Signal processing method, device, equipment and storage medium of photoelectric detector

Abstract

The present application provides a signal processing method, device, equipment and storage medium for a photoelectric detector, the method comprising: extracting an amplitude histogram and preliminary preliminary motion features from an original signal frame, weighting and combining the amplitude histogram according to a partition weighted matrix to obtain a partition weighted amplitude sequence, performing time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, inputting the stable amplitude sequence into a preset Gaussian mixture model for kernel density estimation to obtain an amplitude probability feature; according to a first change feature of each sampling point in a digital signal frame, obtaining a third change feature according to the first change feature, the amplitude probability feature and a cluster analysis algorithm; determining a preselected sampling time zone according to the third change feature and the partition weighted amplitude sequence; constructing a probability prediction model according to a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, inputting the preselected sampling time zone into the probability prediction model, and outputting signal detection information.

Description

Technical Field

[0001] The present application relates to the technical field of optical signal processing, and in particular to a signal processing method, device, equipment and storage medium for a photoelectric detector. Background Art

[0002] At present, in the field of modern security inspection and industrial inspection, the demand for dangerous goods detection is increasing, especially in key places such as airports, ports, chemical plants, etc. Although traditional detection methods, such as X-ray imaging, metal detection and chemical analysis, meet basic security needs to a certain extent, they have obvious limitations in detecting trace or hidden dangerous goods. Especially for some low-concentration and highly concealed dangerous substances, traditional methods often fail to achieve ideal detection accuracy.

[0003] In recent years, photoelectric detection technology has shown great potential in dangerous goods detection due to its high sensitivity and non-contact detection characteristics. Photoelectric detectors can achieve remote and rapid detection of dangerous goods by capturing weak light signals emitted by substances. However, existing photoelectric detection technology still faces many challenges when processing weak signals. On the one hand, the interference of environmental noise and background light seriously reduces the signal-to-noise ratio of the detection signal, resulting in low detection accuracy; on the other hand, the response characteristics and noise level of the photoelectric detector itself also limit its performance in weak signal detection. Summary of the invention

[0004] The present application provides a signal processing method, device, equipment and storage medium for a photoelectric detector, which are used to improve the accuracy of the photoelectric detector and the detection capability of weak signals.

[0005] In a first aspect, an embodiment of the present application provides a signal processing method of a photodetector, the method comprising:

[0006] Calculating amplitude histograms of a plurality of digital signal frames respectively, and calculating inter-frame differences according to every two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector;

[0007] Divide each digital signal frame into a plurality of sampling sub-regions, generate a partition weighting matrix, and perform weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix to obtain a partition weighted amplitude sequence;

[0008] Performing time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, inputting the stable amplitude sequence into a preset Gaussian mixture model for kernel density estimation to obtain an amplitude probability feature;

[0009] According to each sampling point and its adjacent sampling points in the digital signal frame, determine a first change feature of each sampling point, obtain a second change feature according to the first change feature and the amplitude probability feature, and perform cluster analysis on the second change feature to obtain a third change feature;

[0010] Determining a preselected sampling time zone according to the third variation characteristic and the partition weighted amplitude sequence;

[0011] A probability prediction model is constructed according to a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, the preselected sampling time zone is input into the probability prediction model, the target sampling time zone is output, and signal detection information is output according to the target sampling time zone.

[0012] In a second aspect, an embodiment of the present application provides a signal processing device for a photodetector, the device comprising:

[0013] A first calculation module is used to calculate the amplitude histograms of a plurality of digital signal frames respectively, and to calculate the inter-frame difference according to every two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector;

[0014] A second calculation module is used to divide each digital signal frame into a plurality of sampling sub-regions, and generate a partition weighting matrix, and perform weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix to obtain a partition weighted amplitude sequence;

[0015] An amplitude analysis module is used to perform time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, and input the stable amplitude sequence into a preset Gaussian mixture model for kernel density estimation to obtain an amplitude probability feature;

[0016] a change analysis module, configured to determine, according to each sampling point and its adjacent sampling points in the digital signal frame, a first change feature of each sampling point, obtain a second change feature according to the first change feature and an amplitude probability feature, and perform cluster analysis on the second change feature to obtain a third change feature;

[0017] A time zone determination module, used to determine a preselected sampling time zone according to the third change feature and the partition weighted amplitude sequence;

[0018] The result output module is used to build a probability prediction model based on a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, input the pre-selected sampling time zone into the probability prediction model, output the target sampling time zone, and output signal detection information according to the target sampling time zone.

