Portable scanning device and method based on AI recognition

Abstract

This invention discloses a portable scanning device and method based on AI recognition, relating to the field of image scanning technology. The method includes: S1, periodically acquiring scanning imaging data and performing preprocessing on the data; S2, determining the imaging state based on the preprocessed data, and when the device is in a drift-triggered state, evaluating the degree of multi-source drift during the scanning imaging process and determining the drift level based on the evaluation result; S3, performing graded parameter correction according to the drift level; and S4, executing imaging control commands and continuously evaluating the degree of multi-source drift during the scanning imaging process based on the data after command execution, determining whether to repeat parameter correction, and realizing a closed-loop compensation control for scanning imaging. This solves the problem that existing devices lack online drift calibration and temperature drift compensation, leading to light source attenuation, dark current drift of the photosensitive chip, and power supply voltage fluctuations after prolonged use.

Description

Technical Field

[0001] This invention relates to the field of image scanning technology, specifically to a portable scanning device and method based on AI recognition. Background Technology

[0002] With the widespread application of portable optical scanning devices, scanner designs are gradually evolving towards lightweight, intelligent, and high-precision technologies. These devices typically acquire and digitally reconstruct images of target object surfaces through built-in light sources, photosensitive chips, and positioning modules, playing a crucial role in document recognition, surface inspection, and medical imaging. To improve scanning accuracy and imaging consistency, existing technologies generally incorporate multi-module collaborative scanning and path correction mechanisms to reduce imaging errors caused by hand shake or path deviation during scanning.

[0003] For example, invention publication number CN103702011B discloses a portable scanner and its scanning path calculation method. The portable scanner includes a scanning module, a reference module, and a path calculation component. The scanning module is capable of scanning the object in relative motion and includes first and second positioning units. The reference module is disposed at the edge of the object to be scanned and includes first and second reference units. The path calculation component is electrically connected to the first and second reference units. When the scanning module scans the object, the reference module remains relatively stationary with respect to the object. The path calculation component calculates the movement path of the scanning module based on the positional information between the first positioning unit and the first and second reference units, and the positional information between the second positioning unit and the first and second reference units, respectively.

[0004] However, although the aforementioned technologies can achieve geometric correction and pose calculation of the scanning path, thereby improving scanning accuracy, they still have shortcomings in practical applications: they mainly address the motion trajectory problem of the scanning module, failing to consider the impact of multi-source interference factors such as illumination attenuation, temperature drift, and power fluctuations on image quality during the scanning and imaging process. Existing portable scanning devices typically lack comprehensive evaluation and dynamic compensation mechanisms for optical, thermal, and electrical drift, leading to uneven image brightness, increased noise, and decreased signal stability after long-term operation, making it difficult to achieve consistently high image quality control.

[0005] Therefore, in order to address the above problems, there is an urgent need for a portable scanning device and method based on AI recognition. Summary of the Invention

[0006] Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides a portable scanning device and method based on AI recognition, which solves the problems of light source attenuation, dark current drift of the photosensitive chip, and power supply voltage fluctuations caused by the lack of online drift calibration and temperature drift compensation in existing devices after long-term use.

[0008] Technical solution

[0009] To achieve the above objectives, the present invention provides the following technical solution: a portable scanning method based on AI recognition, comprising: S1, periodically acquiring scanning imaging data, and performing time drift correction, abnormal sample removal, dynamic compensation, and standardization processing on the scanning imaging data to obtain preprocessed scanning imaging data; S2, determining the imaging state based on the preprocessed scanning imaging data, and when the device is in a drift-triggered state, evaluating the degree of multi-source drift during the scanning imaging process, and determining the drift level based on the evaluation result to construct a device drift dataset; S3, receiving the device drift dataset, performing graded parameter correction according to the drift level, compensating for LED fill light current, photosensitive chip dark current, and exposure time respectively, and obtaining target adjustment values ​​by combining real-time corresponding data to generate imaging control commands; S4, executing the imaging control commands, and continuously evaluating the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after command execution, determining whether to repeat parameter correction, and realizing a closed-loop compensation control for scanning imaging.

[0010] Furthermore, the scanning imaging data is periodically acquired, and time drift correction, outlier removal, dynamic compensation, and standardization are performed on the preprocessed scanning imaging data. The specific steps for obtaining the preprocessed scanning imaging data are as follows: A fixed sliding time window is used as a sampling period to periodically acquire scanning imaging data, which includes power supply voltage, output current, light intensity, photosensitive chip temperature, ambient temperature, photosensitive chip dark current, white field brightness, exposure time, and LED fill light current. The acquired scanning imaging data is aligned to a unified time base using a multi-channel synchronous resampling and time drift correction algorithm, and the time series is reconstructed using a cubic spline interpolation algorithm to obtain a continuous and aligned data sequence. The scanning imaging data is trend smoothed and anomaly detected using an exponentially weighted moving average algorithm to identify and filter out time series fluctuations and instantaneous interference signals. Missing points in the scanning imaging data are dynamically compensated using an autoregressive moving average prediction algorithm. The data of each channel is numerically normalized using a Z-score standardization algorithm to unify the numerical scale and eliminate dimensional differences.

[0011] Further, based on the preprocessed scanned imaging data, the imaging state is determined. When the device is in the drift trigger state, the specific steps for evaluating the multi-source drift degree during the scanning imaging process are as follows: Receive the preprocessed scanned imaging data, calculate the absolute value of the difference between the real-time light intensity, the dark current of the photosensitive chip, the power supply voltage, and the corresponding calibration data, and compare it with the corresponding threshold. When any absolute value of the difference is greater than the corresponding threshold, it is determined that the device is in the imaging drift state; otherwise, it is determined that the device is in the imaging stable state and no processing is performed. When the device is in the drift trigger state, divide the ratio of the light intensity to the white field brightness by the ratio of the calibrated light intensity to the calibrated white field brightness, take the absolute value after subtracting one from the obtained result to get the optical field response deviation term; multiply the difference between the temperature of the photosensitive chip and the ambient temperature by the dark current temperature coefficient k, and perform an operation on the natural constant e with the obtained product as the exponential power to get the temperature drift exponential term; add one to the ratio of the dark current of the photosensitive chip to the calibrated dark current to get the dark current compensation term; divide the difference between the power supply voltage and the calibrated power supply voltage by the calibrated power supply voltage, and take the absolute value of the obtained ratio to get the voltage offset term; multiply the optical field response deviation term, the temperature drift exponential term, and the dark current compensation term in sequence, add the voltage offset term and then add one to get the scanning imaging drift evaluation value.

