Multi-sensor combined train detection method and system, medium and program product

Abstract

The invention discloses a multi-sensor combined train detection method and system, a medium and a program product, and the method comprises the steps: collecting optical feature data and acoustic feature data; calculating the running speed of the train; determining corrected optical feature data and corrected acoustic feature data; optical feature sequences and acoustic feature sequences of the train head, the train middle and the train tail are determined; extracting vehicle head contour structure features, compartment connection structure features and vehicle tail contour structure features; extracting train starting acoustic features, a wheel and steel rail impact sound spectrogram and train braking acoustic features; determining an optical feature recognition result; determining an acoustic feature recognition result; and according to the weights corresponding to the optical feature recognition result and the acoustic feature recognition result, calculating to obtain a train type recognition result. According to the method and the device, the accuracy of train type identification in severe weather is improved, and the accuracy and the real-time performance of railway transportation scheduling are further improved.

Description

Technical Field

[0001] The present application relates to the field of train model identification, and in particular to a multi-sensor combined train detection method, system, medium and program product. Background Art

[0002] In modern railway transportation systems, accurate identification of train types is of great significance for optimizing line resource allocation and improving transportation efficiency. At present, the identification of train types mainly relies on manual judgment or monitoring by detection devices at fixed locations. This method is prone to identification errors in bad weather or low light conditions, and manual judgment is inefficient, making it difficult to meet the growing demand for railway intelligence.

[0003] In related technologies, cameras can be deployed along the track to collect train images and use computer vision algorithms for intelligent analysis to achieve automatic identification of train types. This method uses a deep learning model to extract and classify image features, improving the automation level and efficiency of identification.

[0004] However, when encountering severe weather conditions such as rain, snow, fog and haze, the image quality deteriorates significantly, making it difficult for the system to accurately identify the type of train, reducing the accuracy and real-time performance of railway transportation scheduling. Summary of the invention

[0005] The present application provides a multi-sensor combined train detection method, system, medium and program product, which are used to improve the accuracy of train type identification in severe weather conditions, thereby improving the accuracy and real-time performance of railway transportation scheduling.

[0006] In a first aspect, the present application provides a multi-sensor combined train detection method, which collects optical feature data and acoustic feature data of a train during operation through an optical sensor array and an acoustic sensor array, respectively, wherein the optical sensor array includes at least two optical sensors spaced apart along a train running direction, and the acoustic sensor array includes at least two acoustic sensors spaced apart along a train running direction; Calculate the train running speed based on the time difference of optical characteristic data collected by adjacent optical sensors and the sensor spacing; Performing time-series alignment and synchronization processing on the optical characteristic data and the acoustic characteristic data according to the running speed of the train to obtain corrected optical characteristic data and corrected acoustic characteristic data; The corrected optical feature data is segmented and intercepted to obtain the optical feature sequences of the train head, the middle of the train and the rear of the train; the corrected acoustic feature data is segmented and intercepted to obtain the acoustic feature sequences of the train head, the middle of the train and the rear of the train; Extract the contour structure features of the train head according to the optical feature sequence of the train head, extract the car connection structure features according to the optical feature sequence of the middle part of the train, and extract the contour structure features of the train tail according to the optical feature sequence of the train tail; extract the train start acoustic features according to the acoustic feature sequence of the train head, extract the wheel and rail impact sound spectrum according to the acoustic feature sequence of the middle part of the train, and extract the train braking acoustic features according to the acoustic feature sequence of the train tail; Compare the feature data in the preset front vehicle profile feature library, the preset vehicle compartment connection structure feature library, and the preset rear vehicle profile feature library to obtain an optical feature recognition result; compare the feature data in the preset startup acoustic feature library, the preset impact sound spectrum feature library, and the preset braking acoustic feature library to obtain an acoustic feature recognition result; When the light intensity is lower than the first threshold or the visibility is lower than the second threshold, the acoustic feature recognition result is assigned a first weight, and the optical feature recognition result is assigned a second weight, and the first weight is greater than the second weight; when the ambient wind speed is higher than the third threshold, the optical feature recognition result is assigned a first weight, and the acoustic feature recognition result is assigned a second weight; The train type recognition result is obtained by calculating the weights corresponding to the optical feature recognition result and the acoustic feature recognition result.

[0007] The above technical solution is adopted to collect the characteristic data of the train running through the optical sensor array and the acoustic sensor array respectively, and the train running speed is calculated by using the time difference of adjacent optical sensors, and the two types of characteristic data are time-series aligned and synchronized based on the speed, thereby improving the time consistency of data collected by different types of sensors. The corrected characteristic data is segmented and the contour structure of the front of the vehicle, the connection structure of the carriage, the contour structure of the rear of the vehicle, the acoustic characteristics of the start-up, the impact spectrogram, and the acoustic characteristics of the brake are extracted, making the feature extraction more targeted and comprehensive. The recognition result is obtained by comparing with the preset feature library, and the weight of the optical and acoustic feature recognition results is dynamically adjusted according to the environmental conditions, so that a good recognition effect can be maintained in different environments. When the light intensity or visibility is poor, the weight of the acoustic feature is increased; when the ambient wind speed is high, the weight of the optical feature is increased. This adaptive weight allocation mechanism based on environmental parameters improves the accuracy of train type recognition and environmental adaptability.

[0008] In combination with some embodiments of the first aspect, in some embodiments, the corrected optical feature data is segmented to obtain optical feature sequences of the train head, the middle of the train, and the rear of the train; the corrected acoustic feature data is segmented to obtain acoustic feature sequences of the train head, the middle of the train, and the rear of the train, specifically including: Calculate the time window for the train to pass each sensor according to the train running speed; Adaptively segmenting the corrected optical feature data based on a time window, reducing the length of the time window when the train speed is greater than a preset speed, and increasing the length of the time window when the train speed is less than the preset speed; Synchronously segmenting the corrected acoustic feature data according to the adaptive segmentation time window; The signal-to-noise ratio of the optical feature data and the acoustic feature data in each time window is calculated, and the data segments with a signal-to-noise ratio lower than a preset threshold are deleted to obtain the optical feature sequence and the acoustic feature sequence of the train head, the middle of the train and the rear of the train.

