Intelligent cabin environment sensing and control method based on vehicle sensor

Abstract

The invention discloses an intelligent cabin environment sensing and control method based on a vehicle sensor, and relates to the technical field of intelligent automobiles, and the method comprises the following steps: in the operation process of an intelligent cabin, an infrared thermal sensing system collects thermal imaging data in an automobile in real time, extracts multi-dimensional feature data based on the obtained thermal imaging data, and sends the multi-dimensional feature data to the intelligent cabin; and forming an infrared thermal image characteristic data parameter set. According to the method, the key features of the infrared thermogram are extracted, and the machine learning model is combined, so that accurate identification and dynamic suppression of the thermo-optical interference state are realized. According to the method, sampling and filtering strategies can be adaptively adjusted in complex environments such as strong light reflection, a real thermal image contour is reconstructed, a non-human body heat source is effectively eliminated, the recognition accuracy and the operation stability of an infrared thermal sensing system are remarkably improved, the misjudgment rate is reduced, and the safety and the reliability of an intelligent cabin are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent vehicle technology, and in particular to an intelligent cockpit environment perception and control method based on vehicle sensors. Background Art

[0002] "Intelligent cockpit environment perception and control based on vehicle sensors" refers to the use of various sensors deployed inside and around the vehicle (such as temperature and humidity sensors, light sensors, Concentration sensors, PM2.5 sensors, noise detectors, infrared thermal sensors, driver status monitoring cameras, etc., provide real-time perception and data collection of the in-vehicle environment and occupant physiological and behavioral characteristics. Intelligent algorithms then fuse and analyze these perception results to determine whether the current cabin environment is within comfortable, safe, or healthy ranges. If an environmental parameter anomaly is detected (such as excessive temperature, degraded air quality, or excessive noise), the system automatically controls actuators such as the vehicle's air conditioning, air purification system, seat ventilation and heating, sunshade system, or ambient lighting, performing adaptive adjustments to achieve intelligent management of the cabin microenvironment, thereby enhancing driving and riding comfort, safety, and a personalized experience. This approach is widely used in intelligent cockpit systems in smart cars and is a crucial component of in-vehicle human-computer interaction and contextual awareness control.

[0003] Existing technologies have the following shortcomings: Infrared thermal sensing systems are often used in existing vehicle intelligent cockpits to provide real-time sensing of occupant physiological and behavioral characteristics (such as movement frequency, duration of eye closure, and body temperature distribution) to enable early identification of driver fatigue, hypoxia, or behavioral abnormalities. However, in practical applications, infrared thermal sensing systems are susceptible to strong environmental interference. Specifically, when the vehicle is exposed to direct sunlight, the window film is faded and aged, or there are reflective metallic surfaces inside the vehicle, external heat sources or reflected infrared energy may directly enter the infrared sensing channel, resulting in the appearance of high-temperature signatures that are not from the occupants in the thermal image. Because such interference highly resembles human thermal signatures in spatial distribution and temperature gradient, existing recognition algorithms can easily misinterpret it as occupant movement, dozing, or abnormal body temperature, causing the system to erroneously trigger unnecessary responses such as fatigue warnings, air conditioning control, ventilation, or cabin safety mode switching. The above problems not only reduce the system's recognition accuracy and control stability, but also may trigger improper control logic due to misjudgment in autonomous driving or remote cockpit management modes, causing driving interference, passenger discomfort and even safety risks.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The application aims to provide a vehicle sensor-based intelligent cabin environment perception and control method, which realizes accurate identification and dynamic suppression of thermal light interference state by extracting infrared thermal map key features and combining a machine learning model. It can adaptively adjust the sampling and filtering strategy in complex environments such as strong light reflection, reconstruct the real thermal image contour, effectively eliminate non-human heat sources, significantly improve the identification accuracy and operation stability of the infrared thermal sensing system, reduce the misjudgment rate, and enhance the safety and reliability of the intelligent cabin, to solve the problems in the above background technology.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme: a vehicle sensor-based intelligent cabin environment perception and control method, comprising the following steps: During the operation of the intelligent cabin, the infrared thermal sensing system collects thermal imaging data in the vehicle in real time, extracts multi-dimensional feature data based on the obtained thermal imaging data, and forms an infrared thermal map feature data parameter set; The collected infrared thermal map feature data is preprocessed, multi-dimensional analysis is performed on the preprocessed data using feature engineering technology, key indicators for characterizing the thermal light interference of the infrared thermal sensing system are extracted, and the extracted key indicators are comprehensively analyzed to quantify the interference degree of the infrared thermal sensing system; The analyzed key indicators are input into a pre-trained machine learning model, the current operation state of the infrared thermal sensing system is intelligently evaluated by the model, and it is determined whether the infrared thermal sensing system is interfered by thermal light; When it is identified that the infrared thermal sensing system is interfered by thermal light, the spatial sampling matrix and the filtering window of the infrared thermal sensing system are dynamically adjusted, the interference area is redundantly sampled and high-pass-low-pass composite filtered, the heat source pixel points that do not match the spatial distribution and human body structure are dynamically eliminated in combination with the occupant activity mode of the short-time history window, and the real human body thermal image contour is reconstructed by the machine learning model to weaken the interference of external thermal light on the thermal sensing system.

[0007] Preferably, during the operation of the intelligent cabin, the infrared thermal sensing system continuously monitors the thermal imaging of the vehicle environment, and the construction of the infrared thermal map feature data parameter set specifically comprises the following steps: Real-time acquisition of the infrared thermal imaging image in the vehicle, acquisition of the original thermal map sequence of consecutive frames, and use for reflecting the temperature distribution of the occupant and the environment surface; Based on the temperature distribution in the infrared image, the heat source area in the image is identified, and the spatial position, shape contour and thermal center coordinates of each heat source are extracted; The key features of the heat source in the spatial dimension are analyzed to characterize whether it meets the characteristics of the human heat source; In combination with the heat source change trend between image frames, the features in the time dimension are extracted, and finally a complete infrared thermal map feature data parameter set is formed.