[0019] In a third aspect, an embodiment of the present application provides a photoelectric detection module, wherein the photoelectric detection module includes a memory and a processor;

[0020] The memory is used to store computer programs;

[0021] The processor is used to execute the computer program and implement the signal processing method of the photoelectric detector as described in any one of the embodiments of the present application when executing the computer program.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements a signal processing method of a photodetector as described in any one of the embodiments of the present application.

[0023] The embodiment of the present application provides a signal processing method for a photoelectric detector, the method comprising: respectively calculating the amplitude histograms of a plurality of digital signal frames, and calculating the inter-frame difference according to each two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector; dividing each digital signal frame into a plurality of sampling sub-regions, and generating a partition weighting matrix, performing weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix, and obtaining a partition weighted amplitude sequence; performing time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, and inputting the stable amplitude sequence into a preset The Gaussian mixture model is used for kernel density estimation to obtain the amplitude probability feature; according to each sampling point in the digital signal frame and its adjacent sampling points, the first change feature of each sampling point is determined, and the second change feature is obtained according to the first change feature and the amplitude probability feature, and the second change feature is clustered to obtain the third change feature; according to the third change feature and the partition weighted amplitude sequence, the preselected sampling time zone is determined; according to the preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, a probability prediction model is constructed, the preselected sampling time zone is input into the probability prediction model, the target sampling time zone is output, and the signal detection information is output according to the target sampling time zone. The photoelectric detector signal processing method provided in the embodiment of the present application effectively improves the accuracy and reliability of weak signal detection through multi-level signal processing and feature extraction, enhances the detection capability of motion features through inter-frame difference and partition weighted matrix, improves the accuracy of amplitude probability features by using Gaussian mixture model and kernel density estimation, realizes accurate identification of target sampling time zone through cluster analysis and probability prediction model, and significantly improves the overall performance and anti-interference capability of the detection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0025] Figure 1 A schematic flow chart of a signal processing method of a photodetector provided in an embodiment of the present application;

[0026] Figure 2 A schematic block diagram of a signal processing device for a photodetector provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0029] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0030] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0031] See also Figure 1 , Figure 1 The embodiment of the present application also provides a schematic flow chart of a signal processing method of a photodetector. Figure 1 As shown, the specific steps of the method include: S101-S106.

[0032] S101, respectively calculating amplitude histograms of a plurality of digital signal frames, and calculating inter-frame differences according to every two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting original signal frames output by a photoelectric detector.

[0033] The photodetector can be a common type such as a photodiode or a photomultiplier tube. The photodetector converts the received light signal into an electrical signal output. Then, the output analog electrical signal is converted into a digital signal sequence, i.e., the original signal frame. The original signal frame is preprocessed, including denoising, filtering and other operations, to improve the accuracy of subsequent processing. The preprocessed original signal frame is divided into several time periods (frames), each frame containing a fixed number of sampling points. For each frame signal, its amplitude distribution histogram is calculated to obtain the amplitude distribution characteristics of the signal, i.e., the amplitude histogram. The inter-frame difference between adjacent frames is calculated to obtain preliminary change information, i.e., preliminary motion characteristics. The preprocessed signal frame, amplitude distribution histogram and inter-frame difference information are used as input for subsequent processing.

[0034] S102, dividing each digital signal frame into a plurality of sampling sub-regions, and generating a partition weighting matrix, performing weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix, to obtain a partition weighted amplitude sequence.

