SD card anti-reverse insertion detection and self-adaptive guiding method, system and equipment
By combining non-uniform hardware sensing with multi-level intelligent algorithms, the problem of insufficient accuracy in SD card insertion direction detection in complex scenarios is solved, achieving high accuracy and stability in insertion direction recognition, which is suitable for high-frequency operating environments.
Patent Information
- Application Number
- CN202511034586.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-25
- Publication Date
- 2025-11-07
AI Technical Summary
Existing SD card insertion orientation detection methods are not accurate enough in complex usage scenarios and have difficulty accurately identifying the insertion orientation in diverse environments. In particular, the misjudgment rate is high when the card is worn, the insertion speed is fast, the angle is incorrect, or there is environmental interference, which affects the reliability of the device and the user experience.
A combination of non-uniform hardware sensing and hierarchical intelligent algorithms is adopted. The sensor nodes are non-uniformly arranged to capture the physical feature difference signals of the cards. Wavelet denoising, adaptive filtering and multi-stage signal optimization are used, combined with support vector machine and decision tree algorithms to make multi-level judgments and generate control signals to drive the device to perform card slot locking or releasing operations.
It significantly improves the accuracy and stability of SD card insertion orientation recognition, adapts to high-speed insertion and removal and angle deviation, reduces the false judgment rate, meets the needs of high-frequency operating environments, and ensures device security and user experience.
Smart Images
Figure CN120911491A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of information technology, and in particular to an SD card anti-reverse insertion detection and adaptive boot method, system and device. BACKGROUND
[0002] SD card insertion direction detection technology is crucial in the field of consumer electronics such as mobile devices and cameras, directly affecting user experience and device reliability. Correct insertion direction detection can prevent card damage, data loss and device failure, and thus becomes a key link in improving product competitiveness. However, existing detection systems have significant defects in complex usage scenarios.
[0003] Traditional methods rely on uniformly distributed sensor nodes to detect signal changes in card physical characteristics to determine direction, but such designs have weak signal differences when facing manufacturing tolerances, card wear or high-speed insertion, making it difficult to accurately distinguish between front and back, especially when the card surface is slightly deformed or the insertion angle is deviated, the misjudgment rate increases significantly. These limitations make the system less robust in diverse usage scenarios, making it difficult to meet the needs of modern devices for high precision and stability.
[0004] With the widespread deployment of SD cards in different terminals, multiple typical application scenarios have higher requirements for insertion direction recognition. For example, in consumer electronics terminal devices (such as mobile phones, tablets, digital / motion cameras, etc.), users often perform hot plug operations in small spaces and environments lacking visual guidance, which can easily result in misplug and reverse plug behavior. Due to the small size and precise interface of micro-SD, traditional solutions are prone to recognition errors due to fast insertion speed, incorrect angle or poor contact, which can damage the card slot or affect device operation. For example, bank ATMs and other self-service terminals and other public self-service devices need to support mass user card insertion behavior, and there are situations such as abnormal insertion, malicious interference or severe card wear, making it difficult for traditional solutions to guarantee the accuracy of direction determination and operation safety. For such devices, there is an urgent need for adaptive recognition capabilities and a protection mechanism that can be linked to perform locking or interception. For example, outdoor high-frequency plug or unattended devices (such as field recorders, remote sensing nodes, etc.): such devices face problems such as strong vibration, strong environmental interference and fast card wear, and the direction recognition system not only needs to have high stability, but also needs to be fault-tolerant and anti-interference, maintaining accurate determination in multiple plug-ins, fast operations.
[0005] The core challenge is how to improve the accuracy and stability of direction recognition through innovative detection mechanisms under dynamic and complex usage conditions. Irregularly spaced sensor node design is a key technical attribute to solve this problem. This design takes advantage of the differences in the physical characteristics of the front and back of the card to form a unique signal pattern through non-uniformly arranged detection points to enhance the accuracy of direction recognition. However, the irregularly spaced arrangement puts higher demands on the signal analysis capabilities of the algorithm. Different insertion speeds, angles, or card wear levels can cause fluctuations in the signal pattern, and the algorithm needs to distinguish between front and back signals stably in these changes. In addition, the need for higher process precision to manufacture irregularly spaced sensor arrays increases production costs and complexity, which further requires the system to strike a balance between accuracy and cost.
[0006] Therefore, how to design an adaptable irregularly spaced detection system that can accurately capture the signal pattern of the SD card insertion direction through non-uniform sensor nodes, maintain the stable analysis capabilities of the algorithm in diverse usage scenarios, and control manufacturing costs has become a key issue in this research. SUMMARY
[0007] In view of the limitations of existing SD card insertion direction detection methods in complex usage scenarios, the present application provides an SD card anti-reverse insertion detection and adaptive guidance method, system, and device, which systematically solves the pain points of the insufficient direction detection accuracy of traditional solutions in complex scenarios through the combination of "non-uniform hardware sensing + hierarchical intelligent algorithm + closed-loop execution", significantly improving the reliability and ease of use of SD cards.
[0008] In a first aspect, the present application provides an SD card anti-reverse insertion detection and adaptive guidance method, which comprises: S1. During the insertion of an SD card, capture the original signal data of the physical feature differences of the card through non-uniformly arranged sensor nodes to generate an initial signal set; S2. Preprocess the initial signal set using a denoising algorithm to generate an optimized signal set; S3. Based on the optimized signal set, use a pattern matching algorithm to compare with pre-established front and back signal templates to obtain a preliminary judgment result of the insertion direction and its confidence value; S4. Determine whether the confidence value of the preliminary judgment result is lower than a first preset threshold, if not, take the preliminary judgment result as the insertion direction determination, if yes, on the basis of the optimized signal set, use a feature smoothing algorithm to correct the time and speed fluctuations during insertion to generate a corrected signal set; S5. Based on the corrected signal set, extract feature parameters related to the insertion angle and speed, use a support vector machine algorithm for classification to obtain a secondary judgment result of the insertion direction and its confidence value; S6, judge whether the confidence value of the secondary judgment result is lower than the first preset threshold value, if not, the secondary judgment result is taken as the insertion judgment direction, if yes, based on the characteristic parameters of the correction signal set, the signal strength distribution characteristics between the sensor nodes are analyzed to generate an enhanced feature set; S7, based on the enhanced feature set, a decision tree algorithm is used for final direction classification to determine the insertion judgment direction of the SD card; S8, according to the insertion judgment direction, a control signal is generated to drive the equipment to perform the card slot locking or releasing operation, and the insertion direction detection process is completed.