[0009] By adopting the above technical solution, the time window is calculated based on the train running speed, and the time window length is adaptively adjusted. When the train speed is fast, the window length is shortened, and when the speed is slow, the window length is extended, so that the appropriate amount of feature data can be obtained under different speed conditions. By adaptively segmenting the optical feature data and synchronously segmenting the acoustic feature data based on this, the time correspondence of the two feature data segments is ensured. The signal-to-noise ratio of the feature data in each time window is calculated and the low signal-to-noise ratio data segments are eliminated, reducing the interference of environmental noise. This adaptive segmentation method combined with the signal-to-noise ratio screening mechanism improves the data quality of the feature sequence and lays a reliable data foundation for subsequent feature extraction and matching.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the train type recognition result is obtained according to the weight calculation corresponding to the optical feature recognition result and the acoustic feature recognition result, specifically including: The matching probabilities of the optical feature recognition results and each preset train type are calculated respectively to construct a first probability array; the matching probabilities of the acoustic feature recognition results and each preset train type are calculated respectively to construct a second probability array; The matching probability of each preset train type in the first probability array is multiplied by the first weight to obtain a first weighted probability array; the matching probability of each preset train type in the second probability array is multiplied by the second weight to obtain a second weighted probability array; Adding the probability values ​​of the corresponding preset train types in the first weighted probability array and the second weighted probability array to obtain a comprehensive probability value of each preset train type; The preset train type corresponding to the largest comprehensive probability value is selected as the train type identification result.

[0011] By adopting the above technical solution, the matching probabilities of the optical feature recognition results and the acoustic feature recognition results with each preset train type are calculated respectively, a probability array is constructed, and weighted calculation is performed according to the weights, thereby achieving effective fusion of the two feature recognition results. By adding the weighted probability values ​​of the corresponding train types in the first probability array and the second probability array, a comprehensive probability value is obtained, and the train type corresponding to the maximum probability value is selected as the recognition result, thus avoiding the one-sidedness that may be caused by single feature recognition. The complementary advantages of different features are utilized to improve the reliability of train type recognition.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after obtaining the train type recognition result according to the weight calculation corresponding to the optical feature recognition result and the acoustic feature recognition result, the method further includes: Obtaining a first confidence level of an optical feature recognition result and a second confidence level of an acoustic feature recognition result; When the difference between the first confidence level and the second confidence level is greater than a preset difference, it is marked as a pending verification state; Re-extract features and perform recognition matching on the recognition results in the pending verification state to obtain supplementary recognition results; The supplementary recognition results are weightedly fused with the train type recognition results to obtain the final train type recognition result.

[0013] By adopting the above technical solution, by calculating the first confidence of the optical feature recognition result and the second confidence of the acoustic feature recognition result, and comparing the confidence difference with the preset difference, the recognition result with a large deviation can be found in time. Re-feature extraction and recognition matching are performed on the recognition results with large confidence differences to obtain supplementary recognition results, which are weighted and fused with the original recognition results to obtain the final train type recognition result. This confidence-based secondary recognition mechanism and weighted fusion method reduce the uncertainty of the recognition results and improve the accuracy of train type recognition.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, obtaining a first confidence level of an optical feature recognition result and a second confidence level of an acoustic feature recognition result specifically includes: Calculating the similarity between the optical feature recognition result and feature data in a preset vehicle head contour feature library, a preset vehicle compartment connection structure feature library, and a preset vehicle tail contour feature library to obtain a first similarity; Calculating the similarity between the acoustic feature recognition result and feature data in a preset startup acoustic feature library, a preset impact sound spectrogram feature library, and a preset braking acoustic feature library to obtain a second similarity; A first confidence level is calculated based on the first similarity, and a second confidence level is calculated based on the second similarity.

[0015] By adopting the above technical solution, the first similarity is obtained by calculating the similarity between the optical feature recognition result and the feature data in the preset feature library, and the second similarity is obtained by calculating the similarity between the acoustic feature recognition result and the feature data in the preset feature library, and the first confidence and the second confidence are calculated accordingly, so that the reliability of the optical feature recognition result and the acoustic feature recognition result can be quantitatively evaluated. Since different types of trains have unique feature patterns in optical and acoustic features, the similarity calculation of the recognition result and the standard features in the preset feature library can accurately reflect the matching degree between the recognition result and the real train feature, thereby improving the reliability of the system evaluation of the recognition result.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, before re-extracting features and performing identification matching on the recognition result in a state to be verified, the method further includes: detecting the signal output strength of the optical sensor array and the acoustic sensor array; When the signal output strength is lower than a preset strength threshold, switching to the backup sensor; Adjust the sampling frequency and sampling accuracy of the optical sensor array and the acoustic sensor array according to the current environmental parameters; After confirming that the sampling frequency and sampling accuracy meet the preset feature extraction requirements, data collection is carried out.

[0017] By adopting the above technical solution, the quality and reliability of data acquisition are improved by detecting the signal output strength of the sensor array and switching to the backup sensor when the signal strength is insufficient, and dynamically adjusting the sampling parameters according to the environmental parameters. In a complex and changeable external environment, the performance of the sensor is easily interfered and affected. Timely detection of the signal output strength and switching to the backup sensor can avoid the degradation of data quality due to signal attenuation. By real-time monitoring of environmental parameters and adjusting the sampling frequency and accuracy accordingly, the system can adapt to the data acquisition requirements under different working conditions. This adaptive data acquisition mechanism ensures the quality of raw data acquisition, improves the accuracy of subsequent feature extraction and recognition matching, and enables the system to maintain stable recognition performance under various environmental conditions.