[0035] Exemplarily, each digital signal is divided into multiple sampling sub-areas through a preset sampling range, and each sampling sub-area contains several sampling points. For each sampling sub-area, different weight coefficients are assigned according to its time position in the whole frame to generate a partition weighting matrix. The weight of the central area is larger, and the weight of the edge area is smaller. The weighted amplitude distribution histogram is calculated for each sampling sub-area according to the partition weighting matrix. The weighted amplitude distribution histograms of all sampling sub-areas are combined according to the partition weighting matrix to obtain an amplitude feature sequence weighted by the time position of the whole frame. The above process is repeated for multiple consecutive digital signal frames to obtain a series of amplitude feature sequences that are continuous in time. Then the time variation law of these amplitude feature sequences is analyzed to extract a stable partition weighted amplitude sequence.

[0036] S103, performing time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, inputting the stable amplitude sequence into a preset Gaussian mixture model for kernel density estimation, and obtaining an amplitude probability feature.

[0037] Exemplarily, when performing time series analysis on the partitioned weighted amplitude sequence, methods such as moving average method, exponential smoothing method or ARIMA model can be used to eliminate random fluctuations and extract the trend and periodic characteristics of the sequence. This step helps to obtain a more stable and reliable amplitude sequence. After obtaining a stable amplitude sequence, it is input into a preset Gaussian mixture model for kernel density estimation. The Gaussian mixture model is a commonly used probability density estimation method, which assumes that the data consists of multiple Gaussian distributions. In this process, the number of Gaussian components needs to be determined, and the Bayesian Information Criterion (BIC) or Akaike Information Criterion (AIC) can be used to select the optimal model complexity. The results of kernel density estimation can be used to describe the probability distribution characteristics of the amplitude. In order to more comprehensively characterize the amplitude probability characteristics, some statistics such as mean, variance, skewness and kurtosis can be further calculated. These statistics can reflect the central tendency, dispersion, symmetry and sharpness of the amplitude distribution.

[0038] In addition, you can consider using nonparametric methods, such as kernel density estimation (KDE), to complement the results of Gaussian mixture models. KDE methods do not assume that the data follow a specific distribution form and may therefore provide more flexible and accurate density estimates in some cases.

[0039] S104. Determine a first variation feature of each sampling point according to each sampling point and its adjacent sampling points in the digital signal frame, obtain a second variation feature according to the first variation feature and the amplitude probability feature, and perform cluster analysis on the second variation feature to obtain a third variation feature.

[0040] Exemplarily, for each sampling point in the digital signal frame, the difference between it and the adjacent sampling points before and after is calculated to form a 3D change vector. Based on the amplitude feature sequence obtained in step S102, the probability that each sampling point belongs to the target is calculated to generate an amplitude probability sequence. The probability value can be calculated using methods such as Gaussian mixture model or kernel density estimation. The neighborhood change model is combined with the amplitude probability sequence to weight the change vector. Cluster analysis is performed on the weighted change vector to extract the main change pattern as the change feature of the target.

[0041] S105. Determine a preselected sampling time zone according to the third change feature and the partition weighted amplitude sequence.

[0042] Exemplarily, when determining the preselected sampling time zone through the third change feature and the partitioned weighted amplitude sequence, the time period with a larger change amplitude, higher frequency or longer duration can be selected as the key focus features, and the time zone to which these features belong can be used as the preselected sampling time zone, so that more dynamic information can be extracted.

[0043] S106. Construct a probability prediction model according to a preset Gaussian distribution function, a partition weighted amplitude sequence and a third change feature, input a preselected sampling time zone into the probability prediction model, output a target sampling time zone, and output signal detection information according to the target sampling time zone.

[0044] Exemplarily, the third change feature and the partition weighted amplitude sequence are placed within the same scale range to facilitate subsequent fusion processing. Based on a preset weight coefficient, it is used to balance the importance of the third change feature and the partition weighted amplitude sequence. This weight coefficient can be fixed or dynamically adjusted according to actual conditions. Then, the normalized third change feature and the partition weighted amplitude sequence are weighted and summed according to the preset weights to obtain a comprehensive feature vector. This feature vector contains both the amplitude information and the change information of the target, and can more comprehensively describe the characteristics of the target.