[0012] Further, based on the evaluation result, the drift level is determined. The specific steps for constructing the device drift data set are as follows: Compare the scanning imaging drift evaluation value P with the multi-level drift thresholds P1 and P2 in real time to determine the device drift level: When P ≤ P1, mark the device as a first-level drift; when P1 < P < P2, mark the device as a second-level drift; when P ≥ P2, mark the device as a third-level drift; Extract the scanning imaging drift evaluation value, the scanning imaging data, and the corresponding drift level to construct the device drift data set.

[0013] Further, the specific steps for receiving the device drift dataset and performing graded parameter correction based on the drift level are as follows: When the device is in level one drift, the current device parameters are kept unchanged, and the data acquisition frequency is increased; when the device is in level two drift, the absolute values ​​of the differences between the real-time LED fill light current, the photosensitive chip dark current, the exposure time, and the corresponding calibration data are extracted, divided by the corresponding thresholds to obtain the optical, thermal, and electrical deviation rates. The main drift types are determined based on the maximum deviation rate, specifically including optical main drift, thermal main drift, and electrical main drift, and the control quantities corresponding to the main drift types are corrected. When the equipment is in the third level of drift, the LED fill light current, photosensitive chip dark current, and exposure time are corrected simultaneously. After determining the correction strategy, real-time scanning imaging data is extracted, and compensation calculations are performed on the LED fill light current, photosensitive chip dark current, and exposure time. The compensation values ​​for fill light current, dark current, and exposure time are added to the real-time LED fill light current, photosensitive chip dark current, and exposure time, respectively, to obtain the corresponding target adjustment values. The target adjustment values ​​are written into the imaging control unit to generate imaging control commands.

[0014] Further, the specific steps for calculating the LED supplementary light current compensation are as follows: Subtract the ratio of the calibrated illuminance to the calibrated white field brightness from the ratio of the real-time illuminance to the white field brightness, and then divide by the ratio of the calibrated illuminance to the calibrated white field brightness to obtain the light field response deviation term; Multiply the scanning imaging drift evaluation value by the light field response deviation term, add one, and take the natural logarithm to obtain the light field modulation coefficient; Multiply the LED supplementary light current by the light field modulation coefficient to obtain the supplementary light current compensation value.

[0015] Further, the specific steps for calculating the dark current compensation of the photosensitive chip are as follows: Divide the absolute value of the difference between the photosensitive chip temperature and the ambient temperature by the ambient temperature to obtain the temperature drift normalization term; divide the absolute value of the difference between the power supply voltage and the calibrated power supply voltage by the calibrated power supply voltage to obtain the voltage offset term; add the temperature drift normalization term and the voltage offset term, and multiply them by the scanning imaging drift evaluation value to obtain the denominator correction factor; divide the photosensitive chip dark current by the denominator correction factor and add one to obtain the dark current compensation value.

[0016] Furthermore, the specific steps for calculating the exposure time compensation are as follows: take the natural logarithm of the ratio of real-time output current to power supply voltage to obtain the electrical signal intensity variation term; multiply the voltage sensitivity coefficient, the scanning imaging drift evaluation value, and the electrical signal intensity variation term in sequence to obtain the electrical offset factor; multiply the electrical offset factor by the exposure time to obtain the exposure time compensation value.

[0017] Furthermore, the specific steps for executing imaging control commands and continuously evaluating the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after command execution to determine whether parameter correction should be repeated, thus realizing the closed-loop compensation control for scanning imaging, are as follows: Receive imaging control commands and drive the scanning device to perform adjustments to the light source, heat source, and exposure control; after the control commands are executed, collect scanning imaging data in real time and recalculate the scanning imaging drift evaluation value to determine the device drift level; when the device is in level one drift for N consecutive sampling cycles, maintain the current parameter configuration and enter a steady-state operation mode; otherwise, perform parameter correction again until the device enters a steady-state operation mode.

[0018] The second aspect of this invention provides a portable scanning device based on AI recognition, comprising: a data acquisition and preprocessing module, an AI feature drift evaluation module, a multi-source drift parameter correction module, and an imaging quality self-calibration module, wherein: the data acquisition and preprocessing module is used to periodically acquire scanning imaging data and perform time drift correction, abnormal sample removal, dynamic compensation, and standardization processing on the scanning imaging data to obtain preprocessed scanning imaging data; the AI ​​feature drift evaluation module is used to determine the imaging state based on the preprocessed scanning imaging data, and when the device is in a drift-triggered state, it evaluates the degree of multi-source drift during the scanning imaging process and determines the drift level based on the evaluation result to construct a device drift dataset; the multi-source drift parameter correction module is used to receive the device drift dataset, perform graded parameter correction according to the drift level, compensate for LED fill light current, photosensitive chip dark current, and exposure time respectively, and obtain target adjustment values ​​by combining real-time corresponding data to generate imaging control commands; the imaging quality self-calibration module is used to execute the imaging control commands and continuously evaluate the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after the command execution, determine whether to repeat parameter correction, and realize a closed loop of compensation control for scanning imaging.

[0019] Beneficial effects

[0020] The present invention has the following beneficial effects:

[0021] (1) A portable scanning device and method based on AI recognition, which for the first time introduces multi-channel imaging feature data such as light intensity, photosensitive chip temperature, power supply voltage and dark current into a unified drift evaluation system, constructs a scanning imaging drift evaluation formula, realizes the comprehensive judgment and quantitative evaluation of optical, thermal and electrical drift, can accurately reflect the imaging stability of the device in complex operating environment, and provide a scientific basis for subsequent correction.

[0022] (2) A portable scanning device and method based on AI recognition, which dynamically divides the drift evaluation value into first-level, second-level and third-level drift levels, and automatically matches the graded parameter correction strategy for different levels to realize differentiated compensation control of optical, thermal and electrical parameters, thereby significantly improving the imaging consistency of the device under long-term operation and environmental changes.

[0023] (3) A portable scanning device and method based on AI recognition, which realizes a closed-loop control process from detection, evaluation, correction to verification through real-time re-acquisition and drift evaluation after the execution of imaging control commands. When the system is in the first-level drift state for multiple consecutive sampling cycles, it automatically enters the steady-state operation mode, effectively avoiding over-adjustment and parameter oscillation, and improving the system's self-stabilization capability and long-term reliability.