[0008] Preferably, the feature engineering technique is applied to perform multi-dimensional analysis on the preprocessed data to extract key indicators for characterizing the thermal interference of the infrared thermal imaging system, the extracted indicators include the edge gradient variation degree of the heat source and the existence continuity of the heat source in the time sequence, the edge gradient variation degree of the heat source and the existence continuity of the heat source in the time sequence are comprehensively analyzed under the detection window to generate the hot spot edge sharpening indicator and the heat source persistence abnormality indicator respectively, and the infrared thermal imaging system interference degree is quantified by the hot spot edge sharpening indicator and the heat source persistence abnormality indicator.

[0009] Preferably, the specific steps of comprehensively analyzing the edge gradient variation degree of the heat source under the detection window to generate the hot spot edge sharpening indicator are as follows: Under the detection window, an edge enhancement operator is applied to the thermal image to calculate the edge gradient amplitude value of each pixel point The edge gradient amplitude reflects the change rate of the temperature near the pixel point in the spatial direction, that is, the boundary strength feature of the thermal image, whether the edge has an abnormal sharpening phenomenon is analyzed, and the local gradient mutation factor of each pixel point is calculated, and the calculation expression is as follows: , wherein is the edge gradient amplitude value of the infrared thermal image at the pixel point, is the pixel point coordinate, is the displacement amount in the horizontal direction in the image coordinate, is the displacement amount in the vertical direction in the image coordinate, is the gradient amplitude value of the horizontally adjacent pixel point on the right side of the current pixel point in the thermal image, is the gradient amplitude value of the vertically adjacent pixel point below the current pixel point in the thermal image, is the local gradient mutation factor of the pixel point in the infrared thermal image; Based on the extracted local gradient mutation factors in all edge regions , a set of pixel points within the heat source boundary is selected, the hot spot edge sharpening indicator is calculated, and the calculation expression is as follows: , wherein is the hot spot edge sharpening indicator, is the pixel set of the heat source boundary region, is the edge direction change rate.

[0010] Preferably, the specific steps of comprehensively analyzing the existence continuity of the heat source in the time sequence under the detection window to generate the heat source persistence abnormality indicator are as follows: In the detection window, the heat source shape information in the infrared image is extracted frame by frame. For each frame, the actual contour area and the center coordinates of the heat core of the heat source are extracted, and a consistency score is calculated to quantify whether the heat source in the current frame has physiological characteristics. The consistency score calculation expression is as follows: , wherein, is the consistency score, is the contour area of the heat source in the i th frame, is the maximum contour area of the heat source in the detection window, is the area consistency weight coefficient, is the contour area of the heat source in the i th frame, is the center coordinates of the heat core in the i th frame, is the reference center coordinates of the heat core, is the heat core position fitting function, is the position offset penalty coefficient; After obtaining all the consistency scores , the proportion of the number of frames with a score lower than the confidence threshold is counted to construct a heat source persistence anomaly index. The calculation expression is as follows: , wherein, is the total number of frames of the detection window, is the lower limit threshold of the consistency score, is the heat source persistence anomaly index. Preferably, the analyzed heat spot edge sharpening index and the heat source persistence anomaly index are input into a pre-trained machine learning model to generate a heat-light interference risk coefficient. Based on the heat-light interference risk coefficient, the current running state of the infrared thermal sensing system is intelligently evaluated to determine whether the infrared thermal sensing system is interfered by heat-light interference.

[0011] Preferably, the heat-light interference risk coefficient is compared and analyzed with a pre-set heat-light interference risk coefficient reference threshold to determine whether the infrared thermal sensing system is interfered by heat-light interference. The judgment logic is as follows: If the heat-light interference risk coefficient is greater than the pre-set heat-light interference risk coefficient reference threshold, it is determined that the infrared thermal sensing system is interfered by heat-light interference. If the heat-light interference risk coefficient is less than or equal to the pre-set heat-light interference risk coefficient reference threshold, it is determined that the infrared thermal sensing system is not interfered by heat-light interference.

[0012]

[0013] ​​Preferably, when the infrared thermal sensing system is identified to be interfered by thermal light, the spatial sampling matrix and the filter window of the infrared thermal sensing system are dynamically adjusted, the disturbed area is redundantly sampled and high-pass-low-pass composite filtered; combined with the occupant activity pattern in the short-time history window, the thermal source pixel points not matching the spatial distribution and the human body structure are dynamically removed, and the real human body thermal image contour is reconstructed by the machine learning model to weaken the interference influence of external thermal light on the thermal sensing system. The specific steps are as follows: When the infrared thermal sensing system is identified to be interfered, based on the position and intensity of the interference area in the current thermal map, the spatial sampling matrix resolution and the filter window combination strategy are dynamically adjusted, the redundant sampling is performed, and the redundant response value is extracted. The redundant response value calculation expression is as follows: , wherein, is the redundant sampling response value, is the thermal map frame after adjustment of the spatial sampling, indicating that the thermal map of the first frame is obtained after re-sampling in space, is the total number of detection window thermal map frames, is the baseline reference thermal map frame, is a high-pass-low-pass composite filter function; Combined with the occupant thermal map contour data in the short-time history window, according to the matching degree of the current frame thermal source and the standard human body thermal structure template, the thermal source pixel points with abnormal spatial distribution are removed, the structure error value after removal is calculated, and the calculation expression is as follows: , wherein, is the structure removal error value, is the number of thermal source pixel points, is the temperature value of the first thermal source pixel in the current frame, is the reference temperature value corresponding to the first thermal source pixel in the history window, is a structure matching factor; Based on the output redundant response value and the structure removal error value , the thermal map reconstruction neural network model trained is input, the human body thermal image reconstruction map is generated, and the enhancement coefficient thereof is calculated. The calculation expression is as follows: , wherein, is the thermal map reconstruction enhancement coefficient, is a thermal map reconstruction enhancement factor, is a thermal light interference risk coefficient, is a thermal light interference risk coefficient reference threshold.