[0045] Then, based on this comprehensive feature vector, a probability model is constructed to describe the relationship between the observed value and the true target state. This probability model can use Gaussian distribution or a more complex non-parametric model. Finally, using this joint observation model, the likelihood probability of each candidate target time period can be calculated. The target sampling time zone is determined from the pre-selected sampling time zone based on the likelihood probability. According to the target digital frame corresponding to the target sampling time zone, dangerous goods detection is performed, and if the target digital frame is close to the relevant information of the dangerous goods, a warning message is output.

[0046] The embodiment of the present application provides a signal processing method for a photoelectric detector, the method comprising: respectively calculating the amplitude histograms of a plurality of digital signal frames, and calculating the inter-frame difference according to each two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector; dividing each digital signal frame into a plurality of sampling sub-regions, and generating a partition weighting matrix, performing weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix, and obtaining a partition weighted amplitude sequence; performing time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, and inputting the stable amplitude sequence into a preset The Gaussian mixture model is used for kernel density estimation to obtain the amplitude probability feature; according to each sampling point in the digital signal frame and its adjacent sampling points, the first change feature of each sampling point is determined, and the second change feature is obtained according to the first change feature and the amplitude probability feature, and the second change feature is clustered to obtain the third change feature; according to the third change feature and the partition weighted amplitude sequence, the preselected sampling time zone is determined; according to the preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, a probability prediction model is constructed, the preselected sampling time zone is input into the probability prediction model, the target sampling time zone is output, and the signal detection information is output according to the target sampling time zone. The photoelectric detector signal processing method provided in the embodiment of the present application effectively improves the accuracy and reliability of weak signal detection through multi-level signal processing and feature extraction, enhances the detection capability of motion features through inter-frame difference and partition weighted matrix, improves the accuracy of amplitude probability features by using Gaussian mixture model and kernel density estimation, realizes accurate identification of target sampling time zone through cluster analysis and probability prediction model, and significantly improves the overall performance and anti-interference capability of the detection system.

[0047] In order to more clearly introduce the technical solution of the present application, the technical solution of the present application will be introduced through specific embodiments below. It should be noted that the specific embodiments are used to expand the technical solution of the present application, but are not intended to limit the present application.

[0048] In some embodiments, amplitude histograms of multiple digital signal frames are calculated respectively, and inter-frame differences are calculated based on every two adjacent digital signal frames to obtain preliminary motion features, including: dividing the original signal frame into multiple digital signal frames according to a preset number of sampling points; calculating the amplitude distribution histogram of each digital signal frame to obtain amplitude distribution features; calculating inter-frame differences between adjacent digital signal frames to obtain differential features; and generating preliminary motion features from the amplitude distribution features and differential features.

[0049] Exemplarily, the segmentation process is based on a preset number of sampling points to ensure that each digital signal frame contains the same number of sampling points. For each digital signal frame, its amplitude distribution histogram is calculated, so that the distribution of the original signal frame in different amplitude ranges can be obtained, and the amplitude distribution characteristics can be obtained. At the same time, the inter-frame difference between adjacent digital signal frames is calculated to obtain the differential features, which helps to capture the dynamic changes of the signal. The amplitude distribution features and the differential features are combined to generate preliminary motion features. This preliminary motion feature integrates the static distribution and dynamic change information of the signal, providing a basis for subsequent signal processing and analysis. Through multi-dimensional feature extraction, the signal characteristics of the photodetector output can be more comprehensively described, which is conducive to subsequent signal analysis and target detection.

[0050] In some embodiments, each digital signal frame is divided into multiple sampling sub-areas, and a partition weighting matrix is ​​generated, including: generating a partition weighting coefficient based on the positional relationship between each sampling sub-area and the time zone midpoint of the digital signal frame, wherein the closer the sampling sub-area is to the time zone midpoint, the larger the corresponding partition weighting coefficient; and generating a partition weighting matrix based on the partition weighting coefficient.

[0051] Exemplarily, the distance of each sampling sub-region relative to the midpoint of the time zone is calculated, and different weights are assigned based on this distance. The sampling sub-regions closer to the midpoint of the time zone are assigned greater weights, reflecting that these regions may contain more key information. This weighting strategy helps to highlight the most representative parts of the signal while reducing the effects of noise or interference that may be introduced by the edge areas. The generated partition weighting coefficients are then organized into a matrix form, namely the partition weighting matrix. This matrix can be directly applied to subsequent signal processing steps, such as weighted combination of amplitude histograms, thereby emphasizing more important information in time while retaining the overall characteristics of the signal. This method not only improves the accuracy of signal analysis, but also may improve the ability to detect transient events or local features.