[0024] (4) A portable scanning device and method based on AI recognition, through AI-driven multi-source information fusion and self-learning calibration mechanism, enables the portable scanning device to autonomously identify drift and automatically correct imaging parameters, and can maintain high-quality imaging output under complex environments such as different light, temperature and voltage, which greatly improves the stability, intelligence level and user experience of the device. Attached Figure Description

[0025] Figure 1 This is a flowchart of a portable scanning method based on AI recognition.

[0026] Figure 2 This is a structural diagram of a portable scanning device based on AI recognition.

[0027] Figure 3 This is a drift level determination chart based on scanned imaging drift evaluation values;

[0028] Figure 4 This is a flowchart for determining the level of drift in scanning imaging. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please see Figures 1-4This invention provides a technical solution: a portable scanning method based on AI recognition, comprising: S1, periodically acquiring scanning imaging data and performing time drift correction, abnormal sample removal, dynamic compensation, and standardization processing on the scanning imaging data to obtain preprocessed scanning imaging data; S2, determining the imaging state based on the preprocessed scanning imaging data, and when the device is in a drift-triggered state, evaluating the degree of multi-source drift during the scanning imaging process, and determining the drift level based on the evaluation result to construct a device drift dataset; S3, receiving the device drift dataset, performing graded parameter correction according to the drift level, compensating for LED fill light current, photosensitive chip dark current, and exposure time respectively, and obtaining target adjustment values ​​by combining real-time corresponding data to generate imaging control commands; S4, executing the imaging control commands, and continuously evaluating the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after the command execution, determining whether to repeat parameter correction, and realizing a closed loop of compensation control for scanning imaging.

[0031] Specifically, the process involves periodically acquiring scanning imaging data and performing time drift correction, outlier removal, dynamic compensation, and standardization on the data to obtain preprocessed scanning imaging data. The specific steps are as follows: A fixed sliding time window is used as a sampling period. Within each sampling period, the following data are sequentially acquired: power supply voltage, output current, illumination intensity, photosensitive chip temperature, ambient temperature, photosensitive chip dark current, white field brightness, exposure time, and LED illumination current. Power supply voltage and output current are acquired through a power sampling circuit; illumination intensity and white field brightness are acquired through an optical detection component; photosensitive chip temperature and ambient temperature are measured through a temperature detection device; and photosensitive chip dark current, exposure time, and LED illumination current are read through imaging control logic. All of these data are timestamped and form a synchronous sampling sequence to characterize the changes in electrical, optical, and thermal features during the scanning imaging process. The acquired scanning imaging data is then aligned to a unified time base using multi-channel synchronous resampling and a time drift correction algorithm to ensure consistent temporal correspondence between signals from different channels. During resampling, using the calibrated sampling frequency as a reference, phase deviations caused by sampling delays are corrected through linear interpolation and a time mapping function. Then, a cubic spline interpolation algorithm is used to continuously reconstruct the time series, resulting in a continuous data sequence with consistent time-domain resolution and time alignment. After time-base alignment, an exponentially weighted moving average algorithm is used to smooth the scanning imaging data and detect anomalies, suppressing short-term spikes and abrupt signals, identifying and filtering out abnormal samples caused by transient noise, voltage fluctuations, temperature disturbances, and light source flicker, ensuring the statistical stability of the imaging feature sequence. For the smoothed scanning imaging data, an autoregressive moving average prediction algorithm is used to dynamically compensate for detected missing points and anomaly removal points. The compensation process automatically establishes a prediction model based on time series correlation, using the autoregressive characteristics and moving average trends of historical data to generate predicted values, continuously repairing data gaps and restoring the integrity and coherence of the scanning imaging data. Finally, the Z-score normalization algorithm is used to normalize the data of each channel, calculate the deviation of each channel sample from the corresponding mean and standard deviation, map data of different dimensions to a unified dimensionless numerical range, thereby unifying the numerical scale, eliminating differences in physical units, and providing highly consistent and comparable input data for subsequent drift feature extraction and AI feature drift evaluation.

[0032] In this implementation scheme, by synchronously acquiring and uniformly aligning the power supply voltage, output current, light intensity, photosensitive chip temperature, ambient temperature, photosensitive chip dark current, white field brightness, exposure time, and LED illumination current within a fixed sampling period, consistency in the temporal dimension and comparability in the numerical dimension of the imaging data are achieved. Multi-channel resampling and time drift correction methods ensure accurate mapping of different types of data under the same time base, significantly reducing timing deviations caused by sampling delays. An exponentially weighted moving average algorithm smoothly suppresses instantaneous noise and abrupt interference, maintaining the continuity of the signal trend. An autoregressive moving average prediction algorithm dynamically compensates for missing and outlier points, improving the integrity and continuity of the data sequence. Finally, Z-score normalization maps the data of each channel to a unified numerical scale, providing a high-precision and highly stable input data foundation for subsequent drift assessment and parameter correction, thereby significantly improving the reliability and repeatability of the scanning imaging process.