[0018] In combination with some embodiments of the first aspect, in some embodiments, weighted fusion of the supplementary recognition result and the train type recognition result is performed to obtain a final train type recognition result, specifically including: Calculate the feature matching probability of the supplementary recognition result; Determining a fusion weight coefficient of the train type recognition result based on the first confidence level and the second confidence level; Determine the fusion weight coefficient of the supplementary recognition result based on the feature matching probability; The train type recognition result and the supplementary recognition result are weighted according to the corresponding fusion weight coefficient to obtain the final train type recognition result.

[0019] By adopting the above technical solution, the optimized fusion of the recognition results is achieved by calculating the feature matching probability of the supplementary recognition results and determining the fusion weight coefficient based on the confidence. The feature matching probability of the supplementary recognition results reflects its matching degree with the preset features. Combined with the confidence index of the original recognition results, the system can assign reasonable fusion weights to the recognition results from different sources. The weight-based fusion method fully utilizes the advantageous information of different recognition results and weakens the impact of recognition results that may be biased. This weighted fusion method that takes into account confidence and feature matching probability improves the accuracy and reliability of the final recognition results, enabling the system to output more accurate train type determination results.

[0020] In the second aspect, an embodiment of the present application provides a multi-sensor combined train detection system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to enable the system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a system, causes the system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer program product, characterized in that when the computer program product runs on a system, the system executes the method described in any possible implementation manner in the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The present application provides a multi-sensor combined train detection method. The optical sensor array and the acoustic sensor array are used to collect the characteristic data of the train during operation, and the time difference of adjacent optical sensors is used to calculate the train speed. Based on the speed, the two types of characteristic data are time-series aligned and synchronized, thereby improving the time consistency of data collected by different types of sensors. The corrected characteristic data is segmented and the contour structure of the front of the vehicle, the connection structure of the carriage, the contour structure of the rear of the vehicle, the acoustic characteristics of the start-up, the impact spectrogram, and the acoustic characteristics of the brake are extracted, so that the feature extraction is more targeted and comprehensive. The recognition result is obtained by comparing with the preset feature library, and the weights of the optical and acoustic feature recognition results are dynamically adjusted according to the environmental conditions, so that a good recognition effect can be maintained in different environments. When the light intensity or visibility is poor, the weight of the acoustic feature is increased; when the ambient wind speed is high, the weight of the optical feature is increased. This adaptive weight allocation mechanism based on environmental parameters improves the accuracy of train type recognition and environmental adaptability.

[0024] 2. The present application provides a multi-sensor combined train detection method, which calculates the first confidence of the optical feature recognition result and the second confidence of the acoustic feature recognition result, and compares the confidence difference with the preset difference, so as to promptly find the recognition result with a large deviation. Re-feature extraction and recognition matching are performed on the recognition results with large confidence differences to obtain supplementary recognition results, which are weighted and fused with the original recognition results to obtain the final train type recognition result. This confidence-based secondary recognition mechanism and weighted fusion method reduce the uncertainty of the recognition results and improve the accuracy of train type recognition.

[0025] 3. The present application provides a multi-sensor combined train detection method, which improves the quality and reliability of data acquisition by detecting the signal output strength of the sensor array and switching to the backup sensor when the signal strength is insufficient, and dynamically adjusting the sampling parameters according to the environmental parameters. In a complex and changeable external environment, the performance of the sensor is easily interfered and affected. Timely detection of the signal output strength and switching of the backup sensor can avoid the degradation of data quality due to signal attenuation. By monitoring the environmental parameters in real time and adjusting the sampling frequency and accuracy accordingly, the system can adapt to the data acquisition requirements under different working conditions. This adaptive data acquisition mechanism ensures the acquisition quality of the original data, improves the accuracy of subsequent feature extraction and recognition matching, and enables the system to maintain stable recognition performance under various environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of a multi-sensor combined train detection method in an embodiment of the present application.

[0027] Figure 2It is a flow chart of an optimization method based on confidence assessment in an embodiment of the present application.

[0028] Figure 3 It is a schematic diagram of the physical device structure of a multi-sensor combined train detection system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more listed items.

[0030] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0031] The following uses an embodiment and combines Figure 1 , a train detection method combining multiple sensors in an embodiment of the present application is described: See also Figure 1 , is a flow chart of a multi-sensor combined train detection method in an embodiment of the present application.

[0032] S101, collecting optical characteristic data and acoustic characteristic data of the train during operation through an optical sensor array and an acoustic sensor array respectively; The system collects optical characteristic data and acoustic characteristic data of the train during operation through an optical sensor array and an acoustic sensor array respectively. The optical sensor array includes at least two optical sensors spaced apart along the direction of the train's operation, and the acoustic sensor array includes at least two acoustic sensors spaced apart along the direction of the train's operation.

[0033] In this step, the system collects optical characteristic data and acoustic characteristic data of the train during operation through the optical sensor array and acoustic sensor array arranged along the track. The optical sensor array includes at least two optical sensors arranged at intervals along the direction of train operation, which are used to collect image data of the train during operation, such as images of the front, carriage, and rear of the train. The acoustic sensor array includes at least two acoustic sensors arranged at intervals along the direction of train operation, which are used to collect sound data generated by the train during operation, such as the sound of train starting, the sound of wheels hitting rails, and the sound of train braking.

[0034] The system can use various types of optical sensors to collect optical feature data, such as visible light cameras, infrared cameras, etc., and obtain image data of different parts of the train by arranging optical sensors at different angles and positions. At the same time, the system can use acoustic sensors such as microphone arrays and vibration sensors to collect acoustic feature data, and obtain sound data when the train is running by reasonably arranging the position and number of acoustic sensors. In addition, the system can also dynamically adjust the sampling frequency and sensitivity of optical sensors and acoustic sensors according to actual needs to adapt to different train speeds and environmental conditions.

[0035] S102, calculating the train running speed according to the time difference of the optical characteristic data collected by adjacent optical sensors and the sensor spacing; In this step, the system calculates the train speed based on the time difference of the optical feature data collected by adjacent optical sensors and the sensor spacing. Specifically, when a train passes through two adjacent optical sensors in sequence, it will be captured by the sensors at different times. By recording the time difference between the two times and combining it with the distance between the two sensors, the average speed of the train passing through this section can be calculated.