[0009] In a second aspect, the application also provides an SD card anti-reverse insertion detection and adaptive guidance system, comprising a signal acquisition module, a data processing module, a preliminary judgment module, a signal correction module, a secondary judgment module, a signal enhancement module, a direction determination module and an operation control module, which are used to realize the above-mentioned SD card anti-reverse insertion detection and adaptive guidance method.
[0010] In a third aspect, the application also provides an SD card anti-reverse insertion detection and adaptive guidance device, comprising a memory and at least one processor, and the memory stores instructions; the at least one processor calls the instructions in the memory, so that the SD card anti-reverse insertion detection and adaptive guidance device executes the SD card anti-reverse insertion detection and adaptive guidance method in any one of claims 1-8.
[0011] The technical scheme provided by the embodiment of the application has the following beneficial effects: 1. By using irregularly spaced multiple sensor points in the card slot, combined with multi-stage signal optimization methods such as wavelet denoising, adaptive filtering and mean smoothing, the distinguishability of the front and back signals of the SD card is significantly enhanced, the interference caused by card wear, insertion angle deviation or speed fluctuation is effectively resisted, and the accuracy of direction recognition is improved.
[0012] 2. By establishing a confidence threshold judgment mechanism, when the preliminary judgment is insufficient, secondary / tertiary enhancement processing such as feature correction and SVM classification, PCA clustering and decision tree fusion judgment is automatically triggered, so that the insertion direction can still be stably judged even in dynamic scenes such as high-speed insertion or angle deviation insertion, the adaptability of the system is improved, and the system is suitable for high-frequency operation environments such as ATM, self-service terminal and outdoor recorder.
[0013] 3. A progressive discrimination algorithm chain is used, and the confidence of each stage output is evaluated and dynamically triggered, so that in the low signal-to-noise ratio or boundary state, the system can adaptively deepen the feature extraction and classification granularity, improve the judgment efficiency and accuracy, avoid misjudgment or mistriggering behavior, and enhance the flexibility and fault tolerance of the overall system.
[0014] 4. Based on the final orientation determination result, the system can generate control commands in real time to drive the card slot to lock / release, realize physical anti-misinsertion protection, effectively avoid problems such as abnormal contact of the card pin and electrical damage, improve equipment safety, and meet the "plug and play" operation experience requirements of consumer electronic devices. Attached Figure Description
[0015] Figure 1 This is a flowchart of an SD card anti-reverse insertion detection and adaptive boot method according to the present invention; Figure 2 Comparison chart of SD card insertion orientation recognition accuracy under different algorithm configurations; Figure 3 This is a schematic diagram of the structure of an SD card anti-insertion reverse detection and adaptive boot system according to the present invention. Detailed Implementation
[0016] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.
[0017] like Figure 1 As shown, this embodiment of an SD card anti-reverse insertion detection and adaptive boot method may specifically include: S1. During the SD card insertion process, the original signal data of the physical feature differences of the card are captured by the non-uniformly arranged sensor nodes to generate an initial signal set.
[0018] In one specific embodiment, the process of performing step S1 may specifically include the following steps: (1) The initial signal set is denoised using a preset signal filtering method to obtain a purified signal set; (2) If the signal strength of any node in the purification signal set is lower than the second preset threshold, then the signal is enhanced to obtain an enhanced signal set; (3) Based on the enhanced signal set, the support vector machine algorithm is used to classify the signal features and determine the category distribution of the physical features of the card; (4) Compare the category distribution results with the preset card feature library to obtain the card identification information with the highest matching degree; (5) Extract key feature parameters of card recognition information, and combine them with time series analysis to track the signal change trend during the insertion process and obtain the dynamic features of the operation process; (6) Compare the dynamic features with the preset operation process template. If the deviation exceeds the preset deviation range, it is marked as an abnormal operation and entry into S2 is prohibited.
[0019] Specifically, during the SD card insertion process, a plurality of sensor nodes arranged at non-uniform intervals collect multi-modal original signal data such as force, displacement, voltage, and capacitance during the contact between the card and the slot, forming an initial signal set.
[0020] Due to physical environmental interference and equipment manufacturing tolerances, some original signals have high-frequency noise or distortion. A preset filtering method is used to perform time-domain or frequency-domain denoising processing on the original signals to obtain a purified signal set. If the signal strength after purification does not reach a distinguishable threshold, a gain control or signal reconstruction method is used to enhance the feature quantity to generate an enhanced signal set. The enhanced signal set reflects the nonlinear response relationship of local feature points in the card insertion trajectory under multiple nodes. A support vector machine algorithm is used to classify the multi-dimensional feature vectors in the set and output the physical category distribution to which the signal belongs. The category distribution result is matched with a preset card feature library to calculate the optimal proximity in the feature space to identify the initial attribution label of the front and back directions of the card. On this basis, time sequence feature parameters such as insertion force change rate and sliding path deviation corresponding to the label are extracted, and a time series modeling method is used to model the dynamic change trend during the insertion process to generate dynamic features reflecting the operation continuity and stability. The dynamic features are compared with a preset operation template. If a deviation is found that exceeds the allowed range, it is identified as an abnormal operation state such as oblique insertion or rapid insertion, and the subsequent direction discrimination process is prohibited to avoid false positives.
[0021] The above algorithm features and sensor structure features are functionally integrated and coordinated. The algorithm relies on the asymmetry of the signals generated by the non-uniformly distributed structure to complete multi-dimensional feature extraction and precision enhancement. The structure design provides high-discrimination data input for the algorithm. Together, they solve the technical problem of low direction recognition accuracy in traditional solutions in weak signal and dynamic insertion situations, and are suitable for use in scenarios such as ATMs and self-service devices where frequent hot plugging and uncontrolled operation occur, thereby improving the robustness and false positive prevention and control capabilities of the overall system.
[0022] S2, using a denoising algorithm to preprocess the initial signal set to generate an optimized signal set.
[0023] In a specific embodiment, the process of performing step S2 can specifically include the following steps: (1) wavelet decomposition of the initial signal set to extract low-frequency components and high-frequency components; (2) determining whether the amplitude of the high-frequency component exceeds a third preset threshold. If so, a wavelet denoising algorithm is used to filter out high-frequency interference to obtain a signal component after filtering out interference; (3) using a mean filter algorithm to smooth all signal sets that have been denoised or not denoised to obtain a smoothed signal set; (4) performing feature extraction on the smoothed signal set to obtain time domain and frequency domain features of the signal, and obtaining a feature enhanced signal set; (5) if a frequency shift caused by the insertion speed is detected, an adaptive filtering algorithm is used to correct the feature enhanced signal set to obtain an optimized signal set; (6) performing data integrity check on the optimized signal set to determine whether the signal quality meets the preset quality standard, and if so, generating the optimized signal set.
