Vehicle internal state automatic monitoring system based on multispectral polarization window-penetrating technology

Abstract

The invention provides a vehicle internal state automatic monitoring system based on a multispectral polarization window-penetrating technology, and aims to solve the problems that in the prior art, a vehicle window solar film enables the outside to not see the condition in a vehicle, and whether members in the vehicle exceed a standard load or are in other dangerous driving states cannot be judged. The system comprises a light source emitting unit, a multi-modal image acquisition unit, a real-time image processing unit and an overcrowding detection unit, a multi-spectral element penetrates through automobile window glass and a solar film, image data of different modes are collected through a multi-mode camera, the obtained images are processed, and a perspective image of an automobile window is generated; and identifying figures in the perspective image by using a passenger detection algorithm, and calculating the number and state of people in the vehicle. When it is detected that the number of passengers in the vehicle exceeds the legal or safe state limit, the overload alarm unit automatically triggers an alarm. According to the transparent window monitoring system, the defects of a traditional manual patrol method are effectively overcome, and the scientificity, efficiency and accuracy of traffic management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and particularly to an automatic vehicle interior state monitoring system based on multi-spectral polarization through-window technology. Background Art

[0002] According to the statistics of the Ministry of Public Security, as of the end of June 2024, the national automobile ownership has reached 345 million vehicles, and the number of automobile drivers is as high as 496 million. Facing such a huge number of automobile ownership and driving population, the management pressure on the public security traffic management department is enormous. In order to improve work efficiency and provide a safer and more convenient travel environment for the public, the introduction of technological means is particularly important. When monitoring and collecting evidence of the interior of a vehicle, the traffic management department has encountered many obstacles. The primary problem is that the dark car film coated on the window or sunlight reflection will cause the conventional monitoring camera to fail. This means that even if there are violations such as overloading in the vehicle, the monitoring equipment may not be able to capture it in time, creating a'monitoring blind spot' and posing a serious threat to traffic safety. Some countries and regions have legal orders prohibiting vehicle window glass from covering solar films in order to always understand the interior conditions of the vehicle and report to the police immediately in case of emergencies.

[0003] On the other hand, the widespread use of one-way perspective solar films on automobile glass, while ensuring the privacy rights of passengers, has also brought new challenges to the law enforcement of traffic management personnel. The unique design of this film makes it difficult for traffic management personnel to directly observe the interior of the vehicle. Especially during peak traffic hours or long-distance driving, it is often difficult to accurately judge the actual number of passengers inside the vehicle.

[0004] Although pasting solar films on vehicles has its advantages, if used improperly or low-quality solar films are selected, it will lead to unclear interior conditions of the vehicle, thus triggering a series of serious safety hazards. The specific hazards are as follows: 1. Overloading problem Difficult to detect: Due to the window being blocked by the solar film, it is difficult to see the actual number of people inside the vehicle from the outside, which easily leads to overloading.

[0005] Safety risks: When a vehicle is overloaded, its total mass will exceed the design standard, which will not only cause the braking performance of the vehicle to decline and the steering stability to deteriorate, but also accelerate the wear of the tires. Furthermore, in case of emergencies, such as sudden braking or avoiding obstacles, it will increase the risk of traffic accidents. In addition, overloading may also lead to overcrowding inside the vehicle, and once a collision or other accidents occur, the risk of injury to passengers will increase significantly.

[0006] 2. Mixing passengers and goods Violation: Some vehicle owners use dark solar films to cover up the behavior of mixing passengers and goods in order to avoid inspection.

[0007] Traffic hazards: Mixing passengers and goods affects the vehicle's center of gravity and braking performance, increasing the probability of traffic accidents.

[0008] 3. Irregularities in the behavior of drivers and passengers Monitoring difficulties: The sun film makes it difficult to observe the situation inside the vehicle from the outside, and irregular behaviors of drivers and passengers inside the vehicle (such as smoking, making phone calls, etc.) are difficult to be detected and stopped in time.

[0009] Legal risks: These behaviors not only violate traffic regulations but also pose a threat to the safety of drivers and other passengers.

[0010] 4. Carrying dangerous goods High concealment: The sun film can effectively hide dangerous goods inside the vehicle (such as knives, weapons, etc.), making it difficult for law enforcement officers to detect.

[0011] Potential threats: These items may be used to harm others in case of emergency, increasing the social security risks.