[0036] The system can use a variety of methods to improve the accuracy of speed calculation, such as considering the impact of train length on time difference, calculating the speed through the time difference between the front and rear of the train passing through the sensor; increasing the number of optical sensors, calculating the average speed through the speed of multiple intervals; combining the data of optical sensors and acoustic sensors, estimating the speed through the frequency change of sound signals, etc. In addition, the system can also introduce algorithms such as Kalman filtering to smooth the calculated speed and reduce the impact of outliers.

[0037] S103, performing time sequence alignment and synchronization processing on the optical characteristic data and the acoustic characteristic data according to the train running speed, to obtain corrected optical characteristic data and corrected acoustic characteristic data; In this step, the system performs time alignment and synchronization processing on the optical feature data and acoustic feature data according to the calculated train speed to obtain corrected optical feature data and acoustic feature data. Due to the different layout positions of the optical sensor and the acoustic sensor, the collected data may be misaligned in time, and time alignment is required. At the same time, due to the change in the train speed, the data collected at different times may correspond to different positions of the train, and synchronization processing is required to unify the data to the same time and space reference.

[0038] The system can use digital signal processing technologies such as interpolation and resampling to achieve timing alignment and synchronization of data. Specifically, the corresponding time when different sensors collect data can be calculated based on the train speed and sensor layout, and the data can be aligned through timestamps; then, the train position corresponding to each data point can be calculated based on the aligned timestamp and train speed, and the data can be synchronized to a unified position reference through spatial mapping. In addition, the system can also use data fusion technology, such as Kalman filtering, to fuse optical features and acoustic features to improve the temporal and spatial consistency of data.

[0039] S104, segmentally intercepting the corrected optical feature data to obtain optical feature sequences of the train head, the middle of the train, and the train tail; segmentally intercepting the corrected acoustic feature data to obtain acoustic feature sequences of the train head, the middle of the train, and the train tail; The system segments the corrected optical feature data to obtain optical feature sequences of the train head, middle and rear parts; segments the corrected acoustic feature data to obtain acoustic feature sequences of the train head, middle and rear parts, specifically: the time window of the train passing through each sensor is calculated according to the train running speed; the corrected optical feature data is adaptively segmented based on the time window, when the train running speed is greater than the preset speed, the length of the time window is reduced, and when the train running speed is less than the preset speed, the length of the time window is increased; the corrected acoustic feature data is synchronously segmented according to the adaptively segmented time window; the signal-to-noise ratio of the optical feature data and the acoustic feature data in each time window is calculated, and the data segments with a signal-to-noise ratio lower than a preset threshold are deleted to obtain the optical feature sequences and acoustic feature sequences of the train head, middle and rear parts.

[0040] In this step, the system segments the corrected optical feature data and acoustic feature data to obtain the optical feature sequence and acoustic feature sequence of the train head, middle and rear. Specifically, the system first calculates the time window of the train passing each sensor according to the train running speed, and then adaptively segments the corrected optical feature data based on the time window. When the train running speed is high, the length of the time window is reduced to ensure the consistency of the train position in each segment; when the train running speed is low, the length of the time window is increased to obtain more feature information. Then, the system synchronously segments the corrected acoustic feature data according to the segmented time window of the optical feature data to ensure the time correspondence between the optical and acoustic features. Finally, the system calculates the signal-to-noise ratio of the optical and acoustic feature data in each time window, eliminates the data segments with a signal-to-noise ratio lower than the preset threshold, and obtains high-quality optical feature sequences and acoustic feature sequences of the train head, middle and rear.

[0041] The system can use a variety of signal processing technologies to achieve adaptive segmentation and noise removal of feature data. For example, the sliding window and dynamic threshold methods can be used to dynamically adjust the size and position of the segmentation window according to the train speed and data quality; time-frequency analysis methods such as wavelet transform and EMD can be used to extract the time-frequency characteristics of the data and identify the characteristic patterns of different parts of the train; blind source separation, independent component analysis and other technologies can be used to separate the acoustic characteristics of the train itself from the noisy environmental noise. In addition, the system can also combine prior knowledge, such as the physical size of the train, typical characteristic patterns, etc., to verify and optimize the segmentation and denoising results, and improve the accuracy and reliability of feature extraction.

[0042] In actual applications, the features of different parts of the train may be similar, and the influence of environmental noise may lead to deviations in feature extraction and segmentation. To address this problem, the system can introduce a multi-sensor fusion strategy, comprehensively utilizing multi-modal information such as optics, acoustics, and vibration to enhance the distinguishability and robustness of features; at the same time, the system can also use an adaptive learning algorithm to accumulate train operation data, continuously optimize feature extraction and segmentation models, improve the system's ability to identify different parts of the train, and lay a good data foundation for subsequent train type identification.

[0043] S105, extracting the contour structure features of the train head according to the optical feature sequence of the train head, extracting the car connection structure features according to the optical feature sequence of the middle part of the train, and extracting the contour structure features of the train tail according to the optical feature sequence of the train tail; extracting the train start acoustic features according to the acoustic feature sequence of the train head, extracting the wheel and rail impact sound spectrogram according to the acoustic feature sequence of the middle part of the train, and extracting the train braking acoustic features according to the acoustic feature sequence of the train tail; In this step, the system extracts the front contour structure features, car connection structure features and rear contour structure features from the optical feature sequences of the train head, middle and rear parts, and extracts the train start acoustic features, wheel and rail impact sound spectrogram and train braking acoustic features from the acoustic feature sequences of the corresponding parts. These features describe the structure and acoustic properties of each part of the train from different angles, providing rich discriminant information for subsequent train type identification.