[0024] Specifically, for the initial signal set collected by the sensor node, a wavelet decomposition method is used to perform multi-scale decomposition on the signal, and low-frequency components and high-frequency components are extracted for analyzing the signal energy distribution. If the amplitude of the high-frequency component exceeds the set interference threshold, it is determined that there is non-target factor interference such as high-speed insertion, mechanical jitter or electromagnetic disturbance, and a wavelet denoising algorithm is used to filter out the signal. After removing the high-frequency noise, the main features of the signal are retained to improve the data stability. The mean filter algorithm is used to perform time series smoothing on the denoised or non-denoised signal set to eliminate local jitter caused by the difference in the microstructure of the card or the wear of the slot, and a smooth signal set with continuous characteristics is obtained. Further feature extraction is performed on the smooth signal set to construct time domain parameters (such as amplitude mean, duration) and frequency domain parameters (such as main frequency, frequency shift) reflecting the change law of the insertion process, and a feature enhanced signal set for model training is generated. If there is a frequency shift trend in the features, it indicates that the insertion speed and attitude have changed, and an adaptive filtering algorithm is used to correct the feature signal, dynamically adjusting the filtering parameters to compensate for the feature shift and ensure the consistency of the data. The integrity check is performed on the corrected signal set to determine whether the data continuity, amplitude stability and noise suppression effect meet the preset standard, and if so, the optimized signal set is generated for subsequent direction discrimination processing.
[0025] S3, based on the optimized signal set, a pattern matching algorithm is used to compare with the pre-established front and back signal templates to obtain a preliminary judgment result of the insertion direction and its confidence value.
[0026] In a specific embodiment, the process of step S3 can specifically include the following steps: (1) using a pattern matching algorithm, the optimized signal set is compared with the preset front and back signal templates respectively to obtain a first matching result and its similarity, and a second matching result and its similarity; (2) based on the similarity comparison result, the insertion direction is preliminarily judged, and the corresponding confidence value is obtained; (3) if the corresponding confidence value is lower than the fourth preset threshold, the optimized signal set is optimized again, and the comparison is re-performed to update the judgment result; (4) If the updated confidence value is still lower than the fourth preset threshold, the secondary optimized signal set is processed through a backup signal enhancement mechanism to obtain a preliminary judgment result and its confidence value.
[0027] Specifically, by applying a pattern matching algorithm to the optimized signal set, the signals are compared with pre-established front and back signal templates, and the similarity between the signals is used to preliminarily determine the insertion direction of the SD card. The optimized signal set reflects the data characteristics after signal denoising, smoothing and enhancement, which play a crucial role in the template matching process. First, the pattern matching algorithm compares the optimized signal set with the front and back templates respectively, calculates the similarity value of each comparison, and obtains the first matching result and its similarity, the second matching result and its similarity. Through these similarity values, the system can determine the preliminary result of the insertion direction, determine whether it is front insertion or back insertion based on the template with higher similarity, and generate a preliminary confidence value for the result.
[0028] If the confidence value is lower than the preset threshold, it indicates that the preliminary judgment is not accurate enough, and the optimized signal set is subjected to secondary optimization. Secondary optimization can include adjusting the filtering parameters of the signals, further smoothing or enhancing the key features in the signals, to improve the clarity and stability of the signals. After secondary optimization, template comparison is performed again to update the direction judgment result. If the confidence value after secondary optimization is still lower than the threshold, it indicates that the quality or pattern of the signal is not clear, at this time, the backup signal enhancement mechanism is activated, and by further enhancing the key features of the signal, such as through a high-order feature extraction algorithm, increasing the resolution of the signal, or using other algorithms such as signal fusion, more clear and accurate insertion direction data is obtained. The enhanced signal set is compared with the front and back templates again, and finally the insertion direction judgment result and its corresponding confidence value are obtained.
[0029] In a specific embodiment, the technical solution of the present application can be applied to an automatic teller machine (ATM) or a self-service terminal device for identifying the insertion direction of an SD card inserted by a user under non-visual assistance. Six non-uniformly spaced sensor nodes are arranged in a card slot, the sensors including pressure sensors, vibration sensors, and electrical contact detection units, the sampling frequency of each node being set to 2000 Hz, and the original signal data being continuously collected for 0.5 seconds during the insertion of the SD card to form an initial signal set containing 3000 samples. Wavelet decomposition is performed on the initial signal set to separate the high-frequency component and the low-frequency component. If the amplitude of the high-frequency component of the signal collected by the node exceeds ±0.3 units (such as 0.6N), the wavelet denoising algorithm is used to filter out the high-frequency interference to generate an optimized signal set after filtering out the interference. The pattern matching algorithm is applied to the optimized signal set to compare it with the preset front insertion signal template and the preset back insertion signal template in the system, respectively, to calculate the similarity. For example, in a typical scenario, the similarity of the optimized signal set with the front template and the back template is calculated, for example, the similarity with the front template is 0.78, and the similarity with the back template is 0.52, according to the similarity comparison result, the system preliminarily determines that the insertion direction is the front, and the confidence is 0.78. If the confidence value is lower than the preset determination threshold (such as 0.8), a secondary optimization processing procedure is triggered, the signal gain parameter is dynamically adjusted, the signal time sequence offset is adjusted in combination with the adaptive insertion speed correction module, an updated optimized signal set is formed, and the pattern matching is re-executed. After the update and comparison, the similarity is improved to 0.85, the confidence meets the determination standard, and the system outputs a control signal accordingly to drive the card slot locking mechanism into a card reading preparation state.
[0030] The test results show that the technical solution can still maintain an identification accuracy of not less than 97.5% within an insertion angle deviation of ±22° and an insertion speed variation range of 0.5 cm / s to 5 cm / s, the average judgment time is about 420 milliseconds, which is significantly better than the traditional solution, and has strong stability, real-time performance, and anti-interference ability.
[0031] S4, determining whether the confidence value of the preliminary determination result is lower than a first preset threshold, if not, taking the preliminary determination result as the insertion determination direction, and if yes, correcting the time and speed fluctuations in the insertion process by using a feature smoothing algorithm on the basis of the optimized signal set to generate a corrected signal set.