[0033] Specifically, based on the pre-processed scanning imaging data, the imaging state is determined. When the device is in a drift-triggered state, the specific steps for evaluating the degree of multi-source drift during the scanning imaging process are as follows: The pre-processed scanning imaging data is received, and the illumination intensity, photosensitive chip dark current, and power supply voltage of each channel are analyzed in real time. The absolute values ​​of the differences between the real-time illumination intensity, photosensitive chip dark current, and power supply voltage and the corresponding calibration data are calculated and compared with the corresponding thresholds to quantify the stability of the optical, thermal, and electrical channels during the imaging process. The calibration data comes from a set of reference parameters collected under a standard imaging environment during the device's maintenance and calibration phase. These parameters include illumination intensity, white field brightness, photosensitive chip dark current, LED fill light current, and exposure time, serving as a comparison benchmark for subsequent real-time data during imaging. When the absolute value of any difference exceeds the corresponding threshold, it indicates a significant deviation in the device's key parameters, and the device is determined to be in an imaging drift state. If the absolute values ​​of all differences are within the threshold range, the device is determined to be in an imaging stable state, and the current operating parameters are maintained without compensation or adjustment. When the device is determined to be in a drift-triggered state, the optical, thermal, and electrical drift during the scanning imaging process are further quantitatively evaluated. First, the ratio of real-time illumination intensity to white field brightness is divided by the ratio of calibrated illumination intensity to calibrated white field brightness to reflect the relative response change of the current light field. The absolute value of the result minus one is used to obtain the light field response deviation term, which characterizes the optical deviation caused by illumination non-uniformity. Second, the difference between the photosensitive chip temperature and the ambient temperature is multiplied by the dark current temperature coefficient, and the product is used as an exponent to calculate the natural constant e, resulting in the temperature drift exponent term, which characterizes the impact of temperature differences on imaging stability. Then, the ratio of the photosensitive chip dark current to the calibrated dark current is added by one to obtain the dark current compensation term, which reflects the imaging noise shift caused by dark current changes. Finally, the difference between the power supply voltage and the calibrated power supply voltage is divided by the calibrated power supply voltage to calculate the voltage offset term, which characterizes the impact of power supply fluctuations on signal amplitude. Finally, the optical field response deviation term, temperature drift index term, and dark current compensation term are multiplied sequentially to form a multi-source drift composite response. This, combined with the voltage offset term, is then incremented by one, and the natural logarithm is taken to obtain the scanning imaging drift evaluation value. This evaluation value comprehensively quantifies the cumulative effects of optical, thermal, and electrical drifts during the imaging process and can serve as a core reference indicator for the stability of equipment operation and the accuracy of parameter compensation. The dark current temperature coefficient is obtained using experimental measurement data of the photosensitive chip's dark current at different ambient temperatures during the historical calibration phase, through a least-squares exponential fitting algorithm, and its value ranges from (0,1).

[0034] The specific formula for calculating the scanning imaging drift evaluation value is as follows:

[0035] ;

[0036] In the formula, This represents the scanned imaging drift assessment value. Indicates light intensity. Indicates white field brightness. Indicates the temperature of the photosensitive chip. Indicates ambient temperature. This indicates the dark current of the photosensitive chip. Indicates the power supply voltage. Indicates the calibrated light intensity. Indicates the calibrated white point brightness. Indicates the calibrated dark current. Indicates the rated power supply voltage. This represents the temperature coefficient of dark current.

[0037] In this embodiment, Table 1 is a data table of scanning imaging drift evaluation values, showing the key imaging parameters and the finally calculated scanning imaging drift evaluation values ​​over five sampling periods. The data in the table includes: illumination intensity, white field brightness, photosensitive chip temperature, ambient temperature, photosensitive chip dark current, power supply voltage, and scanning imaging drift evaluation values. The specific details are as follows: When the calibrated illumination intensity is 1000, the calibrated white point brightness is 250, the calibrated dark current is 2.0, the calibrated power supply voltage is 5.00, and the dark current temperature coefficient is 0.015, in sampling period 1, the illumination intensity is 995, the white point brightness is 248, the photosensitive chip temperature is 36, the ambient temperature is 35, the dark current is 2.1, and the power supply voltage is 4.98, resulting in a final calculated scanning imaging drift evaluation value of 0.010; in sampling period 2, the illumination intensity is 970, the white point brightness is 247, the photosensitive chip temperature is 37, the ambient temperature is 35, the dark current is 2.3, and the power supply voltage is 4.95, resulting in a drift evaluation value of 0.0. 49; In sampling period 3, the illuminance was 940, the white point brightness was 245, the photosensitive chip temperature was 38, the ambient temperature was 35, the dark current was 2.6, the power supply voltage was 4.93, and the drift evaluation value was 0.106; In sampling period 4, the illuminance was 905, the white point brightness was 244, the photosensitive chip temperature was 40, the ambient temperature was 35, the dark current was 3.0, the power supply voltage was 4.90, and the drift evaluation value was 0.196; In sampling period 5, the illuminance was 860, the white point brightness was 242, the photosensitive chip temperature was 42, the ambient temperature was 35, the dark current was 3.4, the power supply voltage was 4.87, and the drift evaluation value was 0.308.

[0038] Table 1. Data on Scanning Imaging Drift Evaluation Values

[0039]

[0040] like Figure 3The figure shows the scanning imaging drift evaluation values ​​and corresponding drift levels for five sampling periods. Different colored dots are used to clearly distinguish the imaging drift levels for each sampling period. Specifically, green dots represent Level 1 drift, orange dots represent Level 2 drift, and red dots represent Level 3 drift. Furthermore, dashed lines are used to plot the Level 1 and Level 2 drift thresholds, providing an intuitive reference for determining the drift level for each sampling period, facilitating rapid identification of changes in imaging system stability and enabling parameter correction. As can be seen from the figure, sampling periods 1 and 2 are in the Level 1 drift range, indicating relatively stable imaging; sampling periods 3 and 4 are in the Level 2 drift range; and sampling period 5 is in the Level 3 drift range, indicating significant drift and requiring multi-source parameter joint correction. Figure 3 This intuitively reflects the sensitivity and classification ability of the scanning imaging drift evaluation method proposed in this invention under complex operating environments, providing a decision-making basis for subsequent adaptive compensation control and imaging quality self-calibration.

[0041] In this implementation scheme, by dynamically comparing and calculating the light intensity, dark current of the photosensitive chip, power supply voltage, photosensitive chip temperature, ambient temperature, and white field brightness during the imaging process, real-time identification of imaging stability can be achieved under the combined influence of multiple factors. By jointly quantifying the light field response deviation, temperature drift index, dark current compensation, and voltage offset, the combined effects of optical drift, thermal drift, and electrical drift are accurately characterized, thereby improving the sensitivity and accuracy of imaging drift identification. This method enables real-time perception of the device drift state under different operating conditions and provides a reliable quantitative basis for subsequent parameter compensation and imaging control, effectively ensuring the stability and repeatability of the scanning imaging process.