[0044] The system can use computer vision and pattern recognition methods to extract the structural features of various parts of the train from the optical image sequence. For example, edge detection, region segmentation and other image processing technologies can be used to extract the contour features of the front and rear of the train; image matching, feature point tracking and other algorithms can be used to identify the connection structure between carriages; deep learning target detection and semantic segmentation models can be used to automatically identify and locate key components of the train. For acoustic features, the system can use speech recognition and signal processing technologies to extract the acoustic fingerprint of the train in different states from the sound signal. For example, short-time Fourier transform and other time-frequency analysis methods can be used to extract the sound spectrum features during the train start and braking process; voiceprint recognition and sound source localization technology can be used to analyze the sound pattern generated by the collision between the wheel and the rail; machine learning classification and regression algorithms can be used to establish a mapping relationship between the train's acoustic features and its operating status.

[0045] S106, comparing feature data in a preset vehicle front profile feature library, a preset vehicle compartment connection structure feature library, and a preset vehicle rear profile feature library to obtain an optical feature recognition result; comparing feature data in a preset startup acoustic feature library, a preset impact sound spectrum feature library, and a preset braking acoustic feature library to obtain an acoustic feature recognition result; In this step, the system compares the extracted optical and acoustic features of each part of the train with the preset feature library to obtain the optical feature recognition results and the acoustic feature recognition results. Among them, the optical feature recognition determines the type of train by comparing the features such as the front contour, the car connection structure and the rear contour; the acoustic feature recognition determines the type of train by comparing the sound features such as the train start, wheel impact and braking. The preset feature library stores the standard feature data of different models of trains, providing a reliable reference for train type identification.

[0046] The system can use methods such as template matching and similarity measurement to compare the extracted features with the preset feature library. For example, use metric functions such as Euclidean distance and cosine similarity to calculate the distance or similarity between the features to be identified and the standard features; use machine learning classifiers such as support vector machines and K nearest neighbors to distinguish the type of train based on feature similarity; use deep learning feature embedding and metric learning technology to map high-dimensional features to low-dimensional space, enhance the discriminative power and generalization ability of features, etc. At the same time, the system can also adopt different comparison strategies for the features of different parts of the train, such as using a shape-based matching algorithm for the contour features of the front and rear of the train, a texture-based recognition method for the structural features of the carriage connection, and a spectrum-based analysis technology for the acoustic features, etc., to give full play to the advantages and complementarity of various features.

[0047] S107, when the light intensity is lower than the first threshold or the visibility is lower than the second threshold, assigning a first weight to the acoustic feature recognition result and a second weight to the optical feature recognition result, wherein the first weight is greater than the second weight; when the ambient wind speed is higher than the third threshold, assigning a first weight to the optical feature recognition result and a second weight to the acoustic feature recognition result; In this step, the system assigns different weights to the optical feature recognition results and the acoustic feature recognition results according to the environmental conditions. When the light intensity is low or the visibility is low, the image quality collected by the optical sensor decreases, and the discriminative power of the optical feature is weakened. At this time, the acoustic feature contributes more to train recognition, so a higher weight is assigned to the acoustic feature recognition result; conversely, when the ambient wind speed is high, the audio data collected by the acoustic sensor is easily disturbed by wind noise, and the discriminative power of the acoustic feature decreases. At this time, the optical feature contributes more to train recognition, so a higher weight is assigned to the optical feature recognition result. By dynamically adjusting the weights of optical and acoustic features, the complementary advantages of the two types of features can be fully utilized to improve the environmental adaptability of train recognition.

[0048] The system can obtain environmental condition information in a variety of ways, such as using illuminance sensors, transmissive light meters, etc. to measure light intensity, using visibility meters, image contrast analysis, etc. to assess visibility levels, and using anemometers, ultrasonic anemometers, etc. to monitor ambient wind speeds. Based on the acquired environmental parameters, the system can set a series of thresholds to divide environmental conditions into different levels, and predefine the weight values ​​of optical and acoustic features at each level. In the actual recognition process, the system detects environmental parameters in real time, determines which level it currently belongs to, and dynamically adjusts the weight coefficients of the optical and acoustic feature recognition results to achieve adaptive optimization of the recognition strategy.

[0049] S108. Obtain a train type recognition result according to the weight calculation corresponding to the optical feature recognition result and the acoustic feature recognition result.

[0050] The system obtains the train type recognition result according to the weight calculation corresponding to the optical feature recognition result and the acoustic feature recognition result. Specifically: the matching probability of the optical feature recognition result and each preset train type is calculated respectively to construct a first probability array; the matching probability of the acoustic feature recognition result and each preset train type is calculated respectively to construct a second probability array; the matching probability of each preset train type in the first probability array is multiplied by the first weight to obtain a first weighted probability array; the matching probability of each preset train type in the second probability array is multiplied by the second weight to obtain a second weighted probability array; the probability values ​​of the corresponding preset train types in the first weighted probability array and the second weighted probability array are added to obtain a comprehensive probability value of each preset train type; the preset train type corresponding to the largest comprehensive probability value is selected as the train type recognition result.

[0051] In this step, the system comprehensively considers the optical feature recognition results and the acoustic feature recognition results, and calculates the final train type recognition result based on the weight coefficients of the two types of features. Specifically, the system first calculates the matching probability of the optical features and acoustic features with various preset train types to obtain two probability arrays; then, each element in the two probability arrays is multiplied by the weight coefficient of the corresponding feature to obtain a weighted probability array; finally, the elements in the corresponding positions in the two weighted probability arrays are added together to obtain the comprehensive matching probability of each preset train type, and the one with the largest probability is selected as the train type recognition result. By weighted fusion of the recognition results of optical features and acoustic features, the limitations of a single feature are overcome, the complementary advantages of multi-source information are brought into play, and the accuracy and reliability of train type recognition are effectively improved.