[0032] In a specific embodiment, the process of performing step S4 can specifically include the following steps: (1) When the confidence value of the preliminary determination is lower than the first preset threshold, extracting the time feature and the speed feature from the original signal data to construct a feature data set; (2) using a moving average algorithm to smooth the feature data set to generate a smoothed feature set, judging whether the time fluctuation or speed fluctuation in the smoothed feature set exceeds the preset fluctuation range, if so, using Kalman filter algorithm for correction; (3) using linear interpolation method to fill in the missing data of the corrected or uncorrected feature data to generate a complete signal set, then calculating and judging whether the time consistency and speed stability of each signal point meet the preset standard, if not, performing secondary smoothing processing on the complete signal set to obtain a corrected signal set, if yes, taking the complete signal set as the corrected signal set.
[0033] Specifically, when the confidence value obtained by the preliminary judgment of the insertion direction based on the optimized signal set is lower than the preset threshold, the time and speed features in the insertion process are extracted from the original signal data corresponding to the optimized signal set to construct a feature data set for fluctuation correction. For the feature data set, a moving average algorithm is used to perform time series smoothing to weaken the local fluctuations caused by uneven user insertion rate, unstable posture or slot contact delay, obtain a smoothed feature set, and statistically judge the time continuity and speed change range of each data segment in the set. If the fluctuation amplitude is detected to exceed the set range threshold, the feature set is dynamically estimated and state corrected by Kalman filter algorithm to improve the signal stability and trend consistency of the insertion process. If there are discontinuities or missing samples in the feature data after Kalman filtering, the system further uses linear interpolation method to fill in the missing data to form a continuous complete signal set. On this basis, the time interval consistency and speed change stability of each signal point in the complete set are evaluated, if it does not meet the signal quality requirements, the complete signal set is subjected to secondary smoothing processing to obtain a corrected signal set; if the conditions are met, the current complete signal set is directly taken as the corrected signal set output.
[0034] This technical feature combines sensor sampling structure with time series-based filtering optimization algorithm, which can dynamically adjust the time distribution and physical amplitude of the signal when the user's card insertion action has a dramatic change in speed or posture deviation, solving the problem of recognition error caused by signal fluctuation due to high-frequency insertion or oblique insertion operation. In ATM or self-service terminal scenarios, the user's behavior of inserting a bank card is uncertain and diverse, and traditional signal processing cannot effectively distinguish between incidental disturbances and real features; this embodiment uses a joint correction mechanism in the time and speed dimensions to construct a corrected signal input with stable sequence characteristics, providing a quality controllable signal basis for subsequent direction recognition algorithms, and improving the robustness and direction determination accuracy of the system under complex use conditions.
[0035] In a specific embodiment, the step S4 described in the present application is applicable to typical application scenarios such as ATM or self-service terminal, when the user inserts the SD card at a non-constant speed and the insertion angle has a deviation, the confidence value obtained after the preliminary direction judgment is 0.62, which is lower than the first preset confidence threshold 0.75, triggering the feature correction process. The system extracts the time feature and speed feature in the insertion process based on the original sampling data corresponding to the optimized signal set, and constructs a feature data set, wherein the speed sequence shows that there is a high-speed change section of 5.5 to 7.8 cm / s in about 0.4 seconds, and there is a sharp fluctuation. The moving average algorithm is used for sliding smoothing processing on the feature data set, and the sliding window width is set to 5 sampling points. The speed change interval is compressed to the range of 2.1 to 3.9 cm / s after smoothing. The feature set after smoothing still has a speed jump that exceeds the system stability range, and the system further uses the Kalman filter algorithm to jointly correct the time and speed state variables, suppresses the abnormal mutation section, and adjusts the speed sequence to be continuous and smooth. For the signal discontinuity in the sampling process, the system uses the linear interpolation method to complete the speed data corresponding to the missing time point to form a complete signal set. The complete set is evaluated for time consistency and speed stability, and the time interval standard deviation is less than 1.5 ms and the speed change rate standard deviation is less than 0.6 cm / s, which meets the system's preset data integrity standard. The signal set generated after the above processing is the corrected signal set, which is used as the input data for subsequent classification and discrimination. The application results show that the correction processing improves the original confidence value from 0.62 to 0.84, improving the reliability of the insertion direction recognition. The feature smoothing and correction mechanism and the previous optimized signal collection process form a linked structure in function, which can effectively resist the non-stable data input caused by user's fast insertion, posture deviation or signal interference, and ensure the direction judgment accuracy and use safety of the system in the typical unassisted card insertion environment.
[0036] S5, based on the corrected signal set, extracting feature parameters related to the insertion angle and speed, using a support vector machine algorithm for classification to obtain a secondary judgment result of the insertion direction and a confidence value thereof.
[0037] In a specific embodiment, the process of executing step S5 can specifically include the following steps: (1) constructing a feature vector set based on the feature parameters related to the insertion angle and speed; (2) classifying and training the feature vector set by a support vector machine algorithm to obtain a preliminary insertion direction result and a confidence value thereof; (3) if the confidence value of the preliminary insertion direction result is lower than the fifth preset threshold, re-extracting angle features and speed features from the corrected signal set to generate a supplementary feature parameter group; (4) Update the feature vector set based on the supplementary feature parameter group, and re-classify and predict to make a second judgment of the insertion direction; (5) Sort the two insertion direction judgment results according to the confidence values to determine the second judgment result and the confidence value thereof.
[0038] Specifically, after the system completes the generation of the correction signal set, feature parameters related to the insertion angle and speed are extracted based on the correction signal set, which are used to construct a feature vector set of a support vector machine classification model. The angle feature can be obtained by analyzing the response amplitude change trend of each sensor node in the insertion process and the relative response time difference between nodes, and the speed feature can be derived from the rate of change of signal amplitude over time. Both of them reflect the directionality behavior and dynamic process of card insertion. After the feature vector set is constructed, it is input into the trained support vector machine classifier for preliminary classification and prediction, and the insertion direction judgment result and the corresponding confidence value are output. If the confidence value is lower than the fifth threshold value (such as 0.80) set by the system, it indicates that the current features are not sufficient to support high-confidence classification. The system will perform local detailed analysis on the original correction signal set to extract supplementary parameters such as higher time density angle change rate, insertion path offset amplitude, speed fluctuation period, etc. to construct a supplementary feature parameter group, which is used to update the feature vector set. The updated feature vector set is input into the support vector machine for classification and prediction again to obtain the second direction judgment result and the confidence value, and the confidence values of the two results are sorted to output the second judgment result with higher confidence as the final insertion direction.