[0052] In some embodiments, a time series analysis is performed on the partitioned weighted amplitude sequence to obtain a stable amplitude sequence, and the stable amplitude sequence is input into a preset Gaussian mixture model for kernel density estimation to obtain amplitude probability characteristics, including: applying Fourier transform to the partitioned weighted amplitude sequence to obtain frequency domain data; calculating power spectrum density based on the frequency domain data to obtain frequency component distribution; performing threshold processing on the frequency component distribution to obtain main frequency components; filtering the partitioned weighted amplitude sequence based on the main frequency components to obtain a denoised amplitude sequence; performing wavelet decomposition on the denoised amplitude sequence to obtain multi-scale coefficients; and The denoised amplitude sequence is reconstructed by multi-scale coefficients to obtain a reconstructed amplitude sequence; a trend analysis is performed on the reconstructed amplitude sequence to obtain a local trend; the residual is calculated based on the local trend to obtain a detrended amplitude sequence; a time series analysis is performed on the detrended amplitude sequence to obtain a stable amplitude sequence; a preliminary statistical analysis is performed on the stable amplitude sequence to obtain data distribution characteristics; a Gaussian mixture model is initialized based on the data distribution characteristics to obtain an initial model; the initial model is optimized to obtain an optimized Gaussian mixture model; the kernel density estimate is calculated using the optimized Gaussian mixture model to obtain amplitude probability characteristics.

[0053] Exemplarily, applying Fourier transform to the partitioned weighted amplitude sequence can convert the signal to the frequency domain, making it easier to identify the main frequency components. Through power spectral density calculation and threshold processing, the noise can be effectively filtered out and the key frequency information in the signal can be retained. The subsequent filtering and wavelet decomposition steps further optimize the signal quality, remove high-frequency noise and retain multi-scale features. Trend analysis and detrending processing help to eliminate long-term trends in the signal and make short-term fluctuations more obvious. Time series analysis can capture the dynamic characteristics of the signal and generate a stable amplitude sequence. Finally, kernel density estimation using a Gaussian mixture model can accurately describe the probability distribution of the amplitude. This series of processing steps not only improves the signal-to-noise ratio of the signal, but also accurately captures the statistical characteristics of the signal, providing a reliable basis for subsequent signal detection and analysis. In this way, the key features in the photodetector signal can be effectively extracted and the accuracy and reliability of signal processing can be improved.

[0054] In some embodiments, according to each sampling point in the digital signal frame and its adjacent sampling points, a first change feature of each sampling point is determined, according to the first change feature and the amplitude probability feature, a second change feature is obtained, and a cluster analysis is performed on the second change feature to obtain a third change feature, including: calculating the amplitude difference value between each sampling point and the adjacent sampling point to obtain the first change feature; performing the first change feature according to the amplitude probability feature to obtain the second change feature; performing a cluster analysis on the second change feature according to the preset amplitude change partition to obtain the third change feature.

[0055] By calculating the amplitude difference between each sampling point and the adjacent sampling point, the first change feature is obtained, which can effectively capture the local change trend of the signal. The first change feature is processed using the amplitude probability feature to obtain the second change feature, which can be combined with global statistical information to correct the local changes and improve the accuracy and robustness of the feature. The second change feature is clustered and analyzed to obtain the third change feature, which can classify similar change patterns and simplify subsequent processing steps. This multi-level feature extraction and optimization method can effectively extract key change information in the signal, reduce the impact of noise and interference, and provide a reliable basis for subsequent sampling time zone selection and signal detection. Through this method, the present invention can more accurately identify effective information in the photodetector signal and improve the accuracy and efficiency of signal processing.