[0052] The system can use uncertainty reasoning methods such as Bayesian decision making and DS evidence theory to achieve weighted fusion of optical and acoustic feature recognition results. For example, using the Bayesian formula, the matching probability of optical and acoustic features is regarded as the prior probability, and the feature weight under environmental conditions is regarded as the likelihood probability. The comprehensive discrimination result of the train type is obtained by calculating the posterior probability; or using DS theory, the matching results of optical and acoustic features are regarded as different sources of evidence, and the confidence of each train type is calculated through evidence combination rules, and the one with the highest confidence is selected as the final recognition result. During the fusion process, the system can also introduce an adaptive weight adjustment mechanism to dynamically optimize the weight distribution of optical and acoustic features based on factors such as the accuracy of historical recognition and the trend of environmental changes, so as to continuously improve the self-learning and optimization capabilities of the train recognition algorithm.

[0053] In the above embodiment, the characteristic data of the train running are collected by the optical sensor array and the acoustic sensor array respectively, the train running speed is calculated by using the time difference of adjacent optical sensors, and the two types of characteristic data are time-series aligned and synchronized based on the speed, thereby improving the time consistency of data collected by different types of sensors. The corrected characteristic data is segmented and the contour structure of the front of the train, the connection structure of the carriage, the contour structure of the rear of the train, the acoustic characteristics of the start-up, the impact spectrogram, and the acoustic characteristics of the brake are extracted, so that the feature extraction is more targeted and comprehensive. The recognition result is obtained by comparing with the preset feature library, and the weights of the optical and acoustic feature recognition results are dynamically adjusted according to the environmental conditions, so that a good recognition effect can be maintained in different environments. When the light intensity or visibility is poor, the weight of the acoustic feature is increased; when the ambient wind speed is high, the weight of the optical feature is increased. This adaptive weight allocation mechanism based on environmental parameters improves the accuracy of train type recognition and environmental adaptability.

[0054] The above embodiment describes the basic train detection process, and the train type is determined through multi-sensor data collection, feature extraction and weighted recognition. However, in actual applications, due to factors such as environmental interference and sensor performance fluctuations, a single recognition result may have a certain degree of uncertainty. In order to further improve the reliability of the recognition result, this application also provides an optimization method based on confidence evaluation. Figure 2 , an optimization method based on confidence evaluation in an embodiment of the present application is described: See also Figure 2 , which is a flow chart of an optimization method based on confidence assessment in an embodiment of the present application.

[0055] S201, obtaining a first confidence level of an optical feature recognition result and a second confidence level of an acoustic feature recognition result; The system obtains a first confidence of the optical feature recognition result and a second confidence of the acoustic feature recognition result, specifically: calculating the similarity between the optical feature recognition result and feature data in a preset front vehicle contour feature library, a preset vehicle compartment connection structure feature library, and a preset rear vehicle contour feature library to obtain a first similarity; calculating the similarity between the acoustic feature recognition result and feature data in a preset starting acoustic feature library, a preset impact sound spectrum feature library, and a preset braking acoustic feature library to obtain a second similarity; calculating a first confidence based on the first similarity, and calculating a second confidence based on the second similarity.

[0056] In this step, the system obtains the first confidence of the optical feature recognition result and the second confidence of the acoustic feature recognition result. Specifically, the system first calculates the similarity between the optical feature recognition result and the feature data in the preset front contour feature library, the car body connection structure feature library and the rear contour feature library to obtain the first similarity; then calculates the similarity between the acoustic feature recognition result and the feature data in the preset startup acoustic feature library, the impact sound spectrum feature library and the braking acoustic feature library to obtain the second similarity; finally, the system calculates the first confidence based on the first similarity and the second confidence based on the second similarity. Confidence reflects the credibility of the recognition result and is an important indicator for evaluating the performance of the recognition algorithm.

[0057] The system can use a variety of similarity measurement methods to calculate the similarity between feature recognition results and the preset feature library, such as Euclidean distance, Mahalanobis distance, cosine similarity, etc. For optical features, key features such as the front contour, car body connection structure, and rear contour in the recognition results can be extracted and compared with the standard features in the preset feature library to calculate the similarity of features such as shape and texture; for acoustic features, key features such as the start-up sound, impact sound spectrogram, and braking sound in the recognition results can be extracted and compared with the standard features in the preset feature library to calculate the similarity of features such as spectrum and energy. When calculating confidence, the system can comprehensively consider multiple similarity indicators, such as weighted average, geometric mean, etc., and can also introduce prior knowledge, such as the importance of different features, historical recognition accuracy, etc., and estimate the distribution of confidence through methods such as Bayesian reasoning.

[0058] S202: When the difference between the first confidence level and the second confidence level is greater than a preset difference, mark the state as pending verification; In this step, the system determines whether the confidence difference between the optical feature recognition result and the acoustic feature recognition result exceeds the preset threshold. If it exceeds, the current recognition result is marked as pending verification. This situation usually indicates that there is a large difference between the recognition results of the optical feature and the acoustic feature, and the recognition credibility of a single feature is insufficient, requiring further verification and optimization. After being marked as pending verification, the system detects the signal output intensity of the optical sensor array and the acoustic sensor array. If it is lower than the preset intensity threshold, it switches to the backup sensor to ensure the continuity and reliability of data collection. At the same time, the system adjusts the sampling frequency and sampling accuracy of the sensor array according to the current environmental parameters (such as light intensity, noise level, etc.), confirms that the adjusted parameters meet the preset feature extraction requirements, and then starts the data re-collection process.

[0059] The system can set multiple confidence difference thresholds related to factors such as train speed and environmental conditions, and dynamically adjust the thresholds according to actual conditions to balance the requirements of recognition efficiency and recognition accuracy. When switching backup sensors, the system can comprehensively consider the spatial distribution, measurement accuracy, failure rate and other characteristics of the sensors, select the optimal backup sensor combination, and ensure the data quality of the backup sensors through sensor calibration, data synchronization and other processing. For the adjustment of sampling frequency and sampling accuracy, the system can adaptively optimize the sampling parameters based on the sampling theorem and reconstruction theory of signal processing, combined with prior knowledge such as train vibration frequency and sound frequency band, to meet the requirements of feature extraction while reducing data redundancy and computational overhead.