[0039] In a specific embodiment, in an ATM or unattended card issuing terminal, the user's card insertion process often exhibits a short-time high-speed insertion feature, and the card insertion posture is affected by hand movements, resulting in direction tilt or angle mutation. When the system initially judges the direction, the confidence value may be low due to signal instability caused by angle changes, which may easily lead to misjudgment. By using the method described in this step, the support vector machine is used to jointly classify multiple fused feature dimensions, and the feature dimensions are dynamically updated in the case of initial classification uncertainty to improve the classification reliability. The algorithm can stably output the direction judgment result in the scenario of card insertion speed interval 2-5 cm / s and angle deviation ±20°. The features of the algorithm and the function of the non-uniform sensor structure are mutually dependent. The sensor provides multi-point asynchronous high-resolution sampling signals to capture the angle / speed differences of the card insertion behavior, and the support vector machine performs model training and discrimination based on the structure of the collected features. Both of them solve the problems of signal ambiguity and unstable classification caused by dynamic card insertion in the prior art. Actual tests show that the direction recognition confidence can be improved by more than 15% on average under this processing path, the misjudgment rate is controlled within 2%, and it has good adaptability and scalability, meeting the actual deployment needs of high-frequency fast operation scenarios.
[0040] S6, judging whether the confidence value of the secondary judgment result is lower than the first preset threshold value, if not, taking the secondary judgment result as the insertion judgment direction, if yes, based on the characteristic parameters of the correction signal set, analyzing the signal strength distribution characteristics between the sensor nodes, and generating an enhanced characteristic set.
[0041] In a specific embodiment, the process of performing step S6 can specifically include the following steps: (1) using principal component analysis algorithm to extract the main characteristic vector of the signal strength distribution characteristics, and generating a first characteristic set; (2) if the dimension of the first characteristic set is higher than the preset dimension, performing dimension reduction processing, calculating the covariance matrix of the characteristic vector after dimension reduction, and obtaining a second characteristic set; (3) based on the first or second characteristic set, clustering analyzing the relationship between the sensor nodes through K-means algorithm, judging whether the clustering number is less than the preset number, if yes, generating an enhanced characteristic set through weighted fusion method combined with the signal distribution characteristics, if not, taking the first or second characteristic set as the enhanced characteristic set.
[0042] Specifically, when the secondary judgment result confidence value of the insertion direction is still lower than the first preset threshold value set by the system, based on the response data of each sensor node contained in the correction signal set, the signal strength distribution parameters representing the spatial characteristics of the insertion process are extracted to generate an enhanced characteristic set. Due to the uneven arrangement of the sensor nodes, the response signals exhibit distribution asymmetry under different directions or angles of card insertion. The system utilizes this distribution characteristic to construct a characteristic vector matrix, uses principal component analysis algorithm to extract the features of the matrix, identifies the dominant factors affecting the signal distribution difference through covariance decomposition, extracts a group of principal component vectors to form a first characteristic set. If the dimension of the set exceeds the characteristic capacity threshold set by the system, dimension reduction processing is performed, the main component information is preserved through linear transformation, and a second characteristic set after dimension reduction is generated. The system performs clustering analysis on the first or second characteristic set, divides the characteristic vectors corresponding to all sensor nodes into several classes using K-means algorithm, calculates the average distance between the class centers and the samples to evaluate the response relationship between the nodes, and judges whether the clustering number is less than the preset standard according to the actual clustering result, if yes, it is considered that the sensor response has obvious distribution concentration, which indicates that there is a distinguishable mode in the spatial characteristics of the current card insertion behavior, and the system reconstructs each characteristic category according to the distribution density and weight through weighted fusion strategy, and generates an enhanced characteristic set combined with the signal distribution characteristics; if the clustering number reaches or exceeds the preset standard, it indicates that the difference between the node responses is insufficient, which does not meet the fusion condition, and the first or second characteristic set is directly taken as the enhanced characteristic set for subsequent processing.
[0043] In a specific embodiment, for uncertain operation situations such as fast card insertion speed, posture skew, etc. in the self-service terminal, when the confidence value obtained after performing secondary judgment of the insertion direction is still lower than the first preset threshold set by the system, the system constructs a distribution feature vector matrix representing the difference in card insertion direction based on the response intensity data of the multiple sensor nodes corresponding to the correction signal set, and improves the subsequent recognition accuracy through enhanced modeling. Taking six non-uniformly arranged sensor nodes in a certain government self-service terminal as an example, the normalized average intensity value, response peak value and response position (in terms of total card insertion time percentage) of each node are collected during the user card insertion process to form an original feature matrix. The system applies principal component analysis algorithm to the matrix to extract main signal distribution characteristic factors, and retains the first two dimensions of the feature set whose cumulative variance contribution rate exceeds 90% to form a first feature set. When the dimension of the extracted feature set is still higher than the system's acceptable processing dimension threshold, the system compresses the feature space through linear dimension reduction to form a second feature set after dimension reduction.
[0044] On this basis, the system uses K-means clustering algorithm to cluster and divide the first or second feature set, identifies the response similarity between the sensor nodes, and judges whether the number of clusters is less than the set maximum threshold (such as k=4). In an example, the system divides the six nodes into three clusters, respectively corresponding to the front, middle and end characteristic response regions of card insertion, wherein the response intensity and timing position of the cluster center have significant differences. The system assigns weight values to the nodes in the cluster according to the center intensity, node distribution density and other factors, and generates an enhanced feature set expressing the response characteristics of the card insertion space by weighted fusion of the node responses. If the number of clusters has reached the set threshold, it means that the signal distribution characteristics have no obvious centralized features, and the first or second feature set is directly used as the enhanced feature set.
[0045] The test results show that in the extreme scenario of user card insertion angle deviation ±25° and insufficient insertion time of 0.4 seconds, the secondary judgment confidence is only 0.66, and after the enhanced feature set generated by this step participates in the subsequent recognition process, the confidence of the system's card insertion direction judgment is improved to 0.86, and the recognition accuracy is improved from 87.1% to 96.8%. The enhanced feature generation mechanism cooperates with the data structure established based on the physical distribution response of the sensor and the clustering algorithm to realize the function of supporting the correlation of the previous signal correction technology features, and is used to improve the stability and reliability of the direction recognition model under complex dynamic operation conditions, which has good engineering application value and practical promotion significance.
[0046] S7, based on the enhanced feature set, using a decision tree algorithm for final direction classification to determine the insertion direction of the SD card.
