correlation of the query image and the reference fingerprint of the candidate equipment for identification. According to the method, the recognition accuracy exceeding 90% can be achieved only through three common images, semantic noise interference is effectively restrained, fingerprint robustness is improved, and the method can be widely applied to scenes such as law enforcement evidence obtaining, copyright protection and safety monitoring.

This invention relates to the field of mathematical modeling technology, and more particularly to a method, apparatus, device, and computer-readable storage medium for predicting the remaining lifespan of lithium-ion batteries. The method includes: establishing a segmented empirical degradation model based on inflection point data of the lithium-ion battery; obtaining an initial prediction result of the remaining lifespan of the lithium-ion battery according to a preset particle filter algorithm and the segmented empirical degradation model, and obtaining an original error sequence based on the initial prediction result; determining a reconstruction error sequence corresponding to the capacity of the lithium-ion battery according to the original error sequence and a preset discrete wavelet transform algorithm; constructing a prediction model using a support vector regression algorithm based on the reconstruction error sequence, and correcting the initial prediction result based on the prediction model to determine the final prediction result of the remaining lifespan. This invention improves the prediction efficiency and accuracy of the remaining lifespan of lithium-ion batteries.

The application discloses a kind of progressive camouflage target detection methods based on dual-domain fusion, obtains and divides public data set;Progressive frequency domain-spatial domain collaborative optimization network is constructed, including encoder, frequency band axial attention module, amplitude guide spatial modulator and multi-scale adaptive fusion module, frequency band axial attention module fuses discrete wavelet transform with axial attention to capture directional local structure clues, amplitude guide spatial modulator introduces learnable frequency domain attention mechanism to realize the adaptive modulation of spatial features in frequency domain, multi-scale adaptive fusion module is fused by pixel-level weighting multi-scale features and introduces edge auxiliary supervision;End-to-end optimization is carried out to network using training set and total loss function, and total loss function combines regional segmentation loss and edge auxiliary supervision loss;Finally, the performance of the network is evaluated using evaluation indicators, realizes the deep interaction of spatial and frequency domain information, improves the detection accuracy and boundary integrity of camouflage target.

The application belongs to the technical field of image processing, and particularly relates to an underwater image restoration method based on an adaptive multi-scale large kernel attention module. An obtained underwater image is decomposed into a low-frequency component and three directional high-frequency components through Haar discrete wavelet transform; a mixed domain attention module is introduced at a highest resolution stage of the underwater image decomposed through the Haar discrete wavelet transform; a composite shape convolution module is added at a bottleneck position of a high-frequency branch; a PolyKernel main body is used as a large kernel encoder-decoder in a low-frequency path, and an adaptive multi-scale large kernel attention module is introduced at a bottleneck stage; after independent processing of each frequency branch, the network re-fuses high-frequency and low-frequency features into a spatial domain through inverse wavelet transform; and a light-weight unified output refining-adjusting head is used to perform structural refining and tone normalization on the fusion result.

This invention discloses an image segmentation method, apparatus, electronic device, and medium based on frequency-domain selective state-space modeling. Addressing the boundary breakage problem caused by the loss of high-frequency details in existing segmentation techniques, this invention proposes a differentiated processing strategy within the frequency domain. The method involves performing a two-dimensional discrete wavelet transform on multi-scale feature maps, decoupling them into low-frequency and high-frequency sub-bands. During this process, the low-frequency sub-band is frozen to reduce computational redundancy, while the high-frequency sub-band is fed into a state-space model for sequential scanning modeling, thereby repairing the broken boundary topology. Finally, features are reconstructed via inverse wavelet transform and decoded for output. Furthermore, this invention introduces a semantically guided fusion mechanism and a residual-based attention enhancement method. This scheme effectively balances modeling efficiency and boundary continuity, making it particularly suitable for fine segmentation of medical images.

The application discloses a power load prediction method and system based on multi-stage time sequence preprocessing and double-branch wavelet Mamba, and relates to the technical field of intelligent power grids and time sequence analysis. In view of the problems that the existing power load prediction technology has insufficient long sequence modeling capability, low prediction robustness and precision, and cannot simultaneously consider global trend fitting and local mutation capturing when facing high-noise non-stationary data, a high-quality target sequence is first constructed through multi-stage time sequence preprocessing, and after channel independence and block embedding processing, the high and low frequency characteristic components are decoupled through discrete wavelet transform, the global long-range dependence is extracted through a bidirectional trend Mamba module, the local mutation characteristics are purified through a detail Mamba module, and noise is suppressed, and finally, the prediction result is output through inverse wavelet reconstruction. The application significantly improves the precision and robustness of long sequence power load prediction and reduces the computational complexity.