[0014] In the above technical solution, the technical effects and advantages provided by the present application are: The present application realizes accurate identification and self-adaptive suppression of thermal light interference state by combining dynamic perception of infrared thermal image feature data, key indicator extraction and machine learning modeling, significantly improving the stability and reliability of the infrared thermal sensing system in complex light and heat environment. This method not only can identify non-physiological heat sources under strong sunlight, aging film or reflection interference and other non-ideal conditions, but also can automatically adjust the perception parameters, optimize the image sampling and filtering strategy after determining the interference, and reconstruct the real passenger thermal profile combined with the historical thermal image and human structure model. Overall, it effectively reduces the risk of passenger state misidentification and system false triggering, enhances the safety protection and user experience of intelligent cabin in key scenarios such as automatic driving and human factor monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0016] Figure 1 The method flow chart of the intelligent cabin environment perception and control method based on vehicle sensors of the present application. DETAILED DESCRIPTION

[0017] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the scope of example implementations to those skilled in the art.

[0018] The present application provides an intelligent cabin environment perception and control method based on vehicle sensors as shown in Figure 1 The method flow chart of the intelligent cabin environment perception and control method based on vehicle sensors of the present application. During the operation of the intelligent cabin, the infrared thermal sensing system acquires real-time thermal imaging data in the vehicle, extracts multi-dimensional feature data based on the acquired thermal imaging data, and forms an infrared thermal image feature data parameter set; In the process of intelligent cockpit operation, the infrared thermal sensing system continuously performs thermal imaging scanning on the in-vehicle environment to generate a sequence of images reflecting temperature distribution. Based on these thermal imaging images, the system extracts multi-dimensional image features including spatial distribution, temperature gradient, time sequence change, and regional stability, and forms a unified structured infrared thermal image feature data parameter set. These parameter data are used to reflect the physiological state of the occupant (such as abnormal body temperature, stillness / shaking) and behavior characteristics (such as dozing off, leaving the seat), and can also be used to identify whether there is an interference signal (such as a non-human structure heat source). Commonly extracted features include: heat source shape contour, heat center coordinates, temperature peak and average value, temperature change slope, regional heat kernel stability, symmetry score, and matching degree with human body template, etc. By constructing this parameter set, the system can achieve accurate identification of the occupant's state and high robustness judgment of abnormal heat sources, which is the basic data support for subsequent interference identification and dynamic control.

[0019] In the process of intelligent cockpit operation, the infrared thermal sensing system continuously performs thermal imaging scanning on the in-vehicle environment to generate a sequence of images reflecting temperature distribution. Based on these thermal imaging images, the system extracts multi-dimensional image features including spatial distribution, temperature gradient, time sequence change, and regional stability, and forms a unified structured infrared thermal image feature data parameter set. These parameter data are used to reflect the physiological state of the occupant (such as abnormal body temperature, stillness / shaking) and behavior characteristics (such as dozing off, leaving the seat), and can also be used to identify whether there is an interference signal (such as a non-human structure heat source). Commonly extracted features include: heat source shape contour, heat center coordinates, temperature peak and average value, temperature change slope, regional heat kernel stability, symmetry score, and matching degree with human body template, etc. By constructing this parameter set, the system can achieve accurate identification of the occupant's state and high robustness judgment of abnormal heat sources, which is the basic data support for subsequent interference identification and dynamic control.

[0020] The collected infrared thermal image feature data is preprocessed, and feature engineering techniques are applied to perform multi-dimensional analysis on the preprocessed data to extract key indicators for characterizing the thermal interference of the infrared thermal sensing system. The extracted key indicators are comprehensively analyzed to quantify the degree of interference of the infrared thermal sensing system. The collected heat map feature data is preprocessed, including the following steps: first, the temperature values of each pixel point in the heat map are standardized, and the temperature range of different frames of images is mapped to a unified scale interval (such as 0-1), so as to eliminate the inconsistency of the amplitude caused by the external environmental temperature difference; second, the isolated hot spots and random noise caused by sensor jitter, electrical signal disturbance or reflective interference are removed through median filtering, Gaussian filtering and other image denoising algorithms; third, the size of all images is normalized, and the heat maps under different resolutions or sampling angles are mapped to standard image size and coordinate system, so as to ensure the comparability of different frames of data in space; finally, the temperature of the background temperature area in the image is suppressed, and the contrast of the heat features of the occupant related area is enhanced, so as to improve the stability and accuracy of subsequent feature extraction and interference judgment. The overall role of the preprocessing process is to improve the structural consistency and information quality of the infrared heat map data, and to provide clean and standardized input basis for subsequent modeling and analysis.

[0021] The preprocessed data is subjected to multi-dimensional analysis by applying feature engineering technology, and key indicators for characterizing the thermal light interference received by the infrared thermal sensing system are extracted. The extracted indicators include the edge gradient change degree of the heat source and the existence continuity of the heat source in the time sequence. The edge gradient change degree of the heat source and the existence continuity of the heat source in the time sequence are analyzed under the detection window to generate the heat spot edge sharpening indicator and the heat source persistence abnormality indicator, respectively. The degree of interference received by the infrared thermal sensing system is quantified by the heat spot edge sharpening indicator and the heat source persistence abnormality indicator.

[0022] When the infrared thermal sensing system detects that the edge gradient of the heat source presents an abnormal mutation, it usually indicates that the system is interfered by thermal light. The reason is that the infrared thermal image of a normal occupant has typical physiological heat conduction characteristics, and the temperature transition from the core area to the edge is usually gradual, with a warm and continuous boundary profile. When the system captures a heat source boundary that presents a significant high gradient jump in the image, i.e. instantaneously transitions from a low temperature area to a high temperature area, forming a "hard boundary" or "sharpened edge", this is usually not a feature that can be produced by a physiological heat source. Such edge abnormalities are usually caused by external thermal light interference, such as high-intensity heat spots formed by direct sunlight or window reflection. These heat spots project a significantly abrupt and irregular temperature edge in the image, which does not conform to the natural diffusion pattern of the occupant's heat source. Especially when the vehicle is moving or the light angle is changing, such abnormal boundaries will frequently appear or move, further exacerbating the misdirection of the thermal sensing system. Therefore, the abnormal mutation of the edge gradient of the heat source is one of the important criteria for the manifestation of thermal light interference in the infrared heat map, and can be used as a key spatial feature indicator in the interference recognition model.