[0047] In a specific embodiment, the process of performing step S7 can specifically include the following steps: (1) The enhanced feature set is trained using a decision tree algorithm to construct a direction classification model and complete the insertion direction judgment, and obtain the classification result judgment and its confidence value; (2) If the confidence value of the classification result judgment is lower than the sixth preset threshold, the enhanced feature set is classified again using a support vector machine algorithm to obtain an optimized classification result judgment; (3) The insertion direction is determined according to the optimized classification result judgment.
[0048] In a specific embodiment, when the generation of the enhanced feature set is completed, a direction classification model is constructed based on the enhanced feature set using a decision tree algorithm, and the final judgment of the insertion direction is performed. In this process, the enhanced feature set includes high-confidence feature vectors obtained through multi-stage filtering, signal distribution modeling, and feature aggregation processing, covering multi-dimensional parameters reflecting insertion angle, speed change, sensor spatial response mode, etc. The system uses this feature set as an input sample to train the existing direction labeled data to form a direction classification model with a path branching structure. The model automatically selects the most discriminative feature dimension for node splitting based on the information gain criterion, so that the judgment process is optimized layer by layer among multiple signal dimensions. After the classification model is trained, the system judges the current enhanced feature set and outputs the direction classification result and the corresponding confidence value.
[0049] If the confidence value corresponding to the classification result is still lower than the sixth preset threshold, for example, less than 0.80, the system further calls a support vector machine algorithm to perform secondary classification processing on the same enhanced feature set. In this processing path, the system uses the aforementioned feature set to construct a support vector space, divides the classification boundary through the maximum interval principle, and then obtains the optimized judgment result of the insertion direction and outputs the corresponding confidence value. The system compares the confidence indicators of the decision tree and the support vector machine under the current sample, and takes the direction result corresponding to the higher confidence as the final judgment result of the insertion direction.
[0050] In this technical solution, the decision tree model has the advantages of clear structure, clear logic branching, and strong real-time performance, and can quickly output reliable classification results when the feature dimension structure is clear and significantly different. The support vector machine has higher fault tolerance and discrimination accuracy for complex sample distribution scenarios with ambiguous boundaries, and the two form a complementary relationship in the judgment process. Through this structure, when processing complex dynamic card insertion behaviors in high-frequency use scenarios such as self-service terminals and ATMs, even if the confidence output by the previous recognition stage still has marginal deficiencies, the final high-confidence direction output can be achieved through this dual-path fusion structure, significantly enhancing the intelligence and robustness of the system.
[0051] In a specific embodiment, to solve the problems of user card insertion angle deviation and uneven node response in government self-service terminals, when the system completes the generation of the enhanced feature set, the final judgment operation of the insertion direction is performed. The enhanced feature set includes average response intensity mean, response position weighted center, sensor maximum response difference, principal component analysis score, cluster gravity deviation, and other multi-dimensional feature parameters, which have good insertion state representation ability. The system inputs the enhanced feature set into the decision tree classification model constructed based on the CART algorithm, and judges the direction of the current sample through the node splitting path. In this example, the first principal component score is 0.49, the maximum response difference is 0.31, and the response center position is 0.58, all of which meet the "positive insertion" path conditions, and the system outputs the judgment result as "positive", with a confidence value of 0.76, which is lower than the sixth preset threshold value 0.80 set by the system.
[0052] To further improve the judgment accuracy, the system triggers the backup support vector machine algorithm for secondary classification processing. The support vector machine uses a radial basis kernel function, constructs a classification boundary based on the input features, and calculates the relative position of the sample, outputs the "positive" judgment result, and calculates the confidence value based on the sample distance as 0.89, which is higher than the decision tree result. The system takes the SVM judgment result with higher confidence as the final classification result of the insertion direction, and controls the module to execute the insertion direction feedback instruction.
[0053] In the above processing flow, the decision tree model quickly outputs the direction conclusion based on feature rules, which is suitable for scenarios with clear response feature distribution; the support vector machine model is suitable for samples with fuzzy boundaries and feature intersections, and the comprehensive judgment ability of the system is improved through joint configuration. This method relies on the data structure expression ability of the pre-enhanced feature set, cooperates with the model switching mechanism to form a composite discrimination path, improves the adaptability and robustness of the direction judgment under dynamic operation, and has good engineering realizability and terminal deployment value. The test shows that this processing method can still maintain a confidence value greater than 0.85 and a classification accuracy of more than 97% in the scenario of card insertion angle deviation ± 15° and insertion time less than 0.6 seconds.
[0054] S8, according to the final result of the insertion direction, generate a control signal to drive the device to perform card slot locking or releasing operation, complete the insertion direction detection process.
[0055] In a specific embodiment, the process of performing step S8 can specifically include the following steps: (1) Collect execution state feedback data and judge whether the operation is successful.
[0056] (2) Record complete logs, analyze process efficiency, optimize control parameters according to the analysis results, and update response logic in a closed loop.
[0057] Specifically, after completing the insertion direction judgment and obtaining the final classification result, the system generates a control signal according to the direction result, drives the electric control actuator to complete the card slot locking or releasing action through the control module, wherein the control signal type is bound to the judgment result, if the direction judgment result is positive, the release instruction is sent to allow the card to continue to be inserted and establish the data channel; if the judgment result is negative, the locking instruction is sent to close the electromagnetic mechanism and trigger the visual or audible warning prompt. After the system outputs the control signal, it collects the execution state data through the feedback channel, including mechanical position confirmation signal, motor response current, electromagnetic execution response time, etc., and judges whether the operation is successfully completed within the set time according to the feedback data, if there is a response delay, action deviation or abnormality such as not closed, it is marked as execution failure and recorded.
[0058] All control processes and response information are written into the log system in real time, and the log content includes insertion timestamp, direction judgment path, confidence index at each algorithm stage, control instruction sending time, execution feedback result and action time consumption, etc. The system performs periodic statistics and process efficiency analysis based on the log data, automatically adjusts the control parameters such as response delay threshold, current trigger value and judgment confidence lower limit by extracting indicators such as average execution delay, misjudgment rate and card slot mislocking times, so as to realize the closed-loop adaptive update of the control logic. This technical means realizes the functional linkage between the direction judgment module and the physical response mechanism, ensures that the whole process of insertion operation forms a data closed loop among discrimination, control and feedback, which can effectively improve the response accuracy and equipment operation safety in complex scenarios, and is suitable for application scenarios such as ATM, self-service terminal and other scenarios that need to ensure the safe operation of equipment and the reliability of user interaction.