[0056] In some embodiments, a preselected sampling time zone is determined based on a third change feature and a partitioned weighted amplitude sequence, including: performing a time continuity analysis on the third change feature to obtain a change duration sequence; performing a threshold screening on the change duration sequence based on the partitioned weighted amplitude sequence to obtain a first candidate time period; merging and splitting the first candidate time period to obtain a second candidate time period; and accurately locating the boundaries of the second candidate time period to obtain a preselected sampling time zone.

[0057] Exemplarily, threshold screening is performed on the change duration sequence based on the partitioned weighted amplitude sequence. By setting a reasonable threshold standard, the first candidate time period with significant characteristics is selected from the change duration sequence. These time periods usually represent potential target activity areas or time periods where important events occur in the signal. Then, by performing merging and splitting operations on the first candidate time period, noise interference can be removed while retaining truly meaningful data segments to form a more accurate second candidate time period. The last step is to accurately locate the boundaries of the second candidate time period. This process involves further refining the starting and ending points of each candidate time period to ensure that the selected time period accurately reflects the actual occurrence range of the target activity, thereby obtaining the pre-selected sampling time zone.

[0058] Through multi-level screening and optimization, key information can be effectively extracted from complex photoelectric detection signals, which not only improves the detection accuracy, but also enhances the system's adaptability to different types of motion patterns.

[0059] In some embodiments, a probability prediction model is constructed based on a preset Gaussian distribution function, a partition weighted amplitude sequence and a third change feature, a preselected sampling time zone is input into the probability prediction model, and a target sampling time zone is output, including: normalizing the partition weighted amplitude sequence and the third change feature to obtain features of a uniform scale; weighted summing the normalized partition weighted amplitude sequence and the normalized third change feature to obtain a comprehensive feature vector; constructing a probability prediction model based on the comprehensive feature vector and a preset Gaussian distribution function; using the probability prediction model to calculate the likelihood probability of each preselected sampling time zone, and taking the preselected sampling time zone whose likelihood probability is greater than a preset probability threshold as the target sampling time zone.

[0060] A probability prediction model is constructed based on the above-mentioned comprehensive feature vector and the preset Gaussian distribution function, and the preset Gaussian distribution function is a multivariate Gaussian distribution function. The signal characteristics are converted into a calculable probability form by mathematical modeling. Using the constructed probability prediction model, the likelihood probability of each pre-selected sampling time zone is evaluated, that is, the possibility of the existence of the target activity in the time period is estimated. A preset probability threshold is set. Only when the likelihood probability of a pre-selected sampling time zone exceeds this probability threshold will it be identified as the target sampling time zone, and finally output as part of the signal detection information.

[0061] By combining signal features at different levels (such as amplitude distribution, time dynamic changes, etc.) and using statistical methods for comprehensive consideration, the accuracy of target recognition can be effectively improved, and the system's sensitivity to weak signals or atypical motion patterns in complex backgrounds can be enhanced.

[0062] See also Figure 2 , Figure 2 The embodiment of the present application also provides a schematic block diagram of a signal processing device of a photodetector, wherein the signal processing device 200 of the photodetector is used to execute the aforementioned signal processing method of the photodetector. The signal processing device 200 of the photodetector can be configured in a server.

[0063] Among them, the server can be an independent server or a server cluster, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN) and big data and artificial intelligence platforms.

[0064] like Figure 2 As shown, the signal processing device 200 of the photoelectric detector includes: a first calculation module 201, a second calculation module 202, an amplitude analysis module 203, a change analysis module 204, a time zone determination module 205 and a result output module 206.

[0065] The first calculation module 201 is used to respectively calculate the amplitude histograms of multiple digital signal frames, and calculate the frame difference based on every two adjacent digital signal frames to obtain preliminary motion features. The multiple digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector.

[0066] The second calculation module 202 is used to divide each digital signal frame into multiple sampling sub-regions and generate a partition weighting matrix, and perform weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix to obtain a partition weighted amplitude sequence.

[0067] The amplitude analysis module 203 is used to perform time series analysis on the partition weighted amplitude sequence to obtain a stable amplitude sequence, and input the stable amplitude sequence into a preset Gaussian mixture model for kernel density estimation to obtain amplitude probability characteristics.