[0023] The specific steps of generating the heat spot edge sharpening indicator by comprehensively analyzing the edge gradient change degree of the heat source under the detection window are as follows: Under the detection window, apply the edge enhancement operator (such as Sobel operator or Laplacian operator) to the thermal image to calculate the edge gradient amplitude value of each pixel , the edge gradient amplitude reflects the rate of change of temperature near the pixel in the spatial direction, that is, the boundary intensity feature of the heat map. Analyze whether there is abnormal sharpening at the edge and calculate the local gradient mutation factor of each pixel. The calculation expression is as follows: , where is the edge gradient amplitude value of the infrared thermal image at the pixel point, is the pixel coordinate, It is the horizontal displacement in the image coordinates, usually taken as 1 pixel unit, that is, adjacent horizontal pixels. It is the vertical displacement in the image coordinates, usually taken as 1 pixel unit, that is, adjacent vertical pixels. It is the gradient amplitude value of the horizontally adjacent pixel to the right of the current pixel in the heat map. It is the gradient amplitude value of the adjacent pixel points in the vertical direction below the current pixel point in the heat map. It is used to limit the gradient amplitude in the denominator to prevent the denominator from being too small and causing meaningless amplification of the result. When , the denominator is fixed to 1, and the control mutation value is not weakened; when , using actual gradients to enhance the ability to distinguish low-temperature regions, In the infrared thermal image, the pixel The local gradient mutation factor measures whether the edge gradient around the pixel has experienced a sharp jump in the horizontal and vertical directions; The purpose of this step is to strengthen those areas with the strongest gradient mutations and the fastest reversal of change directions, peel off smooth thermal edges, and highlight the local features of the sharp edges of hot spots. It has the characteristics of direction independence and amplitude nonlinear enhancement, and is an explicit manifestation of thermal-optical interference characteristics in the image.

[0024] Based on the local gradient mutation factor extracted from all edge regions , select the pixel point set within the heat source boundary and calculate the hot spot edge sharpening index. The calculation expression is as follows: , where It is an indicator of the sharpness of the edge of the hot spot, indicating whether there is a high-intensity, abrupt thermal boundary phenomenon in the current thermal image. The larger the value, the sharper and more unnatural the edge of the heat source in the image is, and the more likely it is a "pseudo heat source" caused by external thermal light interference; if the value is small, it indicates that the heat source boundary is smooth and natural, which usually belongs to physiological human thermal images. is the pixel set in the heat source boundary area, is the edge direction change rate, indicating the change speed of the edge profile direction at the pixel point , i.e. the degree of shift of the edge direction (towards) between adjacent pixels.

[0025] The purpose of this step is to fuse the strong gradient mutation of the heat source boundary and the shape abruptness, and to construct an index that combines the local gradient transition energy and the edge geometric abnormality. The larger the value, the more prominent the non-natural boundary structure in the infrared image, which is most likely to be caused by thermal light interference (such as sharp angle heat spots formed by glass reflection).

[0026] The degree of change of the heat source edge gradient is analyzed under the detection window to generate a heat spot edge sharpening index. The metric value generated by analyzing the degree of change of the temperature gradient of the heat source edge reflects the smoothness and naturalness of the heat source profile. When the index value is large, it indicates that there is a significant high gradient mutation in the heat source boundary, i.e. a sharp jump in temperature within a very short spatial range. This phenomenon is usually caused by external thermal light interference, such as sunlight, metal reflection, or car window heat focusing, which will form a non-physiological "hard boundary" heat spot in the thermal image. In contrast, the edge transition of a human heat source is usually gentle and continuous, and does not exhibit sharp edge sharpening. Therefore, when the heat spot edge sharpening index is small, it can be considered that the heat source in the image has natural physiological characteristics, and the infrared thermal sensing system has not been significantly disturbed.

[0027] The heat source appears irregularly and explosively in time series, which is an important criterion for indicating that the infrared thermal sensing system is disturbed by thermal light. Normal occupant heat sources have physiological persistence and behavioral rhythm, and their thermal images show relatively stable thermal intensity, thermal center position, and profile shape in consecutive time frames. Even if there is slight shaking or posture adjustment, the thermal image changes smoothly and continuously. Thermal light interference sources (such as sunlight reflection, metal heat focusing, or moving heat reflection spots in the environment) often form transient high-temperature regions at the car window or trim panel. This interference heat source usually has no stable shape, appears for a short time, and is distributed randomly, appearing as "heat spot flashes" or "heat source jumps" in the image sequence, and does not have the time consistency and position predictability of normal occupant thermal images. Therefore, when the infrared thermal sensing system detects that the heat source lacks continuity in the time dimension or appears explosively in a short time, and cannot be matched with the occupant behavior pattern, it can be highly suspected that the heat source is an external thermal light interference signal. This feature can be used as a key judgment basis to activate the anti-interference mechanism of the system or reduce the participation weight of the region in occupant behavior recognition.

[0028] The existence of the heat source in the time sequence is analyzed under the detection window to generate a heat source persistence abnormality index. The specific steps are as follows: ​Within the detection window, the heat source morphological information in the infrared image is extracted frame by frame. For each frame, the actual contour area and heat core center coordinates of the heat source are extracted, and the consistency score is calculated to quantify whether the heat source in the current frame has physiological characteristics. The consistency score calculation expression is as follows: , where It is a consistency score that reflects whether the heat source in the current frame has the spatial continuity and structural rationality of a typical occupant thermal image. The value range is 0, 1. The higher the value, the higher the credibility of the heat source in the frame. It is The outline area of ​​the frame heat source, is the maximum contour area of ​​the heat source within the detection window, It is the area consistency weight coefficient, which indicates the weight of the heat source outline area stability in the score. It is used to measure the relative ratio between the current frame heat source area and the historical maximum heat source area. The value range is 0.6-0.8. It is Frame hot nucleus center coordinates, are the coordinates of the center of the reference hot nucleus, It is the heat kernel position fitting function, used to determine the heat kernel center of the current frame With reference center position The degree of deviation, is the position offset penalty coefficient, which is used to indicate the importance of heat source position stability (whether the heat core center is stable) in the score. Used together, this function determines whether the heat source center offset exceeds the set threshold. If the offset is large, then , multiplied by The total score will be penalized and the credibility of the frame will be reduced; The heat kernel position fitting function is a discriminant function used to determine whether the center position of the heat source in the current frame is consistent with the reference position. It is mainly used to measure the spatial stability of the heat source in the time series. In infrared thermal image analysis, the position of the heat kernel (i.e., the center of mass of the heat source) of the real occupant in the image is relatively stable, and there will be no drastic jumps even with slight movements. However, thermal interference such as sunlight reflection or dynamic hot spots often causes the heat kernel to mutate or drift significantly between different frames. The heat kernel position fitting function is usually designed as a binary logic function (i.e., an indicator function). By calculating the current heat kernel With the reference heat kernel Euclidean distance, if the distance is within the preset threshold, the function takes the value of 0 (indicating good fit), if it exceeds the threshold, it takes the value of 1 (indicating obvious position deviation). The function is not a function in the general mathematical function library, but a commonly used custom spatial consistency detection function in the application, which is widely used in image sequence analysis, target tracking and behavior recognition. Its core function is to assist in identifying abnormal heat source jump phenomenon, and it is a key logic component for judging "position reliability" in heat source consistency scoring.