[0059] In an embodiment, in order to verify the recognition performance of the "support vector machine + decision tree + enhanced feature fusion algorithm model" proposed in the present application in complex card insertion scenarios, four different algorithm configurations are constructed for comparison experiment group, which are support vector machine model, decision tree model, fusion model (combination of SVM and decision tree) and fusion model with enhanced feature processing. The comparative experiment is carried out in the simulation of ATM terminal card insertion scenario, 1000 groups of sample data with different insertion speed and angle disturbance are used for training and testing, and the accuracy of each model in direction recognition task is counted.
[0060] As shown in Figure 2 , the accuracy of the support vector machine model is 84.3%, the accuracy of the decision tree model is 82.7%, and the accuracy of the fusion model is 91.5% after the fusion of the two. On this basis, the enhanced feature set based on signal strength distribution is introduced, and after further optimizing the feature discrimination ability, the recognition accuracy reaches 96.2%. The results verify that the fusion algorithm model has excellent discrimination performance, can adapt to complex scenarios such as high-speed insertion and insertion angle deviation, and significantly improves the direction recognition accuracy and system stability.
[0061] The above describes the SD card anti-reverse insertion detection and adaptive boot method in the embodiments of the present application. The following describes an SD card anti-reverse insertion detection and adaptive boot system in the embodiments of the present application. Please refer to Figure 3 An embodiment of the SD card anti-reverse insertion detection and adaptive boot system in the embodiments of the present application includes a signal acquisition module 10, a data processing module 20, a preliminary judgment module 30, a signal correction module 40, a secondary judgment module 50, a signal enhancement module 60, a direction determination module 70, and an operation control module 80.
[0062] The signal acquisition module 10 is configured to capture the original signal data of the physical feature difference of the card during the insertion of the SD card by the non-uniformly arranged sensor nodes, and generate an initial signal set.
[0063] The data processing module 20 is configured to pre-process the initial signal set by using a denoising algorithm, and generate an optimized signal set.
[0064] The preliminary judgment module 30 is configured to compare the optimized signal set with the pre-established front and back signal templates by using a pattern matching algorithm, and obtain a preliminary judgment result of the insertion direction and a confidence value thereof.
[0065] The signal correction module 40 is configured to judge whether the confidence value of the preliminary judgment result is lower than a first preset threshold. If not, the preliminary judgment result is taken as the insertion determination direction. If yes, the time and speed fluctuations during the insertion are corrected by using a feature smoothing algorithm on the basis of the optimized signal set, and a corrected signal set is generated.
[0066] The secondary judgment module 50 is configured to extract the feature parameters related to the insertion angle and speed from the corrected signal set, classify by using a support vector machine algorithm, and obtain a secondary judgment result of the insertion direction and a confidence value thereof.
[0067] The signal enhancement module 60 is configured to judge whether the confidence value of the secondary judgment result is lower than the first preset threshold. If not, the secondary judgment result is taken as the insertion determination direction. If yes, the signal strength distribution characteristics among the sensor nodes are analyzed based on the feature parameters of the corrected signal set, and an enhanced feature set is generated.
[0068] The direction determination module 70 is configured to classify the final direction by using a decision tree algorithm based on the enhanced feature set, and determine the insertion determination direction of the SD card.
[0069] The operation control module 80 is configured to generate a control signal based on the insertion determination direction, drive the equipment to perform the card slot locking or releasing operation, and complete the insertion direction detection process.
[0070] The application further provides an SD card anti-reverse insertion detection and adaptive boot device, which comprises a memory and a processor, and the memory stores computer readable instructions, and the computer readable instructions are executed by the processor to make the processor execute the steps of the SD card anti-reverse insertion detection and adaptive boot method in any of the above embodiments.
[0071] The above description and the above embodiments are only used to illustrate the technical solutions of the application, rather than limit the application; although the application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and the modification or replacement does not make the essence of the corresponding technical solution deviate from the spirit and scope of the technical solutions of the embodiments.
Claims
1. A method for SD card anti-reverse insertion detection and adaptive booting, characterized in that, The method comprises: S1, in the SD card insertion process, the original signal data of the difference of the card physical characteristics is captured by the non-uniformly arranged sensor nodes to generate an initial signal set; S2, the initial signal set is preprocessed by using a denoising algorithm to generate an optimized signal set; S3, based on the optimized signal set, a pattern matching algorithm is used to compare with the pre-established front and back signal templates to obtain the preliminary judgment result of the insertion direction and its confidence value; S4, whether the confidence value of the preliminary judgment result is lower than the first preset threshold is judged, if not, the preliminary judgment result is taken as the insertion judgment direction, if yes, the time and speed fluctuations in the insertion process are corrected based on the optimized signal set by using a feature smoothing algorithm to generate a corrected signal set; S5, based on the corrected signal set, the feature parameters related to the insertion angle and speed are extracted, a support vector machine algorithm is used for classification to obtain the secondary judgment result of the insertion direction and its confidence value; S6, whether the confidence value of the secondary judgment result is lower than the first preset threshold is judged, if not, the secondary judgment result is taken as the insertion judgment direction, if yes, based on the feature parameters of the corrected signal set, the signal strength distribution characteristics between the sensor nodes are analyzed to generate an enhanced feature set; S7, based on the enhanced feature set, a decision tree algorithm is used for final direction classification to determine the insertion judgment direction of the SD card; S8, a control signal is generated according to the insertion judgment direction to drive the equipment to perform the card slot locking or releasing operation, and the insertion direction detection process is completed.
2. The method of claim 1, wherein the method further comprises: The S1 comprises: The initial signal set is denoised by using a preset signal filtering method to obtain a purified signal set; If the signal strength of any node in the purified signal set is lower than a second preset threshold, the signal is enhanced to obtain an enhanced signal set; Based on the enhanced signal set, the signal features are classified by using a support vector machine algorithm to judge the category distribution of the card physical characteristics; The category distribution result is compared with a preset card feature library to obtain the card identification information with the highest matching degree; The key feature parameters of the card identification information are extracted, and the signal change trend in the insertion process is tracked by combining a time series analysis method to obtain the dynamic characteristics of the operation process; The dynamic characteristics are compared with a preset operation process template, if the deviation exceeds a preset deviation range, it is marked as an abnormal operation to prohibit entering the S2.