[0012] 5. Health problems of drivers Difficult to detect emergency illness: If a driver suddenly falls ill during driving, it is difficult for the outside world to detect and provide rescue in time because the window is blocked by the sun film.

[0013] Serious consequences: This may lead to out-of-control driving and then cause serious traffic accidents.

[0014] Currently, the monitoring method of the vehicle interior state relies on manual inspections. The traffic control department usually needs to arrange a large number of staff to conduct inspections in areas such as roads, stations, and parking lots to check for dangerous driving phenomena such as overloading and mixing passengers and goods. The traffic network is huge and widely distributed. Especially at the junction of urban and suburban areas, the road conditions are complex, and it is impossible for traffic police officers to cover all key areas. Manual inspections cannot achieve real-time data feedback. Staff often need to conduct one-by-one inspections over a long period of time, resulting in the inability to respond quickly. A large number of traffic police, auxiliary police and other personnel are needed to participate in the inspections, and these personnel also need to be trained regularly, equipped with necessary equipment, and provided with transportation tools, etc., all of which require financial support. Vehicles may be delayed for a long time due to multiple stop-and-checks. The monitoring of dangerous driving inside the car relying on manual inspections has led to a series of problems such as waste of human resources, low work efficiency, and many monitoring blind spots. These problems not only increase the cost of social management but also affect the effect of traffic management.

[0015] To address this challenge, we propose an innovative solution - a monitoring method using a see-through window film. This method uses a spectral see-through element to accurately measure the optical properties of standard automotive glass and various window films, realizing the function of monitoring the situation inside the vehicle through the window (through the film). Summary of the Invention

[0016] The object of the present invention is to provide an automatic monitoring system for the interior state of a vehicle based on multi-spectral polarization through-window technology, so as to solve the problems in the prior art that the automotive window solar film makes it impossible for the outside to see the situation inside the vehicle and it is impossible to judge whether the passengers inside the vehicle exceed the standard load, etc.

[0017] An automatic overloading monitoring system for an automobile based on multi-spectral polarization through-window technology, comprising: A light source emission unit, which uses a multi-spectral element to penetrate the automotive window glass and the solar film; The multi-modal image acquisition unit is provided with a plurality of multi-modal cameras for acquiring image data of different modalities; A real-time image processing unit, which processes the image data of different modalities obtained by the multi-modal image acquisition unit to obtain a perspective image of the window; An overloading detection unit, which uses a passenger detection algorithm to identify the people in the perspective image of the window; An overloading alarm unit, which is used to automatically trigger an overloading alarm when the overloading detection unit detects that the number of passengers in the vehicle exceeds the legal or safety limit.

[0018] A further improvement to the solution of the present application is that the multi-spectral element includes a laser diode, a filter, a beam splitter, and a polarization light sheet; Among them, the laser diode is used to emit light of different bands; the filter is used to selectively pass light of a specific wavelength and block unwanted wavelengths; the beam splitter is used to separate light of different wavelengths; the polarization light sheet is used to generate polarized light in a specific direction, and by rotating the polarization light sheet, the polarization state of the light is changed to enhance the detection of solar films with different light transmittances.

[0019] It should be noted that the laser diode is configured in a wavelength band invisible to the human eye to avoid glare or other visual interference to the driver.

[0020] For further illustration, for solar films with different light transmittances, the configuration of the multi-spectral element is adjusted, and the selection can be made according to the following relationship: In order to adapt to solar films with different light transmittances and adjust the configuration of the multi-spectral element, Let T(λ) be the light transmittance of the solar film for wavelength λ, I(λ) be the light intensity emitted by the laser diode, and I f (λ) be the light intensity after passing through the filter, then there is a relationship: ; For the beam splitter, it divides the light into multiple wavelength components, and the light intensity I s (λ) of each component is expressed as: ; Where λi is the efficiency of the beam splitter; Let P(θ,λ) be the transmittance of the polarizing filter at angle θ and wavelength λ. Then the light intensity I p (θ,λ) for imaging is: .

[0021] In the solution of this application, the multi-modal image acquisition unit acquires polarization images at different polarization angles, performs polarization filtering on the obtained polarization images to separate the target light inside the vehicle and the ambient light, and obtains multiple spectral images of different bands; The real-time image processing unit uses the polarization through-window imaging algorithm to perform fusion calculation on the obtained multiple spectral images of different bands, extracts the polarization difference information between the target and the background, and obtains the perspective image of the window by reconstructing the target polarization information.