The application relates to the field of industrial optimization, and particularly discloses a production efficiency global optimization method applied to an industrial production line, which comprises the following steps: collecting vibration frequency of a processing device and differential current data of a servo system of the processing device in real time during operation of the processing device, extracting vibration and current abnormal characteristic values through discrete wavelet transform and fast Fourier transform respectively, inputting a multi-dimensional running state characteristic vector into a trained deep learning efficiency evaluation model for fusion analysis, outputting a current efficiency coefficient of the processing device and dividing the efficiency level, triggering a corresponding power improvement strategy according to the efficiency level, and realizing closed-loop control from state perception to active intervention. The application improves the efficiency identification precision and self-adaptive response capability of the processing device under complex working conditions, and has good practicability and popularization value.

This invention discloses a method and system for monitoring leaks in urban low-pressure gas pipelines based on negative pressure waves, comprising: S1, acquiring the medium-pressure signal of the medium-pressure pipeline upstream of the pressure regulating box and the low-pressure signal of the low-pressure pipeline downstream of the pressure regulating box; S2, performing discrete wavelet transform on the medium-pressure signal and the low-pressure signal; S3, if the low-pressure signal shows a sudden pressure drop, it is determined that a low-pressure gas leak has occurred under no-gas-use conditions; otherwise, proceed to step S4; S4, if the maximum fluctuation amplitude of the detail signal after two splits of the medium-pressure signal exceeds a set value or the low-pressure signal shows low-frequency changes, it is determined that a low-pressure gas leak has occurred under gas-use conditions. This invention, based on negative pressure waves, only requires collecting and analyzing the internal pressure of the pipeline to achieve real-time online monitoring. Compared with traditional manual inspections, it saves manpower, is highly accurate, safe, and convenient, and has a short response time and high processing efficiency in leak situations.

This invention discloses a reversible image steganography method based on multi-scale feature enhancement and dynamic attention. First, discrete wavelet transform is performed on the carrier image and the secret image to obtain multi-channel wavelet domain features. Then, these features are input into multiple parameter-shared reversible hidden blocks. During the coupled transform process, a multi-scale spatial information enhancement module and a dynamic attention feature refiner module are introduced to extract local and global contextual information and dynamically adjust the importance of channel and spatial features, achieving high-quality embedding of the secret information. Finally, the cryptic image is obtained through inverse wavelet transform. In the inverse recovery stage, the cryptic image and auxiliary variables are input into a symmetrical reversible recovery path to separate and reconstruct the recovered secret image and the recovered carrier image. This invention achieves near-lossless recovery of the secret information while maintaining high visual quality, significantly improving the concealment and recovery accuracy of steganography.

The application provides a discrete wavelet transform-based echocardiogram segmentation method and system, which comprises the following steps: obtaining an echocardiogram to be segmented; performing feature extraction processing and two-dimensional discrete wavelet transform on the echocardiogram to be segmented by using a pre-trained echocardiogram segmentation model to determine spatial domain features and frequency domain features; performing space-frequency cross enhancement fusion processing on the spatial domain features and the frequency domain features to determine space-frequency fusion features; inputting the echocardiogram to be segmented and random Gaussian noise into a preset denoising network to perform high-frequency perception down-sampling enhancement processing to determine frequency perception features; and inputting the space-frequency fusion features into the preset denoising network to perform preset times of diffusion iteration processing on the space-frequency fusion features and the frequency perception features to determine a heart multi-part segmentation map of the echocardiogram to be segmented. Through the application, the discrete wavelet transform and the conditional diffusion architecture are used to realize high-precision positioning and segmentation of a boundary-fuzzy heart multi-part.

This invention discloses a multi-image encryption method based on 3D face keys and multiplexed digital holography. The encryption steps include: using 3D face high-order data acquisition technology, chaotic face structured light phase mask generation technology, grating modulation technology, multiplexed digital holographic coding technology, and watermark embedding technology based on discrete wavelet domain singular value decomposition, multiple plaintext information images can be encrypted into an amplitude-type holographic ciphertext and embedded into a host image; the decryption steps include: the user must first calculate the face similarity using a 3D face multilayer perceptron neural network. If the similarity exceeds a preset threshold, authentication is successful. Then, the final decryption result is generated using watermark image extraction technology based on discrete wavelet transform, image decryption technology based on the Cramer-Kroni relation, and spectral filtering technology; otherwise, authentication fails. This method has advantages such as a large key space, high sensitivity, strong robustness of the 3D face key, and low decryption crosstalk.