[0029] The scoring design emphasizes the spatial consistency and position stability of the heat source form. If the heat source appears instantaneously, the form is abnormal or drifts seriously in a frame, it will be given a lower value. The entire detection window will get a sequence of scores as the basis for subsequent calculation.

[0030] After obtaining all the consistency scores , the frame number ratio of the scores below the reliable threshold is calculated to construct the heat source persistence anomaly index, and the calculation expression is as follows: , in the formula, is the total number of frames in the detection window, is the lower threshold of the consistency score, which is the score critical value set for judging whether the frame is reliable, is the heat source persistence anomaly index, the value range is: 0, 1, the higher the value, the higher the time coverage of "unreliable heat source" in the system.

[0031] The final index represents the proportion of unreliable heat source frames detected by the system in the entire detection period. When tends to 1, it means that there are a large number of frames with heat source discontinuity, structure distortion or position drift in the detection window, and the system is most likely to be under the influence of thermal light interference; on the contrary, when approaches 0, it means that the heat source shows high consistency and stability in time, which can be identified as a normal passenger heat source. This method has the advantages of accurate identification, clear logic and light calculation, and is especially suitable for heat map stability judgment in high dynamic thermal environment in intelligent cockpit.

[0032] The continuity of the heat source in the time sequence is analyzed in the detection window to generate a heat source persistence anomaly index. By analyzing the frequency of the heat source appearing in the continuous image frames, the position stability, the temperature change amplitude, and other time dimension characteristics, it can be determined whether the heat source exhibits the persistence and consistency of a normal passenger thermal image. If the heat source presents intermittent appearance, rapid flashing, obvious position drift, and other abnormal phenomena on the time axis, the value of the index will increase significantly, indicating that its behavior does not conform to the persistent and stable characteristics of human heat sources, and it is most likely a false heat source caused by sunlight reflection, thermal fluctuations, or other external interference. Therefore, the greater the performance value of the heat source persistence anomaly index, the more it indicates that the infrared thermal sensing system is currently being disturbed by thermal light; on the contrary, if the index value is low, it means that the heat source identified by the system is continuous in time, stable in shape, and has high reliability, and it can be basically determined that the infrared thermal sensing system is not disturbed.

[0033] The analyzed key indicators are input into a pre-trained machine learning model (such as random forest, support vector machine, or lightweight neural network), and the current running state of the infrared thermal sensing system is intelligently evaluated by the model to determine whether the infrared thermal sensing system is disturbed by thermal light. The analyzed heat spot edge sharpening index and heat source persistence anomaly index are input into a pre-trained machine learning model (such as random forest, support vector machine, or lightweight neural network), and a thermal light interference risk coefficient is generated by the model. The current running state of the infrared thermal sensing system is intelligently evaluated based on the thermal light interference risk coefficient to determine whether the infrared thermal sensing system is disturbed by thermal light.

[0034] In this scheme, the "pre-trained machine learning model" refers to an intelligent recognition algorithm obtained by learning and optimizing a large number of labeled sample data. Its core function is to receive the interference feature indicators (such as heat spot edge sharpening index and heat source persistence anomaly index) extracted from the infrared thermal image as input, and make a probabilistic judgment or risk score output on whether the current image data has thermal light interference. "Pre-training" means that before the system is formally deployed, a large number of thermal image samples containing normal and interference situations are used for model training, so that the model learns to recognize the interference state labels corresponding to different feature patterns. During the training process, the model compares and learns the input feature vectors (such as edge gradient, heat source time consistency, shape symmetry, etc.) with the actual labeled system state (interference / non-interference), and gradually optimizes its internal weight structure, forming a classification or regression model with discrimination ability. Common models include random forest (which can handle complex relationships between features and is robust to outliers), support vector machine (SVM) (which has good boundary division ability when the feature distribution is nonlinear), and lightweight neural network (such as MLP or TinyCNN) (which is suitable for embedded platforms and has strong generalization ability).

[0035] Through such a pre-training process, the model can not only extract the rules between explicit features, but also implicitly learn some non-linear interference patterns that traditional rules are difficult to capture, such as complex behaviors caused by the superposition of multiple factors, such as edge and time feature collaborative anomalies. When the system is running, the hot spot edge sharpening index and heat source persistence anomaly index calculated in the current detection window are input into the model, and the model can output a thermal light interference risk coefficient (such as a continuous value between 0-1 or an interference level label) according to the mapping relationship obtained by training, indicating the possibility or severity of thermal light interference on the infrared thermal sensing system. The system intelligently evaluates its own running state accordingly: if the risk coefficient is low, the system is determined to be in a stable sensing state and normal function is maintained; if the risk coefficient is high, the infrared thermal image has been seriously contaminated, and the anti-interference processing procedure needs to be triggered immediately, such as dynamically adjusting the spatial sampling, filtering parameters, or performing thermal image reconstruction compensation. This intelligent judgment method based on machine learning models can significantly improve the identification accuracy and response efficiency of the system to thermal sensing interference in complex environments, avoid misjudgment caused by single indicator fluctuations, and effectively enhance the robustness and intelligence level of the cabin sensing system.