[0068] The change analysis module 204 is used to determine the first change feature of each sampling point in the digital signal frame and its adjacent sampling points, obtain the second change feature according to the first change feature and the amplitude probability feature, and perform cluster analysis on the second change feature to obtain the third change feature.

[0069] The time zone determination module 205 is used to determine the pre-selected sampling time zone according to the third change feature and the partition weighted amplitude sequence.

[0070] The result output module 206 is used to build a probability prediction model based on a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, input the pre-selected sampling time zone into the probability prediction model, output the target sampling time zone, and output signal detection information according to the target sampling time zone.

[0071] An embodiment of the present application provides a photoelectric detection module, which includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement a signal processing method of a photoelectric detector such as any one of the embodiments of the present application when executing the computer program.

[0072] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor implements a signal processing method of a photodetector as described in any one of the embodiments of the present application.

[0073] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A signal processing method for a photodetector, characterized in that: The method comprises: Calculating amplitude histograms of a plurality of digital signal frames respectively, and calculating inter-frame differences according to every two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector; Divide each digital signal frame into a plurality of sampling sub-regions, generate a partition weighting matrix, and perform weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix to obtain a partition weighted amplitude sequence; Apply Fourier transform to the partition weighted amplitude sequence to obtain frequency domain data; calculate power spectrum density based on the frequency domain data to obtain frequency component distribution; perform threshold processing on the frequency component distribution to obtain main frequency components; filter the partition weighted amplitude sequence according to the main frequency components to obtain a denoised amplitude sequence; perform wavelet decomposition on the denoised amplitude sequence to obtain multi-scale coefficients; reconstruct the denoised amplitude sequence according to the multi-scale coefficients to obtain a reconstructed amplitude sequence; perform trend analysis on the reconstructed amplitude sequence to obtain local trends; calculate residuals based on the local trends to obtain a detrended amplitude sequence; perform time series analysis on the detrended amplitude sequence to obtain a stable amplitude sequence; perform preliminary statistical analysis on the stable amplitude sequence to obtain data distribution characteristics; initialize a Gaussian mixture model based on the data distribution characteristics to obtain an initial model; perform parameter optimization on the initial model to obtain an optimized Gaussian mixture model; use the optimized Gaussian mixture model to calculate kernel density estimation to obtain amplitude probability characteristics; According to each sampling point and its adjacent sampling points in the digital signal frame, determine a first change feature of each sampling point, obtain a second change feature according to the first change feature and the amplitude probability feature, and perform cluster analysis on the second change feature to obtain a third change feature; Determining a preselected sampling time zone according to the third variation characteristic and the partition weighted amplitude sequence; A probability prediction model is constructed according to a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, the preselected sampling time zone is input into the probability prediction model, the target sampling time zone is output, and signal detection information is output according to the target sampling time zone.

2. The signal processing method of the photodetector according to claim 1, characterized in that: The step of respectively calculating the amplitude histograms of the plurality of digital signal frames and calculating the inter-frame difference according to every two adjacent digital signal frames to obtain preliminary motion features includes: According to a preset number of sampling points, the original signal frame is divided into a plurality of digital signal frames; Calculating the amplitude distribution histogram of each digital signal frame to obtain amplitude distribution characteristics; Calculate the inter-frame difference between adjacent digital signal frames to obtain differential features; The amplitude distribution feature and the differential feature are combined to generate the preliminary motion feature.

3. The signal processing method of the photodetector according to claim 1, characterized in that: The step of dividing each digital signal frame into a plurality of sampling sub-regions and generating a partition weighting matrix comprises: Generate a partition weighting coefficient according to the positional relationship between each sampling sub-region and the midpoint of the time zone of the digital signal frame, wherein the closer the sampling sub-region is to the midpoint of the time zone, the larger the corresponding partition weighting coefficient; The partition weighting matrix is ​​generated according to the partition weighting coefficients.