The application discloses a method for angiogram image segmentation based on self-supervised contrast learning and frequency domain feature enhancement. The method takes UNet as the basis, inputs the original angiogram image into the symmetrical encoder-decoder structure, integrates the discrete wavelet transform path in the encoding stage, injects the high-frequency texture component into the down-sampling feature of the encoder, realizes the detail compensation of the slender branch of the blood vessel, and connects the projection layer at the end of the bottleneck layer of the encoder to perform feature mapping. In the first stage, the decoder is frozen, self-supervised pre-training is performed by minimizing the contrast loss, and the feature capturing ability of the encoder is strengthened. In the second stage, the pre-training weight is loaded and the decoder is unfrozen, the Dice similarity coefficient and the binary cross entropy loss are combined for supervised fine-tuning, and accurate blood vessel segmentation is realized. The method can deeply fuse the global semantic topology and the local detail information, significantly improve the segmentation performance and edge accuracy, and maintain high calculation speed and low resource consumption.

The application relates to the technical field of industrial equipment state monitoring and fault diagnosis, and discloses a centrifugal fan fault prediction method and system. The method comprises the following steps: collecting bearing mixed vibration data and reconstructing an initial vibration sequence, removing strong base frequency interference by using empirical mode decomposition and energy frequency screening mechanism to obtain a residual vibration signal; performing discrete wavelet transform and adaptive soft threshold denoising on the residual vibration signal, and cooperating with Hilbert envelope slice quantization to obtain a time-varying impact sequence; calculating a dynamic change rate of the time-varying impact sequence, combining a base frequency component to classify and fit an independent fault characteristic curve; constructing a fault trend distribution surface based on the fault characteristic curve, separating eccentric evolution and damage expansion trajectories by energy flow decoupling, and comparing a threshold to determine a prediction result. The application can effectively overcome the covering of weak impact characteristics by strong base frequency vibration, and realize high-precision independent prediction of centrifugal fan impeller imbalance and bearing pitting under complex coupling faults.

This invention discloses an image fusion method and system based on cross-modal attention explicit task guidance. The method includes: normalizing and extracting features from infrared and visible light images to obtain infrared and visible light image features; performing feature decomposition on the infrared and visible light image features using discrete wavelet transform to obtain low-frequency and high-frequency components; generating an explicit task guidance signal and injecting it into the high and low-frequency components through a cross-modal attention mechanism for target perception feature enhancement, resulting in enhanced high and low-frequency components; and classifying and adaptively fusing the enhanced high and low-frequency components to obtain a target-perception fused image. This invention effectively preserves infrared thermal energy distribution and visible light details, improving the visual fusion effect of the images. As an image fusion method and system based on cross-modal attention explicit task guidance, this invention can be widely applied in the field of cross-modal image fusion technology.

This invention discloses an image compression method, system, and apparatus for LED display control systems, belonging to the field of image compression. The method includes the following steps: hardware decoding of the HDMI or DVI video interface input signal to parse the original RGB image data stream; caching a complete image frame data in DRAM memory; performing overall compression encoding on the complete image frame data; packaging the image data after overall compression encoding into Ethernet frame data according to two-dimensional information; parsing the compressed image data from the Ethernet frame data; performing partitioned decompression decoding on the parsed image data; and caching the partitioned decompression decoding image frame data in DRAM memory. This invention employs an intra-frame compression transmission technology based on discrete wavelet transform, using overall compression at the encoding end and partitioned decompression at the decoding end, enabling decompression of compressed images at a very low cost.

The application discloses a compression method and device of seismic data, electronic equipment, storage medium and product. The method comprises the following steps: extracting pre-stack seismic data into common offset data; performing discrete wavelet transform on the common offset data to obtain transformed data; performing multi-level coding on the transformed data based on a multi-level tree set splitting coding algorithm to obtain compressed data, wherein the coding level number of the multi-level coding includes an integer from a high coding level number to a maximum coding level number, the maximum coding level number is the total number of thresholds that can be used to determine the importance of wavelet coefficients, and the high coding level number is an integer greater than 1 and less than the maximum coding level number. The above scheme realizes seismic data compression by performing multi-level coding on the wavelet domain data of the seismic data through the multi-level tree set splitting coding algorithm, can guarantee the compression quality of the seismic data, and starts positioning and coding of the data from the high coding level number, so that the total coding level number is reduced, and the compression efficiency of the seismic data is improved.