[0036] The thermal light interference risk coefficient is compared and analyzed with the pre-set thermal light interference risk coefficient reference threshold to determine whether the infrared thermal sensing system is interfered by thermal light. The judgment logic is as follows: If the thermal light interference risk coefficient is greater than the pre-set thermal light interference risk coefficient reference threshold, it is determined that the infrared thermal sensing system is interfered by thermal light; if the thermal light interference risk coefficient is less than or equal to the pre-set thermal light interference risk coefficient reference threshold, it is determined that the infrared thermal sensing system is not interfered by thermal light.

[0037] When it is identified that the infrared thermal sensing system is interfered by thermal light, the spatial sampling matrix and the filtering window of the infrared thermal sensing system are dynamically adjusted, and the disturbed area is redundantly sampled and high-pass-low-pass composite filtered; combined with the activity pattern of the occupant in the short-time historical window, the heat source pixel points that do not match the spatial distribution and human body structure are dynamically removed, and the real human body thermal image contour is reconstructed by the machine learning model to weaken the interference influence of external thermal light on the thermal sensing system; When it is identified that the infrared thermal sensing system is subject to thermal and optical interference, the core role of this series of dynamic response steps executed by the system is to actively suppress the influence of the interference source, restore the accuracy of infrared perception, and ensure that the system can still stably and accurately identify the true status of the occupants in a complex optical and thermal environment. Specifically, the spatial sampling matrix and filter window of the infrared thermal sensing system are first adjusted dynamically to improve the thermal map resolution and signal purification capabilities of the interference area. After identifying the area where the abnormal heat source is concentrated (such as near the window or dashboard), the system will increase the sampling density of the area and flexibly reduce the sampling grid unit to obtain more detailed temperature gradient and heat source morphology information. At the same time, a composite processing method of high-pass and low-pass filtering is applied to remove the slowly changing thermal drift background and the high-frequency sudden change of thermal reflection spike signals, respectively, to enhance the image cleanliness and stability from the frequency domain level.

[0038] After completing basic sampling optimization, the system further combines occupant thermal image data from a short-term historical window to perform comparative analysis of heat source pixels in the current frame, dynamically identifying heat source areas that do not conform to the human body structure. For example, if a high-temperature area does not appear in previous frames and its position, shape, and symmetry significantly deviate from the historical thermal image template, it can be identified as an external thermal-optical disturbance signal. Such heat sources are eliminated in real time to prevent them from participating in the occupant status recognition process, thereby reducing false alarms and the risk of miscontrol. Finally, to compensate for missing thermal image information due to interference or elimination, the system invokes a machine learning-based image reconstruction model (such as a convolutional neural network or a body structure thermal image prediction network). This model intelligently reconstructs the thermal image using historical image data and prior knowledge of the human body's thermal image, restoring a clear and continuous occupant thermal image outline. This reconstruction not only helps maintain the integrity of occupant behavior recognition but also improves the ability to track physiological signals such as fatigue, abnormal body temperature, or hypoxia, ensuring the high reliability and practical value of the infrared thermal sensing system even in interference environments. Therefore, the overall purpose of this step is to realize a multi-level compensation and correction mechanism from the perception layer to the algorithm layer, significantly reducing the system uncertainty caused by external thermal and optical interference, and ensuring the continuous and stable operation of the occupant status recognition and safety control functions in the intelligent cockpit.

[0039] When the infrared thermal sensing system is identified as being interfered with by thermal light, the system's spatial sampling matrix and filter window are dynamically adjusted, and redundant sampling and high-pass and low-pass composite filtering are performed on the interfered area. Combined with the occupant's activity patterns within a short-term historical window, heat source pixels whose spatial distribution does not conform to the human body structure are dynamically eliminated, and a machine learning model is used to reconstruct the true human thermal image contour. The specific steps to reduce the interference of external thermal light on the thermal sensing system are as follows: When the infrared thermal sensing system is identified as being interfered with, the spatial sampling matrix resolution and filter window combination strategy are dynamically adjusted based on the position and intensity of the interference area in the current heat map, redundant sampling is performed, and redundant response values ​​are extracted. The redundant response value calculation expression is as follows: , where It is the redundant sampling response value, which indicates the average temperature response deviation between the reference image and the current thermal image after enhanced sampling and filtering in the interference area. It is used to quantify the degree of deviation between the current thermal image and the historical stable thermal image after processing. The higher the value, the more significant the temperature anomaly is in the current frame and the greater the influence of the interference. On the contrary, it indicates that the thermal image is close to a stable state after processing. It is the heat map frame under the adjusted spatial sampling, which refers to the The heatmap obtained after the frame heatmap is spatially resampled (reconstructed resolution), is the total number of detection window heatmap frames, is the baseline reference heat map frame, indicating the difference between the The historical “clean frame” or benchmark heatmap that the frame heatmap is closest to in time or space, It is a high-pass-low-pass composite filter function, which consists of a high-pass filter (used to remove slowly changing background) and a low-pass filter (used to suppress high-frequency spikes); Represents a high-pass-low-pass composite filter function, which is a combined signal processing function that combines the processing logic of a high-pass filter (High-PassFilter) and a low-pass filter (Low-PassFilter). It is often used for multi-band purification of interference signals in images, thermal maps, or time series data. Its basic principle is: the low-pass filter is used to remove high-frequency noise or sharp hot spot edge interference in the image to make the image smoother; while the high-pass filter is used to eliminate low-frequency interference caused by slowly changing background (such as long-term heat accumulation or infrared thermal drift), thereby retaining local mutation characteristics. When used in combination, It can be applied to thermal map data through cascading, weighting, or convolution, achieving simultaneous suppression of high- and low-frequency interference while preserving mid-frequency features related to human anatomy. This function is not a fixed mathematical function, but rather a customizable combined filter framework. Its structure and parameters can be dynamically configured based on the infrared thermal sensing system's sampling frequency, image noise model, and interference characteristics. It is widely used in tasks such as thermal map enhancement, image stabilization, and interference removal, and exhibits excellent interpretability and engineering adaptability.

[0040] This step generates redundant response values , which is used to measure the purification effect of interference signals in the heat map after filtering and serves as the input basis for subsequent elimination and reconstruction modules.