[0042] Specifically, based on the evaluation results, the drift level is determined, and the specific steps for constructing the device drift dataset are as follows: The scanning imaging drift evaluation value P is compared with the multi-level drift thresholds P1 and P2 in real time, and the current operating stable state and drift level of the device are determined according to the relative relationship between the evaluation value and the thresholds. When P ≤ P1, it indicates that the optical, thermal, and electrical states of the device under the current working conditions are all within the stable range, and the imaging performance is close to the calibration reference, and it is determined to be in the first-level drift state; when P1 < P < P2, it means that there are certain fluctuations in the imaging characteristics of the device, and some channels deviate from the calibration conditions, and it is determined to be in the second-level drift state; when P ≥ P2, it shows that significant offsets have occurred in the light response, temperature characteristics, and power supply voltage of the device, and the imaging performance is significantly affected, and it is determined to be in the third-level drift state. After the drift level is determined, the scanning imaging drift evaluation value, scanning imaging data, and the determined drift level corresponding to the corresponding time are extracted, and the above data are associated in chronological order to form a device drift dataset with time series tags and level characteristics. This dataset is used to record the drift evolution law of the device in different operating cycles, providing data support for subsequent parameter compensation, imaging control, and long-term stability evaluation.

[0043] In this implementation scheme, by introducing the multi-level drift thresholds P1 and P2 to perform hierarchical determination on the scanning imaging drift evaluation value P, a refined quantitative division of the imaging stability is achieved, enabling the device to timely identify different degrees of drift states caused by optical, thermal, and electrical factors during operation. This method effectively enhances the dynamic monitoring ability of the imaging process, establishing a clear corresponding relationship between the drift level and the imaging performance, providing an accurate basis for subsequent parameter compensation and imaging control, and thus significantly improving the imaging stability and reliability of the device in a complex environment.

[0044] Specifically, after receiving the device drift dataset, the hierarchical parameter correction is carried out according to the drift level, and the LED supplementary light current, the dark current of the photosensitive chip, and the exposure time are compensated respectively, and the target adjustment value is obtained by combining the real-time corresponding data. The specific steps for generating the imaging control instruction are as follows: After receiving the device drift dataset, first, the imaging control parameters are hierarchically determined and the strategy is selected according to the drift level to ensure the pertinence and hierarchy of the compensation process. For example Figure 4As shown, when the device is in the first-level drift state, it indicates that the scanning imaging drift evaluation value P is within a stable threshold range, and the imaging state fluctuates relatively little. To prevent over-correction from causing an imbalance in the system's dynamic response, the current device parameters are kept unchanged, and only the data acquisition frequency is increased to enhance subsequent monitoring accuracy and accumulate steady-state operation samples. When the device is in the second-level drift state, the system's optical, thermal, and electrical characteristics experience slight shifts. To identify the main drift type, the absolute values ​​of the differences between the real-time LED illumination current, photosensitive chip dark current, exposure time, and the corresponding calibration data are extracted and divided by their respective set thresholds to obtain the optical deviation rate, thermal deviation rate, and electrical deviation rate. The dominant drift factor is determined by comparing the magnitudes of these three values. Targeted corrections are implemented based on the drift type corresponding to the maximum deviation rate to ensure that the correction amount matches the drift source, improving the convergence efficiency and correction accuracy of parameter adjustments. When the device is in the third-level drift state, it indicates that the scanning imaging drift evaluation value P has exceeded the upper limit threshold, meaning that the light intensity, photosensitive chip dark current, and exposure time all significantly deviate from the calibration reference. At this point, the three control parameters mentioned above need to be simultaneously corrected to avoid secondary shifts caused by single-dimensional compensation. After determining the correction strategy, real-time scanning imaging data is extracted, and the compensation values ​​for LED fill light current, photosensitive chip dark current, and exposure time are calculated respectively. The compensation results are then superimposed on the real-time LED fill light current, photosensitive chip dark current, and exposure time to obtain the corrected target adjustment value. Finally, the target adjustment value is written to the imaging control unit to generate a new imaging control command, driving the device to perform dynamic parameter updates in the next sampling cycle, ensuring that the imaging process maintains high stability and uniform optical response under multi-source drift conditions.

[0045] In this implementation scheme, by implementing graded parameter correction based on drift level, a layered response and precise correction of optical, thermal, and electrical offsets during the scanning imaging process is achieved. This method can automatically match the optimal correction strategy according to different degrees of drift, making the adjustment of LED illumination current, photosensitive chip dark current, and exposure time more sensitive and controllable. By extracting scanning imaging data in real time and dynamically updating imaging control commands, not only is secondary drift caused by overcorrection effectively avoided, but the imaging consistency and data stability of the equipment in complex environments are also significantly improved, providing a reliable guarantee for achieving high-precision, low-drift imaging control.

[0046] Specifically, the calculation steps for LED supplementary lighting current compensation are as follows: Subtract the ratio of calibrated illumination intensity to calibrated white field brightness from the ratio of real-time illumination intensity to white field brightness, then divide by the ratio of calibrated illumination intensity to calibrated white field brightness to obtain the light field response deviation term. This ratio reflects the degree of deviation of the current imaging light source output from the calibration conditions, thus reflecting the dynamic changes in the light field energy distribution. To further improve response sensitivity, time synchronization correction and transient noise filtering are performed on the sampling signals of real-time illumination intensity and white field brightness to ensure that the calculated light field response deviation term truly reflects the changing trend of the optical channel during imaging. After obtaining the light field response deviation term, multiply the scanning imaging drift evaluation value by the light field response deviation term, add one, and take the natural logarithm to obtain the light field modulation coefficient. This light field modulation coefficient is used to comprehensively characterize the overall optical attenuation trend and dynamic drift amplitude during imaging, thereby achieving quantitative adjustment of illumination changes. To ensure modulation accuracy, a logarithmic compression function is introduced in the calculation process of the light field modulation coefficient to weaken the influence of extreme values ​​on the overall compensation process. Finally, the LED supplementary light current is multiplied by the light field modulation coefficient to obtain the supplementary light current compensation value. This compensation value is used to correct the LED drive current during subsequent control command generation, ensuring that the LED light output remains dynamically consistent with the calibration state.

[0047] The specific formula for calculating the compensation value of the supplementary light current is as follows:

[0048] ;

[0049] In the formula, This indicates the compensation value for the supplemental light current. Indicates the LED fill light current. This represents the scanned imaging drift assessment value. Indicates light intensity. Indicates white field brightness. Indicates the calibrated light intensity. This indicates the calibrated white field brightness.

[0050] In this implementation scheme, a joint compensation mechanism involving the light field response deviation term and the light field modulation coefficient is introduced to achieve fine-tuning of the LED supplementary light current, ensuring high consistency between the light source output and the calibrated illumination intensity and calibrated white field brightness. This method can dynamically correct brightness shifts caused by illumination changes during imaging, significantly improving the stability of exposure energy. Through time synchronization correction and noise suppression processing of the real-time illumination signal, the compensation result is smoother and more continuous, effectively avoiding local brightness fluctuations caused by light field jitter, thereby improving the uniformity of scanning imaging brightness and detail reproduction, providing highly robust optical compensation support for maintaining image quality.