The application discloses a kind of new energy power grid transient voltage instability critical state determination method and system, comprising: obtaining the voltage time series data of bus after new energy power grid fault;The voltage time series data is processed with discrete wavelet transform, to obtain multiple different frequency band wavelet coefficients;Based on the wavelet coefficient, high-frequency detail energy and high-low frequency energy ratio are calculated;The high-frequency detail energy and high-low frequency energy ratio are fused to determine transient voltage stability index;When the transient voltage stability index is greater than or equal to preset index threshold, it is determined that power grid is in transient voltage instability critical state.The present application only relies on external measurement, and does not require internal parameters that are difficult to obtain, is physically clear, and the result is reliable, providing a new and effective solution for online safety and stability analysis of new energy power grid.

The application discloses a network traffic anomaly detection method and device, a terminal and a medium. The method comprises the following steps: collecting network traffic data in real time, performing discrete wavelet transform on the network traffic data, calculating the variance and expectation of wavelet coefficients at each scale, calculating the Hurst coefficient of the network traffic, dynamically adjusting the Hurst coefficient reference value based on the fluctuation variable control value of the network traffic, judging whether the network traffic is abnormal according to the difference between the Hurst coefficient and the Hurst coefficient reference value, and if the network traffic is abnormal, searching for the abnormal time of the network traffic and performing network attack tracing. Therefore, the embodiment of the application can measure the wavelet coefficient variance at each scale and the wavelet coefficient expectation at each scale, calculate the fluctuation variable control value of the network traffic, and dynamically update the Hurst coefficient reference value, so that the false positive rate of network traffic anomaly is reduced.

The application discloses a wide field of view video image change detection method and device, and the method comprises the following steps: using an improved adaptive fast guided filter for video image change detection; after filtering the multi-temporal video image, a logarithmic ratio operator considering neighborhood information is proposed, the mean ratio difference graph generation mode is improved, and a difference graph is generated; the improved MR image and the improved LR image are subjected to image fusion through discrete wavelet transform, and a fused difference graph is obtained; a soft threshold function is used to perform initial classification on the fused difference graph, and an initial change area and an unchanged area are obtained; a cumulative distribution function is used to compress the pixel value of the classification result from [0, 255] to [0, 1]; a super-fast robust constraint fuzzy C-Means clustering algorithm is proposed, and an improved adaptive median filter is used for denoising processing. The device comprises a processor and a memory.

The application provides a camera PRNU fingerprint extraction method and system based on high-frequency enhancement, and belongs to the technical field of digital multimedia forensics and image processing. The method comprises the following steps: constructing a fingerprint extraction network based on a Swin Transformer; performing unsupervised training by using a patch-level fingerprint contrast learning loss, narrowing the fingerprints of the same device and widening the fingerprints of different devices; designing a high-frequency enhancement module loss, generating a high-frequency mask by using a discrete wavelet transform, and guiding the network to focus on the high-frequency area related to the fingerprint; introducing a frequency domain fingerprint regularization loss to constrain the uniform distribution of the power spectral density of the fingerprint; jointly optimizing the network by using the above losses; and identifying by extracting the normalized cross-correlation between the query image and the candidate device reference fingerprint during inference. The application can achieve an identification accuracy of more than 90% by using only 3 ordinary images, effectively suppresses semantic noise interference, improves the robustness of the fingerprint, and can be widely applied to law enforcement, copyright protection, security monitoring and other scenes.

The application discloses a surgical stage recognition method based on time-frequency feature fusion, and is realized based on a constructed TFRCDLFormer network, a multi-level frequency domain feature module performs multi-layer cavity convolution RCDL processing on an input picture, then discrete wavelet transformation is performed to obtain multi-level frequency information containing high frequency and low frequency, and after processing the frequency information, inverse discrete wavelet transformation is performed; a multi-scale time domain feature module performs RCDL processing on the input picture, then average pooling and up-sampling processing are performed to obtain feature information of different scales, and the feature information of each scale is comprehensively processed by adopting an RCDL and a Transformer method; a multi-level frequency domain feature enhancement module performs feature fusion on the multi-level frequency domain feature and the multi-scale time domain feature, then frequency domain feature enhancement is performed on the fused features, and a surgical stage recognition prediction result is output after a full connection layer. The recognition speed and accuracy are effectively improved.