[0060] After that, the system detects the signal output strength of the optical sensor array and the acoustic sensor array; When the signal output strength is lower than a preset strength threshold, switching to the backup sensor; Adjust the sampling frequency and sampling accuracy of the optical sensor array and the acoustic sensor array according to the current environmental parameters; After confirming that the sampling frequency and sampling accuracy meet the preset feature extraction requirements, data collection is carried out.

[0061] For situations where the confidence difference is large and secondary recognition is required, the system will first detect the working status of the sensor array. By real-time monitoring of the signal output strength of the optical sensor array and the acoustic sensor array, once the signal strength is found to be lower than the preset threshold, indicating that the current sensor performance may be attenuated or faulty, the system will immediately enable the backup sensor to ensure the quality of data collection. At the same time, the system will dynamically adjust the sampling parameters of the sensor array, including sampling frequency and sampling accuracy, according to the current environmental parameters (such as lighting conditions, temperature, humidity, etc.) to adapt it to the current working environment. Only after confirming that the adjusted sampling parameters meet the minimum requirements for feature extraction, the system will start a new round of data collection. This adaptive data collection mechanism ensures the reliability of the data required for secondary recognition.

[0062] S203, re-extracting features and performing recognition matching on the recognition result in the pending verification state to obtain a supplementary recognition result; In this step, the system starts the re-identification process for the recognition results marked as pending verification, and obtains supplementary recognition results by re-extracting features, matching and identifying, etc. Specifically, the system first uses the sensor data re-collected in step S202 to extract optical features and acoustic features; then, the extracted features are matched and identified with the preset feature library to obtain supplementary optical feature recognition results and acoustic feature recognition results; finally, the supplementary recognition results are compared and verified with the original recognition results, and the consistency and complementarity of the recognition results are comprehensively analyzed to provide more reliable criteria for subsequent weighted fusion.

[0063] When re-extracting features, the system can use feature extraction algorithms and parameter settings that are different from the original recognition process, such as finer-grained image segmentation, higher-order spectrum analysis, etc., to mine deeper feature information in the data; at the same time, the system can also introduce new feature representation and feature selection methods, such as feature embedding based on deep learning, feature weighting based on attention mechanism, etc., to highlight the discriminative role of key features and suppress the interference of redundant features. When re-identifying and matching, the system can use stricter similarity thresholds and more sophisticated decision rules to improve the generalization performance and anti-interference ability of the recognition algorithm; in addition, the system can also fully utilize the complementary advantages of different recognizers and reduce the limitations of a single recognizer through strategies such as multi-algorithm integration and multi-model fusion.

[0064] S204: weighted fusion of the supplementary recognition result and the train type recognition result to obtain a final train type recognition result.

[0065] The system performs weighted fusion of the supplementary recognition results and the train type recognition results to obtain the final train type recognition results. Specifically: the feature matching probability of the supplementary recognition results is calculated; the fusion weight coefficient of the train type recognition results is determined based on the first confidence level and the second confidence level; the fusion weight coefficient of the supplementary recognition results is determined based on the feature matching probability; the train type recognition results and the supplementary recognition results are weightedly calculated according to the corresponding fusion weight coefficients to obtain the final train type recognition results.

[0066] In this step, the system comprehensively utilizes the original recognition results and the supplementary recognition results, and obtains the final train type recognition results by weighted fusion. Specifically, the system first calculates the matching probability of the optical features and acoustic features in the supplementary recognition results; then, based on the first confidence and second confidence of the original recognition results, the corresponding fusion weight coefficient is determined; at the same time, based on the feature matching probability of the supplementary recognition results, the corresponding fusion weight coefficient is determined; finally, the original recognition results and the supplementary recognition results are weighted and summed according to their respective fusion weight coefficients to obtain the final train type recognition results. By fusing the information of the original recognition and the supplementary recognition, the accuracy and reliability of train type recognition can be effectively improved, and the risk of recognition errors and missed judgments can be reduced.

[0067] When determining the fusion weight coefficient, the system can comprehensively consider factors such as the confidence of the original recognition results and the supplementary recognition results, the probability of feature matching, etc., and transform the qualitative decision rules into quantitative weight allocation strategies by setting empirical thresholds and constructing discriminant functions. For example, when the confidence of the original recognition result is high and consistent, it is given a larger weight coefficient; when the feature matching probability of the supplementary recognition result is significantly higher than that of the original recognition result, it is given a larger weight coefficient. At the same time, the system can also introduce an adaptive weight adjustment mechanism to adjust the fusion weight in real time according to dynamic factors such as train operating conditions and recognition task difficulty, and weigh the contribution of original recognition and supplementary recognition. For example, in bad weather, the discriminability of acoustic features may decrease, and the weight coefficient of optical features should be appropriately increased at this time.

[0068] In the above embodiment, by calculating the first confidence of the optical feature recognition result and the second confidence of the acoustic feature recognition result, and comparing the confidence difference with the preset difference, the recognition result with a large deviation can be found in time. Re-feature extraction and recognition matching are performed on the recognition results with large confidence differences to obtain supplementary recognition results, which are weighted and fused with the original recognition results to obtain the final train type recognition result. This confidence-based secondary recognition mechanism and weighted fusion method reduce the uncertainty of the recognition results and improve the accuracy of train type recognition.

[0069] The following describes the system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of the physical device structure of a multi-sensor combined train detection system provided in an embodiment of the present application.

[0070] It should be noted that Figure 3 The structure of the system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0071] like Figure 3As shown, the system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0072] The following components are connected to the I / O interface 305: an input section 306 including a camera, an infrared sensor, etc.; an output section 307 including a liquid crystal display (LCD) and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0073] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0074] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.

[0075] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0076] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the system described in the above embodiment; or may exist independently without being assembled into the system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of a system, the system implements the method provided in the above embodiment.