4. The signal processing method of the photodetector according to claim 1, characterized in that: The method of determining a first change feature of each sampling point and its adjacent sampling points in the digital signal frame, obtaining a second change feature according to the first change feature and the amplitude probability feature, and performing cluster analysis on the second change feature to obtain a third change feature includes: Calculating the amplitude difference between each sampling point and an adjacent sampling point to obtain the first change feature; The first change feature is processed according to the amplitude probability feature to obtain a second change feature; According to the preset amplitude change partitions, cluster analysis is performed on the second change feature to obtain the third change feature.

5. The signal processing method of the photodetector according to claim 1, characterized in that: The step of determining the preselected sampling time zone according to the third variation feature and the partition weighted amplitude sequence comprises: Performing time continuity analysis on the third change feature to obtain a change duration sequence; Performing threshold screening on the change duration sequence according to the partition weighted amplitude sequence to obtain a first candidate time period; Merging and splitting the first candidate time periods to obtain second candidate time periods; The boundary of the second candidate time period is accurately located to obtain the preselected sampling time zone.

6. The signal processing method of the photodetector according to claim 1, characterized in that: The method of constructing a probability prediction model according to a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, inputting the preselected sampling time zone into the probability prediction model, and outputting a target sampling time zone comprises: Normalizing the partition weighted amplitude sequence and the third variation feature to obtain features of a unified scale; Perform weighted summation on the normalized partition weighted amplitude sequence and the normalized third change feature to obtain a comprehensive feature vector; Constructing a probability prediction model based on the comprehensive feature vector and a preset Gaussian distribution function; The probability prediction model is used to calculate the likelihood probability of each of the pre-selected sampling time zones, and the pre-selected sampling time zones whose likelihood probability is greater than a preset probability threshold are used as target sampling time zones.

7. A signal processing device for a photodetector, characterized in that: The signal processing device of the photoelectric detector comprises: A first calculation module is used to calculate the amplitude histograms of a plurality of digital signal frames respectively, and to calculate the inter-frame difference according to every two adjacent digital signal frames to obtain preliminary motion features, wherein the plurality of digital signal frames are obtained by segmenting the original signal frames output by the photoelectric detector; A second calculation module is used to divide each digital signal frame into a plurality of sampling sub-regions, and generate a partition weighting matrix, and perform weighted combination on the amplitude histograms corresponding to the sampling sub-regions according to the partition weighting matrix to obtain a partition weighted amplitude sequence; an amplitude analysis module, for applying Fourier transform to the partition weighted amplitude sequence to obtain frequency domain data; calculating power spectrum density according to the frequency domain data to obtain frequency component distribution; performing threshold processing on the frequency component distribution to obtain main frequency components; filtering the partition weighted amplitude sequence according to the main frequency components to obtain a denoised amplitude sequence; performing wavelet decomposition on the denoised amplitude sequence to obtain multi-scale coefficients; reconstructing the denoised amplitude sequence according to the multi-scale coefficients to obtain a reconstructed amplitude sequence; performing trend analysis on the reconstructed amplitude sequence to obtain a local trend; calculating residuals according to the local trend to obtain a detrended amplitude sequence; performing time series analysis on the detrended amplitude sequence to obtain a stable amplitude sequence; performing preliminary statistical analysis on the stable amplitude sequence to obtain data distribution characteristics; initializing a Gaussian mixture model according to the data distribution characteristics to obtain an initial model; performing parameter optimization on the initial model to obtain an optimized Gaussian mixture model; calculating kernel density estimation using the optimized Gaussian mixture model to obtain amplitude probability characteristics; a change analysis module, configured to determine a first change feature of each sampling point in the digital signal frame and its adjacent sampling points, obtain a second change feature according to the first change feature and the amplitude probability feature, and perform cluster analysis on the second change feature to obtain a third change feature; A time zone determination module, used to determine a preselected sampling time zone according to the third change feature and the partition weighted amplitude sequence; The result output module is used to build a probability prediction model based on a preset Gaussian distribution function, the partition weighted amplitude sequence and the third change feature, input the pre-selected sampling time zone into the probability prediction model, output the target sampling time zone, and output signal detection information according to the target sampling time zone.

8. A photoelectric detection module, characterized in that: The photoelectric detection module includes a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the signal processing method of the photoelectric detector according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the signal processing method of the photodetector according to any one of claims 1 to 6.

Images