[0051] Specifically, the steps for calculating the dark current compensation of the photosensitive chip are as follows: The absolute value of the difference between the photosensitive chip temperature and the ambient temperature is divided by the ambient temperature to obtain the temperature drift normalization term. This term characterizes the degree of temperature drift of the photosensitive chip under the current operating environment, quantifying the impact of temperature changes on dark current stability. Before performing this calculation, the real-time sampling signals of the photosensitive chip temperature and ambient temperature are synchronized and filtered multiple times to ensure the continuity and accuracy of the temperature input data, providing a reliable basis for subsequent thermal drift compensation. The absolute value of the difference between the power supply voltage and the calibrated power supply voltage is divided by the calibrated power supply voltage to obtain the voltage offset term. This term reflects the impact of power supply fluctuations on the dark current output of the photosensitive chip, and can dynamically correct signal deviations caused by unstable power supply. An exponential weighted smoothing algorithm is introduced during voltage acquisition to suppress transient voltage fluctuations, making the calculation results more stable and reliable. The temperature drift normalization term and the voltage offset term are added together and multiplied by the scanning imaging drift evaluation value, serving as the denominator correction factor. The denominator correction factor is used to comprehensively characterize the coupling effect of thermal drift and electrical drift, thereby achieving a composite correction of the dark current variation trend. This multi-parameter superposition method maintains a balance between computational stability and response sensitivity under multi-source interference conditions. Finally, the dark current of the photosensitive chip is divided by the denominator correction factor and one is added to obtain the dark current compensation value. This compensation value is used to adjust the output response of the photosensitive chip in real time, thereby significantly reducing dark noise interference and improving the signal-to-noise ratio of low-light imaging.

[0052] The specific formula for calculating the dark current compensation value is as follows:

[0053] ;

[0054] In the formula, This represents the dark current compensation value. This indicates the dark current of the photosensitive chip. This represents the scanned imaging drift assessment value. Indicates the temperature of the photosensitive chip. Indicates ambient temperature. Indicates the power supply voltage. This indicates the rated power supply voltage.

[0055] In this implementation scheme, a composite compensation mechanism is constructed, centered on a temperature drift normalization term, a voltage offset term, and a scanning imaging drift evaluation value. This mechanism enables precise dynamic correction of the dark current of the photosensitive chip, allowing the dark current output to remain stable under different ambient temperatures and power supply voltages. This method effectively reduces the interference of temperature changes and voltage fluctuations on the photosensitive chip output, significantly improving the linearity and stability of dark current measurement. By introducing a coupled calculation method with a denominator correction factor, the comprehensive ability to offset the effects of multi-source drift during the compensation process is enhanced, resulting in a more balanced low-light response during imaging and improving the overall imaging signal-to-noise ratio and grayscale restoration accuracy.

[0056] Specifically, the steps for calculating exposure time compensation are as follows: The natural logarithm of the ratio of real-time output current to power supply voltage is taken to obtain the electrical signal intensity variation term. This term characterizes the dynamic response relationship between output current and power supply voltage during imaging, reflecting the changing trend of signal transmission stability in the electrical channel. Before performing the ratio calculation, the sampled data of output current and power supply voltage are synchronously resampled and noise suppressed to ensure the continuity and accuracy of the ratio calculation results, thereby avoiding calculation errors caused by transient signal fluctuations. The voltage sensitivity coefficient, the scanning imaging drift evaluation value, and the electrical signal intensity variation term are multiplied sequentially to obtain the electrical offset factor. This factor quantifies the influence of voltage changes on exposure time control and reflects the electrical drift deviation caused by differences in power supply stability and current response during imaging. The voltage sensitivity coefficient is obtained by exponentially fitting the logarithmic relationship between voltage changes and electrical signal intensity changes based on multiple sets of power supply voltage and output current measurement data under constant exposure time and constant illumination conditions during the calibration phase. It characterizes the sensitivity of voltage fluctuations to the amplitude of the electrical signal response and ranges from 0 to 1. The electrical offset factor is multiplied by the exposure time to obtain the exposure time compensation value. This value is used to correct the exposure duration in real time during the scanning imaging process, enabling precise matching of exposure control based on changes in electrical characteristics. This maintains image brightness consistency and signal output stability under conditions of light source fluctuations, current offsets, and drift accumulation.

[0057] The specific formula for calculating the exposure time compensation value is as follows:

[0058] ;

[0059] In the formula, This represents the exposure time compensation value. Indicates the exposure time. Indicates the output current. Indicates the power supply voltage. This represents the scanned imaging drift assessment value. This represents the voltage sensitivity coefficient.

[0060] In this implementation scheme, by establishing a coupled compensation mechanism between the electrical signal intensity variation term and the electrical offset factor, adaptive correction of exposure time to voltage fluctuations and current response changes is achieved, making exposure control more sensitive and stable. This method can sense the dynamic offset of the electrical channel in real time during the scanning imaging process, ensuring consistency between exposure time compensation and drift evaluation results, thereby significantly improving the stability of image brightness and detail reproduction. The precise introduction of the voltage sensitivity coefficient enables higher linear control accuracy in the exposure time adjustment process, effectively improving the grayscale uniformity and optical signal consistency of the imaging output.