[0041] Combined with the occupant thermal map contour data within the short-term history window, according to the matching degree between the current frame heat source and the standard human thermal structure template, the heat source pixels with abnormal spatial distribution are eliminated, and the structural error value after elimination is calculated. The calculation expression is as follows: , where is the structural elimination error value, which represents the total error intensity between the thermal map and the historical reference map at the structural level after eliminating thermal and optical interference. is the number of heat source pixels, that is, the total number of all pixels considered to belong to the heat source area, The current frame The temperature value of each heat source pixel represents the observation data in the current infrared image and is a direct manifestation of the possible impact of the interfering heat source. It is the first The reference temperature value corresponding to the heat source pixel, It is a structural matching factor, which is used to measure the matching degree between the current heat source pixel and the standard structure template of the human body in terms of spatial position, contour shape, heat core connectivity, etc. ; This step outputs the structure elimination error value , which is used to reflect the degree of recovery of the overall temperature structure of the current heat map after removing the interfering heat source. If the value is too high, it indicates that abnormal signals still remain in the interfered area.

[0042] Redundant response values ​​based on output and structure elimination error , input to the trained heat map reconstruction neural network model In , the human body thermal image reconstruction image is generated and its enhancement coefficient is calculated. The calculation expression is as follows: , where It is the thermal image reconstruction enhancement coefficient, which is used to measure the degree of consistency between the output thermal image of the infrared thermal sensing system and the standard human thermal image template after reconstruction. is the heatmap reconstruction enhancement factor, which represents a trained machine learning model (such as lightweight convolutional neural network, variational autoencoder, structural regression network, etc.) with redundant sampling response value and structure elimination error As input, it outputs a normalized reconstruction quality factor with a value range of (0,1]. The larger the value, the stronger the model's ability to restore the current frame's thermal image. If it is 1, it means that the system output thermal image quality is close to the ideal human body thermal image. is the thermal-optical interference risk factor, It is the reference threshold of the thermal-optical interference risk factor.

[0043] The specific steps of the process of generating the human body thermal image reconstruction are as follows: First, extract the thermal image feature input after the thermal light interference processing from the current frame, including the redundant response value , structure elimination error value and optional historical heat map sequence as context information; secondly, input the above-mentioned multi-source features into a pre-trained heat map reconstruction neural network model The model extracts local spatial features in the heat map through a convolution layer, and captures heat source distribution rules through an encoding-decoding structure; then, the model fills and reconstructs the current distorted region according to the structure template and the historical heat source trajectory at the decoding end, to generate a continuous and physiologically reasonable thermal image output graph; finally, post-processing operations are performed on the reconstructed image, including dynamic range normalization, contrast enhancement and morphological boundary optimization, to ensure that the output human thermal image graph has stable temperature levels and contour integrity, thereby providing clear and reliable heat map input for occupant behavior recognition and state perception.

[0044] Reconstruction enhancement coefficient The method is used for dynamically evaluating the interference repair quality of the current infrared image, and can be used as a criterion for whether the thermal sensing system continues to sample and correct or cancels the interference mode, so that intelligent self-recovery control of the heat map in the interference state is realized.

[0045] The present application realizes accurate identification and adaptive suppression of the thermal light interference state by combining dynamic perception, key index extraction and machine learning modeling of infrared heat map feature data, which significantly improves the stability and reliability of the infrared thermal sensing system in complex light and heat environments. The method not only can identify non-physiological heat sources under strong sunlight, aged film or reflection interference and other non-ideal conditions, but also can automatically adjust the perception parameters, optimize the image sampling and filtering strategy after determining the interference, and reconstruct the real occupant heat contour combined with the historical thermal image and the human structure model. Overall, the risk of occupant state misidentification and system false triggering is effectively reduced, and the safety protection and user experience of the intelligent cabin in the key scenarios of autonomous driving and human factor monitoring are enhanced.

[0046] The above formulas are dimensionless numerical calculations, the formulas are obtained by software simulation of a large amount of data to obtain a formula of the nearest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0047] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above drawings and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

[0048] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0049] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0050] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0051] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0052] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0054] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0055] The foregoing merely describes certain exemplary embodiments of this application by way of illustration. Obviously, for those skilled in the art, modifications and changes can be made to the described embodiments in various ways without departing from the spirit and scope of the application. Therefore, the above drawings and descriptions are illustrative in nature, and should not be construed as limiting the scope of the claims of the present application.

Claims

1. Intelligent cockpit environment perception and control method based on vehicle sensors, characterized in that: The following steps are involved: During the operation of the smart cockpit, the infrared thermal sensing system collects real-time thermal imaging data from the vehicle, extracts multi-dimensional feature data based on the acquired thermal imaging data, and forms an infrared thermal map feature data parameter set; Preprocess the collected infrared thermal image feature data, apply feature engineering technology to perform multidimensional analysis on the preprocessed data, extract key indicators used to characterize the thermal and optical interference of the infrared thermal sensing system, conduct a comprehensive analysis of the extracted key indicators, and quantify the degree of interference to the infrared thermal sensing system; The analyzed key indicators are input into a pre-trained machine learning model, which then performs an intelligent assessment of the current operating status of the infrared thermal sensing system to determine whether the infrared thermal sensing system is subject to thermal light interference. When it is identified that the infrared thermal sensing system is interfered with by thermal light, the spatial sampling matrix and filter window of the infrared thermal sensing system are dynamically adjusted, and redundant sampling and high-pass-low-pass composite filtering are performed on the interfered area; combined with the occupant activity pattern in the short-term historical window, heat source pixels whose spatial distribution does not match the human body structure are dynamically eliminated, and the real human body thermal image contour is reconstructed using a machine learning model to weaken the interference of external thermal light on the thermal sensing system.

2. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 1 is characterized in that: During the operation of the smart cockpit, the infrared thermal sensing system continuously monitors the interior environment through thermal imaging. The specific steps for constructing the infrared thermal image feature data parameter set are as follows: Real-time acquisition of infrared thermal imaging images inside the vehicle, obtaining a sequence of original thermal images of continuous frames to reflect the temperature distribution of occupants and environmental surfaces; Based on the temperature distribution in the infrared image, the heat source area in the image is identified, and the spatial position, shape outline and thermal center coordinates of each heat source are extracted; Analyze the key characteristics of the heat source in the spatial dimension to characterize whether it conforms to the heat source characteristics of the human body; Combined with the heat source change trend between image frames, the features in the time dimension are extracted to finally form a complete infrared thermal image feature data parameter set.

3. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 1 is characterized in that: Feature engineering technology is applied to perform multidimensional analysis on the preprocessed data to extract key indicators for characterizing the thermal light interference of the infrared thermal sensing system. The extracted indicators include the degree of change of the heat source edge gradient and the continuity of the heat source in the time series. The degree of change of the heat source edge gradient and the continuity of the heat source in the time series are comprehensively analyzed under the detection window to generate the hot spot edge sharpness index and the heat source persistence anomaly index respectively. The hot spot edge sharpness index and the heat source persistence anomaly index are used to quantify the degree of interference of the infrared thermal sensing system.

4. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 3 is characterized in that: The specific steps for comprehensively analyzing the degree of gradient change of the heat source edge under the detection window to generate the hot spot edge sharpness index are as follows: Under the detection window, the edge enhancement operator is applied to the thermal image to calculate the edge gradient amplitude value of each pixel. , the edge gradient amplitude reflects the rate of change of temperature near the pixel in the spatial direction, that is, the boundary intensity feature of the heat map. Analyze whether there is abnormal sharpening at the edge and calculate the local gradient mutation factor of each pixel. The calculation expression is as follows: , where is the edge gradient amplitude value of the infrared thermal image at the pixel point, is the pixel coordinate, is the horizontal displacement in image coordinates, is the vertical displacement in image coordinates, It is the gradient amplitude value of the horizontally adjacent pixel to the right of the current pixel in the heat map. It is the gradient amplitude value of the adjacent pixel points in the vertical direction below the current pixel point in the heat map. In the infrared thermal image, the pixel The local gradient mutation factor; Based on the local gradient mutation factor extracted from all edge regions , select the pixel point set within the heat source boundary and calculate the hot spot edge sharpening index. The calculation expression is as follows: , where is the hot spot edge sharpening indicator, is the pixel set in the heat source boundary area, is the rate of change of edge direction.

5. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 3 is characterized in that: The specific steps for comprehensively analyzing the continuity of the heat source in the time series within the detection window to generate the heat source persistence anomaly index are as follows: Within the detection window, the heat source morphological information in the infrared image is extracted frame by frame. For each frame, the actual contour area and heat core center coordinates of the heat source are extracted, and the consistency score is calculated to quantify whether the heat source in the current frame has physiological characteristics. The consistency score calculation expression is as follows: , where is the consistency score, It is The outline area of ​​the frame heat source, is the maximum contour area of ​​the heat source within the detection window, is the area consistency weight coefficient, It is Frame hot nucleus center coordinates, are the coordinates of the center of the reference hot nucleus, is the thermal core position fitting function, is the position offset penalty coefficient; Get all consistency scores After that, the statistical score is lower than the credible threshold The frame ratio is used to construct the heat source persistence anomaly index. The calculation expression is as follows: , where is the total number of frames in the detection window, is the lower limit threshold of the consistency score, It is an indicator of persistent abnormal heat source.

6. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 3 is characterized in that: The analyzed hot spot edge sharpness index and heat source persistence abnormality index are input into a pre-trained machine learning model, and the thermal-light interference risk coefficient is generated through the model. Based on the thermal-light interference risk coefficient, an intelligent evaluation is performed on the current operating status of the infrared thermal sensing system to determine whether the infrared thermal sensing system is subject to thermal-light interference.

7. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 6 is characterized in that: The thermal-light interference risk coefficient is compared with the pre-set thermal-light interference risk coefficient reference threshold to determine whether the infrared thermal sensing system is subject to thermal-light interference. The judgment logic is as follows: If the thermal-light interference risk coefficient is greater than a preset thermal-light interference risk coefficient reference threshold, it is determined that the infrared thermal sensing system is subject to thermal-light interference; If the thermal-light interference risk coefficient is less than or equal to a preset thermal-light interference risk coefficient reference threshold, it is determined that the infrared thermal sensing system is not subject to thermal-light interference.

8. The intelligent cockpit environment perception and control method based on vehicle sensors according to claim 7 is characterized in that: When the infrared thermal sensing system is identified as being interfered with by thermal light, the system's spatial sampling matrix and filter window are dynamically adjusted, and redundant sampling and high-pass and low-pass composite filtering are performed on the interfered area. Combined with the occupant's activity patterns within a short-term historical window, heat source pixels whose spatial distribution does not conform to the human body structure are dynamically eliminated, and a machine learning model is used to reconstruct the true human thermal image contour. The specific steps to reduce the interference of external thermal light on the thermal sensing system are as follows: When the infrared thermal sensing system is identified as being interfered with, the spatial sampling matrix resolution and filter window combination strategy are dynamically adjusted based on the position and intensity of the interference area in the current heat map, redundant sampling is performed, and redundant response values ​​are extracted. The redundant response value calculation expression is as follows: , where is the redundant sampling response value, It is the heat map frame under the adjusted spatial sampling, which refers to the The heatmap obtained by spatially resampling the frame heatmap, is the total number of detection window heatmap frames, is the baseline reference heatmap frame, It is a high-pass-low-pass composite filter function; Combined with the occupant thermal map contour data within the short-term history window, according to the matching degree between the current frame heat source and the standard human thermal structure template, the heat source pixels with abnormal spatial distribution are eliminated, and the structural error value after elimination is calculated. The calculation expression is as follows: , where is the structural elimination error value, is the number of heat source pixels, The current frame The temperature value of each heat source pixel, It is the first The reference temperature value corresponding to the heat source pixel, is the structural matching factor; Redundant response values ​​based on output and structure elimination error , input to the trained heat map reconstruction neural network model In , the human body thermal image reconstruction image is generated and its enhancement coefficient is calculated. The calculation expression is as follows: , where is the heatmap reconstruction enhancement coefficient, is the heatmap reconstruction enhancement factor, is the thermal-optical interference risk factor, It is the reference threshold of the thermal-optical interference risk factor.

Images