The application discloses a building group load prediction method based on causal inference. First, the building group load sequence is acquired and pretreated; the discrete wavelet transform is used to decompose the building load sequence to obtain a plurality of high-frequency components and a low-frequency component; the building load fluctuation characteristics are reconstructed by using the high-frequency components and the low-frequency component; then, the static and dynamic causal diagrams are constructed by taking the building as a node, the building load fluctuation characteristics are coded into latent representations, the building load state structure equation is constructed by using the latent representations, each building load state structure equation is solved by optimization, and the adjacency matrix of the causal diagram is obtained; finally, the load prediction model is constructed, the adjacency matrix of the static causal diagram and the dynamic causal diagram is taken as input, the building group load at a future time step is predicted according to the building group historical load sequence, and the building group load prediction sequence is obtained. The problems that it is difficult to capture the dynamic propagation effect of the load change among the buildings and the modeling of the complex interaction among the buildings is insufficient are solved.

The application discloses a load identification method based on an improved fuzzy algorithm, and comprises the following steps: step one, collecting three-phase voltage, current and power parameters of a low-voltage side of a transformer in a transformer area through a terminal device in real time, and constructing an original time sequence data set; the time sequence data set is windowed according to a preset time step, and a data segment sequence with a time scale is generated; step two, performing multi-level discrete wavelet transform (DWT) on the active power time sequence data segment obtained in step one, extracting high-frequency detail coefficient extreme values and low-frequency approximation coefficient extreme values as load mutation features; and a composite feature vector sequence is constructed; step three, performing iterative analysis on the feature vector sequence by using a dynamic weight fuzzy C-means clustering algorithm (DFCM), and establishing an initial load feature library; through three-level criteria of active power interval pre-screening, reactive power overlap rate checking and voltage and current feature consistency matching, fast classification and power interval identification of a new load sample are realized, and decision support is provided for intelligent dispatching of a distribution network.

The application belongs to the technical field of computing processing, and particularly relates to a disaster environment distributed collaborative sensing method based on multi-agent reinforcement learning, which comprises the following steps: obtaining an initial sequence of disaster environment sensed by an agent, and performing discrete wavelet transform denoising and normalization processing to obtain standardized state observation values; extracting spatial entropy of the agent and calculating the sensing difference degree between the agent and neighboring agents, and obtaining collaborative sensing interaction gain through a collaborative sensing interaction gain model; obtaining an environment field change rate, and using the environment field change rate to dynamically compensate an original action evaluation value output by a reinforcement learning network to obtain a corrected action evaluation value; obtaining a sensing task saturation degree of the agent, and completing physical control vector mapping based on the corrected action evaluation value and the sensing task saturation degree, and adjusting the pulse width modulation duty cycle of a driving motor and the sampling frequency of a laser scanner through the physical control vector to realize dynamic adjustment of sensing operation intensity.

This invention discloses a method and system for event detection using an optical time-domain reflectometry (OTDR). The method includes the following steps: S1: Measuring the optical fiber using an OTD to obtain an OTD test curve; S2: Denoising the test curve data to obtain denoised test curve data; S3: Performing discrete wavelet transform and a constant false alarm rate (CFAR) algorithm based on the transform exponent to obtain wavelet coefficient reflection event points and wavelet coefficient non-reflection event points; S4: Merging the wavelet coefficient reflection event points and wavelet coefficient non-reflection event points and sorting them according to their positions; S5: Mapping the sorted wavelet coefficient reflection event points and wavelet coefficient non-reflection event points onto the test curve to obtain the true positions of the reflection event points and non-reflection event points, thus completing the OTD event detection. This invention can quickly and accurately locate targets, including nearby multiple targets and weak non-reflection events.

This invention proposes an underwater mollusk larvae image enhancement method integrating wavelet transform and Mamba architecture, comprising: establishing a U-Net network based on wavelet transform; in the encoding stage, constructing a filter bank based on discrete wavelet transform (DWT) to perform frequency band separation of multi-scale features; constructing a three-layer wavelet decomposition cascaded downsampling architecture, and realizing information enhancement of multi-frequency band features through an interactive dual-branch framework, wherein the interactive dual-branch framework includes a low-frequency Mamba processing module and a high-frequency window enhancement module; in the decoding stage, performing spatial domain reconstruction of multi-frequency band features using inverse wavelet transform (IDWT); and realizing visible light reconstruction of the target image based on the above-mentioned U-Net network enhancement. The model constructed by the method of this invention is used to effectively suppress feature loss caused by traditional downsampling while inheriting the advantages of U-shaped network topology through frequency band decoupling, dual-path interaction, and full information reconstruction mechanisms, thereby achieving high-fidelity restoration of image content.