[0061] Specifically, the steps for executing imaging control commands and continuously evaluating the degree of multi-source drift during the scanning imaging process based on the scanned imaging data after command execution to determine whether parameter correction should be repeated, thus achieving a closed-loop compensation control for scanning imaging, are as follows: Receiving imaging control commands drives the scanning device to perform adjustments to the light source, heat source, and exposure control. After receiving the control commands, target adjustment values ​​for LED supplemental lighting current, photosensitive chip dark current, and exposure time are synchronously issued and executed item by item to ensure consistent dynamic responses in light intensity, chip temperature, and exposure duration. During execution, the internal clock signal is used to align the command triggering and response delays, ensuring that each control action is completed synchronously within the scanning cycle, thereby avoiding response mismatches between optical, thermal, and electrical parameters in different channels. After the control commands are executed, scanned imaging data is acquired in real time, and the scanned imaging drift evaluation value is recalculated to determine the device drift level. This step immediately initiates the data sampling process after each imaging adjustment, synchronously measuring key physical quantities such as light intensity, output current, photosensitive chip temperature, power supply voltage, and photosensitive chip dark current. By recalculating the scanning imaging drift evaluation value and comparing the trend of data changes before and after, the imaging stability and parameter consistency of the adjusted device are identified, thereby achieving dynamic verification of the compensation effect and continuous tracking of the drift trend. When the device is in the first level of drift for N consecutive sampling cycles, the current parameter configuration is maintained, and the device enters the steady-state operation mode. Here, N is a positive integer greater than one. In the steady-state operation mode, the scanning device maintains the existing light source drive, power supply voltage, and exposure time, and no longer performs parameter correction operations to ensure the long-term consistency of the imaging output and the stable maintenance of the signal-to-noise ratio. Otherwise, parameter correction is performed again until the device enters the steady-state operation mode. This process automatically triggers the next round of parameter adjustment through a feedback correction mechanism, dynamically updates the compensation strategy according to the drift level, and achieves continuous optimization of illumination compensation, dark current compensation, and exposure time compensation, ultimately forming adaptive coupling and long-term closed-loop control among multi-source parameters.

[0062] In this implementation scheme, a closed-loop feedback mechanism is introduced after the imaging control is executed, achieving dynamic coordination between light source driving, heat source adjustment, and exposure control, thus enabling parameter correction to possess self-learning and progressive stabilization characteristics. This method can evaluate the scanning imaging drift value in real time after each instruction execution, and achieve state recognition and parameter maintenance through continuous sampling period level determination, avoiding imaging fluctuations caused by frequent adjustments. By establishing an automatic backoff and re-correction mechanism based on drift trends, the illumination intensity, photosensitive chip dark current, and exposure time can continuously maintain an optimal matching state under complex environmental changes, thereby significantly improving the stability and long-term accuracy of scanning imaging.

[0063] like Figure 2 As shown, a second aspect of the present invention provides a portable scanning device based on AI recognition, comprising: a data acquisition and preprocessing module, an AI feature drift evaluation module, a multi-source drift parameter correction module, and an imaging quality self-calibration module, wherein: the data acquisition and preprocessing module is used to periodically acquire scanning imaging data and perform time drift correction, abnormal sample removal, dynamic compensation, and standardization processing on the scanning imaging data to obtain preprocessed scanning imaging data; the AI ​​feature drift evaluation module is used to determine the imaging state based on the preprocessed scanning imaging data, and when the device is in a drift-triggered state, evaluate the scanning imaging process. The system assesses the degree of multi-source drift and determines the drift level based on the evaluation results, constructing a device drift dataset. A multi-source drift parameter correction module receives the device drift dataset, performs graded parameter correction according to the drift level, compensates for LED fill light current, photosensitive chip dark current, and exposure time, and obtains target adjustment values ​​by combining real-time corresponding data, generating imaging control commands. An imaging quality self-calibration module executes the imaging control commands and continuously evaluates the degree of multi-source drift during the scanning imaging process based on the scanned imaging data after command execution, determining whether to repeat parameter correction to achieve a closed-loop compensation control for scanning imaging.

[0064] In this implementation scheme, a closed-loop design integrating data acquisition, feature evaluation, parameter correction, and imaging self-calibration is structurally employed, enabling the scanning imaging process to achieve real-time learning and adaptive adjustment capabilities. The device can automatically identify optical, thermal, and electrical drift sources during imaging and dynamically adjust the LED illumination current, photosensitive chip dark current, and exposure time for different drift levels, effectively eliminating imaging deviations caused by environmental fluctuations. Through continuous feedback and hierarchical correction mechanisms, the device can achieve parameter self-balancing and long-term image quality stability under complex operating conditions, significantly improving scanning accuracy and output consistency, providing technical support for achieving highly reliable portable intelligent imaging.

[0065] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0066] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A portable scanning method based on AI recognition, characterized in that, Includes the following steps: S1 periodically acquires scanning imaging data and performs time drift correction, abnormal sample removal, dynamic compensation and standardization on the scanning imaging data to obtain preprocessed scanning imaging data; S2, determine the imaging status based on the preprocessed scanning imaging data. When the device is in a drift-triggered state, evaluate the degree of multi-source drift during the scanning imaging process, determine the drift level based on the evaluation results, and construct the device drift dataset. S3 receives the device drift dataset, performs graded parameter correction based on the drift level, compensates for the LED fill light current, the dark current of the photosensitive chip and the exposure time respectively, and obtains the target adjustment value by combining the real-time corresponding data, and generates imaging control commands. S4 executes imaging control commands and continuously evaluates the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after the commands are executed, determines whether to repeat parameter correction, and realizes the compensation control closed loop of scanning imaging.

2. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for periodically acquiring scanning imaging data and performing time drift correction, outlier removal, dynamic compensation, and standardization on the scanning imaging data to obtain preprocessed scanning imaging data are as follows: A fixed sliding time window is used as a sampling period to periodically collect scanning imaging data, which includes power supply voltage, output current, light intensity, photosensitive chip temperature, ambient temperature, photosensitive chip dark current, white field brightness, exposure time, and LED fill light current. The acquired scanning imaging data is aligned to a unified time base using a multi-channel synchronous resampling and time drift correction algorithm, and the time series is reconstructed using a cubic spline interpolation algorithm to obtain a continuous and aligned data sequence. The scanning imaging data is then smoothed for trend and anomaly detection using an exponential weighted moving average algorithm to identify and filter out temporal fluctuations and instantaneous interference signals. Missing points in the scanning imaging data are dynamically compensated using an autoregressive moving average prediction algorithm. Finally, the data from each channel is normalized using a Z-score normalization algorithm to unify the numerical scale and eliminate dimensional differences.

3. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for determining the imaging state based on preprocessed scanning imaging data, when the device is in a drift-triggered state, to evaluate the degree of multi-source drift during the scanning imaging process are as follows: The system receives preprocessed scanning imaging data, calculates the absolute value of the difference between real-time light intensity, photosensitive chip dark current, power supply voltage, and corresponding calibration data, and compares it with the corresponding threshold. If any absolute value of the difference is greater than the corresponding threshold, the device is determined to be in an imaging drift state; otherwise, the device is determined to be in an imaging stable state and no processing is performed. When the device is in the drift trigger state, divide the ratio of the light intensity to the white field brightness by the ratio of the calibrated light intensity to the calibrated white field brightness. Take the absolute value of the result after subtracting one to obtain the optical field response deviation term. Multiply the difference between the photosensitive chip temperature and the ambient temperature by the dark current temperature coefficient k, and perform an operation on the natural constant e with the obtained product as the exponential power to obtain the temperature drift exponential term. Add one to the ratio of the photosensitive chip dark current to the calibrated dark current to obtain the dark current compensation term. Divide the difference between the power supply voltage and the calibrated power supply voltage by the calibrated power supply voltage, and take the absolute value of the obtained ratio to obtain the voltage offset term. Multiply the optical field response deviation term, the temperature drift exponential, and the dark current compensation term in sequence, then add the voltage offset term and one to obtain the scanning imaging drift evaluation value.

4. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for determining the drift level based on the evaluation result and constructing the device drift data set are as follows: Compare the scanning imaging drift evaluation value P with the multi-level drift thresholds P1 and P2 in real time to determine the device drift level: When P ≤ P1, mark the device as a first-level drift; When P1 < P < P2, mark the device as a second-level drift; When P ≥ P2, mark the device as a third-level drift; Extract the scanning imaging drift evaluation value, the scanning imaging data, and the corresponding drift level to construct the device drift data set.

5. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for receiving the device drift data set, performing hierarchical parameter correction based on the drift level, compensating the LED supplementary light current, the photosensitive chip dark current, and the exposure time respectively, and obtaining the target adjustment value by combining with the real-time corresponding data to generate the imaging control instruction are as follows: Receive the device drift data set and perform hierarchical parameter correction based on the drift level: When the device is in the first-level drift, keep the current device parameters unchanged and increase the data acquisition frequency; when the device is in the second-level drift, extract the absolute values of the differences between the real-time LED supplementary light current, the photosensitive chip dark current, the exposure time, and the corresponding calibrated data, divide them by the corresponding thresholds respectively to obtain the optical, thermal, and electrical deviation rates. Determine the main drift type based on the maximum deviation rate, specifically including the optical main drift, the thermal main drift, and the electrical main drift, and correct the control quantity corresponding to the main drift type; When the device is in the third-level drift, correct the LED supplementary light current, the photosensitive chip dark current, and the exposure time simultaneously; After determining the correction strategy, extract the real-time scanning imaging data, perform compensation calculations on the LED supplementary light current, the photosensitive chip dark current, and the exposure time. Add the supplementary light current compensation value, the dark current compensation value, and the exposure time compensation value to the real-time LED supplementary light current, the photosensitive chip dark current, and the exposure time respectively to obtain the corresponding target adjustment values; write the target adjustment values into the imaging control unit to generate the imaging control instruction.

6. The portable scanning method based on AI recognition according to claim 5, characterized in that: The specific steps for compensating and calculating the LED supplementary light current are as follows: Subtract the ratio of the calibrated light intensity to the calibrated white field brightness from the ratio of the real-time light intensity to the white field brightness, and then divide by the ratio of the calibrated light intensity to the calibrated white field brightness to obtain the optical field response deviation term; multiply the scanning imaging drift evaluation value by the optical field response deviation term and then add one, and take the natural logarithm to obtain the optical field modulation coefficient; multiply the LED supplementary light current by the optical field modulation coefficient to obtain the supplementary light current compensation value.

7. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for calculating the dark current compensation of the photosensitive chip are as follows: The absolute value of the difference between the photosensitive chip temperature and the ambient temperature is divided by the ambient temperature to obtain the temperature drift normalization term; the absolute value of the difference between the power supply voltage and the calibration power supply voltage is divided by the calibration power supply voltage to obtain the voltage offset term; the temperature drift normalization term and the voltage offset term are added together and multiplied by the scanning imaging drift evaluation value to obtain the denominator correction factor; the dark current of the photosensitive chip is divided by the denominator correction factor and one is added to obtain the dark current compensation value.

8. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for calculating the exposure time compensation are as follows: The natural logarithm of the ratio of real-time output current to power supply voltage is taken to obtain the electrical signal intensity variation term; the voltage sensitivity coefficient, the scanning imaging drift evaluation value, and the electrical signal intensity variation term are multiplied in sequence to obtain the electrical offset factor; the electrical offset factor is multiplied by the exposure time to obtain the exposure time compensation value.

9. The portable scanning method based on AI recognition according to claim 1, characterized in that: The specific steps for executing imaging control commands, continuously evaluating the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after command execution, and determining whether to repeatedly perform parameter correction to achieve a closed-loop compensation control for scanning imaging are as follows: Upon receiving imaging control commands, the scanning device is driven to perform adjustments to the light source, heat source, and exposure control. After the control commands are executed, scanning imaging data is acquired in real time, and the scanning imaging drift evaluation value is recalculated to determine the device drift level. When the device is in the first-level drift for N consecutive sampling cycles, the current parameter configuration is maintained and the device enters the steady-state operation mode. Otherwise, the parameters are readjusted until the device enters the steady-state operation mode.

10. A portable scanning device based on AI recognition, characterized in that: include: The module includes a data acquisition and preprocessing module, an AI feature drift evaluation module, a multi-source drift parameter correction module, and an imaging quality self-calibration module, among which: The data acquisition and preprocessing module is used to periodically acquire scanning imaging data and perform time drift correction, abnormal sample removal, dynamic compensation and standardization on the scanning imaging data to obtain preprocessed scanning imaging data. The AI ​​feature drift evaluation module is used to determine the imaging state based on the preprocessed scanning imaging data. When the device is in a drift-triggered state, it evaluates the degree of multi-source drift during the scanning imaging process, determines the drift level based on the evaluation results, and constructs a device drift dataset. The multi-source drift parameter correction module is used to receive the device drift dataset, perform graded parameter correction according to the drift level, compensate for the LED fill light current, the dark current of the photosensitive chip and the exposure time respectively, and obtain the target adjustment value by combining the real-time corresponding data to generate imaging control commands. The imaging quality self-calibration module is used to execute imaging control commands and continuously evaluate the degree of multi-source drift during the scanning imaging process based on the scanning imaging data after the command execution, determine whether to repeat parameter correction, and realize the compensation control closed loop of scanning imaging.

Images