[0077] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0078] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0079] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0080] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A multi-sensor combined train detection method, characterized in that: include: The optical characteristic data and the acoustic characteristic data of the train during running are collected respectively by an optical sensor array and an acoustic sensor array, wherein the optical sensor array comprises at least two optical sensors spaced apart along the running direction of the train, and the acoustic sensor array comprises at least two acoustic sensors spaced apart along the running direction of the train; Calculating the train running speed according to the time difference of the optical characteristic data collected by adjacent optical sensors and the sensor spacing; Performing time-series alignment and synchronization processing on the optical characteristic data and the acoustic characteristic data according to the running speed of the train to obtain corrected optical characteristic data and corrected acoustic characteristic data; The corrected optical feature data is segmented to obtain optical feature sequences of the train head, the middle of the train and the rear of the train; the corrected acoustic feature data is segmented to obtain acoustic feature sequences of the train head, the middle of the train and the rear of the train; Extracting the contour structure features of the train head according to the optical feature sequence of the train head, extracting the car connection structure features according to the optical feature sequence of the middle part of the train, and extracting the contour structure features of the train tail according to the optical feature sequence of the train tail; extracting the train start acoustic features according to the acoustic feature sequence of the train head, extracting the wheel and rail impact sound spectrogram according to the acoustic feature sequence of the middle part of the train, and extracting the train braking acoustic features according to the acoustic feature sequence of the train tail; Compare the feature data in the preset vehicle head contour feature library, the preset vehicle body connection structure feature library and the preset vehicle tail contour feature library to obtain the optical feature recognition result; Comparing feature data in a preset startup acoustic feature library, a preset impact sound spectrogram feature library, and a preset braking acoustic feature library to obtain an acoustic feature recognition result; When the light intensity is lower than a first threshold or the visibility is lower than a second threshold, assigning a first weight to the acoustic feature recognition result and assigning a second weight to the optical feature recognition result, wherein the first weight is greater than the second weight; When the ambient wind speed is higher than a third threshold, assigning the first weight to the optical feature recognition result, and assigning the second weight to the acoustic feature recognition result; The train type recognition result is obtained by calculating the weights corresponding to the optical feature recognition result and the acoustic feature recognition result.

2. The method according to claim 1, characterized in that The step of segmenting the corrected optical feature data to obtain optical feature sequences of the train head, the middle of the train, and the rear of the train; and segmenting the corrected acoustic feature data to obtain acoustic feature sequences of the train head, the middle of the train, and the rear of the train specifically includes: Calculate the time window for the train to pass each sensor according to the running speed of the train; Adaptively segmenting the corrected optical feature data based on the time window, and when the train running speed is greater than a preset speed, reducing the length of the time window, and when the train running speed is less than the preset speed, increasing the length of the time window; Synchronously segmenting the corrected acoustic feature data according to the adaptive segmented time window; The signal-to-noise ratio of the optical feature data and the acoustic feature data in each of the time windows is calculated, and the data segments with the signal-to-noise ratio lower than a preset threshold are deleted to obtain the optical feature sequence and the acoustic feature sequence of the train head, the train middle and the train tail.

3. The method according to claim 1, characterized in that The train type recognition result is obtained by calculating the weights corresponding to the optical feature recognition result and the acoustic feature recognition result, specifically including: Calculating the matching probability between the optical feature recognition result and each preset train type respectively, and constructing a first probability array; calculating the matching probability between the acoustic feature recognition result and each preset train type respectively, and constructing a second probability array; Multiplying the matching probability of each preset train type in the first probability array by the first weight respectively to obtain a first weighted probability array; multiplying the matching probability of each preset train type in the second probability array by the second weight respectively to obtain a second weighted probability array; Adding the probability values ​​of the preset train types in the first weighted probability array and the second weighted probability array to obtain a comprehensive probability value of each preset train type; The preset train type corresponding to the largest comprehensive probability value is selected as the train type identification result.

4. The method according to claim 1, characterized in that After obtaining the train type recognition result by calculating the weights corresponding to the optical feature recognition result and the acoustic feature recognition result, the method further includes: Obtaining a first confidence level of the optical feature recognition result and a second confidence level of the acoustic feature recognition result; When the difference between the first confidence level and the second confidence level is greater than a preset difference, marking it as a pending verification state; Re-extracting features and performing identification matching on the identification result in the pending verification state to obtain a supplementary identification result; The supplementary recognition result is weightedly fused with the train type recognition result to obtain a final train type recognition result.

5. The method according to claim 4, characterized in that The obtaining of the first confidence level of the optical feature recognition result and the second confidence level of the acoustic feature recognition result specifically includes: Calculating the similarity between the optical feature recognition result and feature data in the preset vehicle head contour feature library, the preset vehicle compartment connection structure feature library, and the preset vehicle tail contour feature library to obtain a first similarity; Calculating the similarity between the acoustic feature recognition result and feature data in the preset startup acoustic feature library, the preset impact sound spectrogram feature library, and the preset braking acoustic feature library to obtain a second similarity; The first confidence level is calculated according to the first similarity, and the second confidence level is calculated according to the second similarity.

6. The method according to claim 4, characterized in that Before re-extracting features and performing identification matching on the recognition result in the to-be-verified state, the method further includes: detecting the signal output strength of the optical sensor array and the acoustic sensor array; When the signal output intensity is lower than a preset intensity threshold, switching to a backup sensor; Adjusting the sampling frequency and sampling accuracy of the optical sensor array and the acoustic sensor array according to current environmental parameters; After confirming that the sampling frequency and the sampling accuracy meet the preset feature extraction requirements, data collection is performed.

7. The method according to claim 4, characterized in that The weighted fusion of the supplementary recognition result and the train type recognition result to obtain the final train type recognition result specifically includes: Calculating a feature matching probability of the supplementary recognition result; Determining a fusion weight coefficient of the train type identification result based on the first confidence level and the second confidence level; Determining a fusion weight coefficient of the supplementary recognition result based on the feature matching probability; The train type recognition result and the supplementary recognition result are weighted and calculated according to the corresponding fusion weight coefficient to obtain a final train type recognition result.

8. A multi-sensor combined train detection system, characterized in that: The system comprises: One or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the system to execute the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a system, the system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a system, the system is caused to execute the method according to any one of claims 1 to 7.

Images