3. The method of claim 1, wherein the method further comprises: The S2 comprises: The initial signal set is wavelet decomposed to extract low-frequency components and high-frequency components; Whether the high-frequency component amplitude exceeds a third preset threshold is judged, if yes, the wavelet denoising algorithm is used to filter out the high-frequency interference to obtain the signal components after filtering out the interference; The denoised or non-denoised all signal sets are smoothed by using a mean filtering algorithm to obtain the smoothed signal set; The smoothed signal set is feature extracted to obtain the time domain and frequency domain features of the signal to obtain the feature enhanced signal set; If the frequency offset caused by the insertion speed is detected, an adaptive filtering algorithm is used to correct the feature enhanced signal set, so as to obtain an optimized signal set; The optimized signal set is subjected to data integrity checking, and it is judged whether the signal quality meets the preset quality standard, if yes, the optimized signal set is generated.
4. The method of claim 1, wherein the method further comprises: determining whether the SD card is inserted in the reverse direction; and outputting a signal to the SD card controller to prevent the SD card from being recognized as the SD card if the SD card is determined to be inserted in the reverse direction. The S3 comprises: The mode matching algorithm is used to compare the optimized signal set with the preset positive signal template and negative signal template respectively, so as to obtain the first matching result and its similarity, and the second matching result and its similarity; Based on the similarity comparison result, the insertion direction is preliminarily judged, and the corresponding confidence value is obtained; If the corresponding confidence value is lower than the fourth preset threshold value, the optimized signal set is subjected to secondary optimization, and is compared again to update the judgment result; If the updated confidence value is still lower than the fourth preset threshold value, the secondary optimized signal set is processed through a backup signal enhancement mechanism, so as to obtain the preliminary judgment result and its confidence value.
5. The method of claim 1, wherein the method further comprises: determining whether the SD card is inserted in the reverse direction; and outputting a signal to the SD card controller to prevent the SD card from being recognized as the SD card if the SD card is determined to be inserted in the reverse direction. The S4 comprises: When the confidence value of the preliminary judgment is lower than the first preset threshold value, the time feature and the speed feature are extracted from the original signal data to construct a feature data set; The moving average algorithm is used to smooth the feature data set to generate a smooth feature set, and it is judged whether the time fluctuation or the speed fluctuation in the smooth feature set exceeds the preset fluctuation range, if yes, the Kalman filtering algorithm is used for correction; The linear interpolation method is used to fill in the missing data of the corrected or uncorrected feature data to generate a complete signal set, and then it is calculated and judged whether the time consistency and the speed stability of each signal point meet the preset standard, if not, the complete signal set is subjected to secondary smoothing to obtain the corrected signal set, if yes, the complete signal set is taken as the corrected signal set.
6. The method of claim 1, wherein the method further comprises: The S5 comprises: Based on the feature parameters related to the insertion angle and speed, a feature vector set is constructed; The support vector machine algorithm is used to classify and train the feature vector set and to make prediction, so as to obtain the preliminary insertion direction result and its confidence value; If the confidence value of the preliminary insertion direction result is lower than the fifth preset threshold value, the angle feature and the speed feature are extracted from the corrected signal set again to generate a supplementary feature parameter group; Based on the supplementary feature parameter group, the feature vector set is updated, and classification training and prediction are performed again to make secondary judgment of the insertion direction; Combined with the confidence value, the two insertion direction judgment results are sorted to determine the secondary judgment result and its confidence value.
7. The SD card anti-reverse insertion detection and adaptive boot method according to claim 1, characterized in that, The S6 comprises: The principal component analysis algorithm is used to extract the main feature vector of the signal intensity distribution feature to generate a first feature set; If the dimension of the first feature set is higher than the preset dimension, dimension reduction processing is performed, the covariance matrix of the dimension-reduced feature vector is calculated, and a second feature set is obtained; Based on the first or second feature set, the relationship between sensor nodes is analyzed by K-means algorithm clustering, it is judged whether the clustering number is less than the preset number, if yes, the enhanced feature set is generated by weighted fusion method combined with signal distribution characteristics, if no, the first or second feature set is taken as the enhanced feature set.
8. The SD card anti-reverse insertion detection and adaptive boot method according to claim 1, characterized in that, The S7 comprises: The enhanced feature set is trained by using the decision tree algorithm, the direction classification model is constructed, and the insertion direction judgment is completed to obtain the classification result judgment and its confidence value; If the confidence value of the classification result judgment is lower than the sixth preset threshold, the enhanced feature set is classified again by using the support vector machine algorithm to obtain the optimized classification result judgment; According to the optimized classification result judgment, the final result of the insertion direction is determined.
9. A system for SD card anti-reverse insertion detection and adaptive booting based on the method of any one of claims 1 to 8, characterized in that, It comprises a signal acquisition module, a data processing module, a preliminary judgment module, a signal correction module, a secondary judgment module, a signal enhancement module, a direction determination module and an operation control module. The signal acquisition module is used for capturing the original signal data of the physical feature difference of the card in the SD card insertion process by the non-uniformly arranged sensor nodes to generate an initial signal set; The data processing module is used for pre-processing the initial signal set by using the denoising algorithm to generate an optimized signal set; The preliminary judgment module is used for comparing the optimized signal set with the pre-established front and back signal templates by using the pattern matching algorithm to obtain the preliminary judgment result of the insertion direction and its confidence value; The signal correction module is used for correcting the time and speed fluctuation in the insertion process by using the feature smoothing algorithm based on the optimized signal set if the confidence value of the preliminary judgment result is lower than the first preset threshold to generate a corrected signal set; The secondary judgment module is used for extracting the feature parameters related to the insertion angle and speed based on the corrected signal set, classifying by using the support vector machine algorithm to obtain the secondary judgment result of the insertion direction and its confidence value; the signal enhancement module is used for generating an enhanced feature set based on the feature parameters of the corrected signal set if the confidence value of the secondary judgment result is still lower than the first preset threshold to analyze the signal intensity distribution characteristics between the sensor nodes; The direction determination module is used for finally classifying the insertion direction by using the decision tree algorithm based on the enhanced feature set to determine the final result of the insertion direction of the SD card; The operation control module is used for generating a control signal according to the final result of the insertion direction to drive the equipment to perform the card slot locking or releasing operation to complete the insertion direction detection process.
10. An SD card anti-reverse insertion detection and adaptive boot device, comprising: The SD card anti-reverse insertion detection and adaptive guiding equipment comprises a memory and at least one processor, and the memory stores instructions; The at least one processor calls the instructions in the memory to make the SD card anti-reverse insertion detection and adaptive guiding equipment execute the SD card anti-reverse insertion detection and adaptive guiding method in any one of claims 1-8.