[0022] The overcrowding detection unit first preprocesses the obtained perspective image of the window, including resizing the image, normalizing, etc., to meet the input requirements of the model; Uses a pre-trained model to extract features from the image; Applies a target detection algorithm to identify the people in the image; For each detected person, the algorithm outputs a bounding box and a corresponding confidence score. By calculating the number of bounding boxes, the total number of people in the image is obtained.

[0023] Suppose there are n detected people in an image, and each person i corresponds to a bounding box and a confidence score C i . If the confidence threshold is set to T, only when C i ≥T, the target is counted in the total number of people; The total number of people N is: where [·] is the indicator function, which takes the value of 1 when the condition is true and 0 otherwise.

[0024] In the solution of this application, a further improvement is that the multi-modal image acquisition unit is also provided with an ordinary camera to capture the license plate information of the vehicle. According to the obtained vehicle license plate information, the overcrowding detection unit retrieves the vehicle model information and obtains the load capacity of the vehicle for comparison with the number of people identified from the perspective image of the window.

[0025] Advantages of the present invention: 1. Reducing manual intervention: The system can replace a large number of manual inspections through automated detection, greatly improving the efficiency and coverage of monitoring. Through real-time image analysis and automatic alarm mechanisms, traffic management departments can more quickly and accurately detect overcrowding problems and reduce traffic safety hazards.

[0026] 2. High efficiency and precision: The combination of multi-spectral data fusion and deep learning algorithms enables the system to not only accurately identify passengers in the vehicle but also overcome interference caused by various complex factors, such as window reflections and occlusion of passengers in the vehicle, reducing misjudgments and missed judgments.

[0027] 3. Privacy protection and law enforcement balance: The system can determine whether a vehicle is overloaded only by analyzing the information on the window surface, without the need for excessive snooping inside the vehicle, effectively balancing the needs of privacy protection and traffic law enforcement. It analyzes the situation inside the vehicle through specific band spectra and polarized light, avoiding direct infringement of the privacy of vehicle owners or passengers.

[0028] 4. Real-time feedback and linkage mechanism: The system has a real-time alarm function and can provide timely feedback to traffic management personnel through text messages, voice prompts, etc., and even directly link with the traffic monitoring system to achieve a rapid response.

[0029] 5. Application scenarios and potential: Urban traffic monitoring: During peak traffic hours, especially in crowded areas such as urban roads and stations, the system can automatically detect overloaded vehicles, improving the efficiency of traffic safety management.

[0030] Parking lot and road section inspections: For fixed areas such as parking lots and passenger stations, the system can be installed at intersections or entrances and exits to automatically detect overloading of vehicles entering and leaving, greatly reducing the workload of manual inspections. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0032] Figure 1 : System block diagram provided by the present invention; Figure 2 : Flow chart of the number of people recognized by the overloading detection unit provided by the present invention; Figure 3 : Chart of the spectral transmittance of different common automotive window tints provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all of them. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0034] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] As Figure 1 and Figure 2 shown, an automatic overloading monitoring system for vehicles based on multi-spectral polarization window technology designed in the present application includes: A light source emission unit that uses a multi-spectral element to penetrate the vehicle window glass and solar film; The multi-modal image acquisition unit is provided with multiple multi-modal cameras for acquiring image data of different modalities; A real-time image processing unit that processes the image data of different modalities obtained by the multi-modal image acquisition unit to obtain a perspective image of the internal state of the window; An overloading detection unit that uses a passenger detection algorithm to identify the people in the perspective image of the window; An overloading alarm unit that automatically triggers an overloading alarm when the overloading detection unit detects that the number of passengers in the vehicle exceeds the legal or safety limit.

[0036] A further improvement to the solution of the present application is that the multi-spectral element includes a laser diode, a filter, a beam splitter, and a polarization light sheet; Among them, the laser diode is used to emit light of different bands; the filter is used to selectively pass light of a specific wavelength and block unwanted wavelengths; the beam splitter is used to separate light of different wavelengths; the polarization light sheet is used to generate polarized light in a specific direction, and by rotating the polarization light sheet, the polarization state of the light is changed to enhance the detection of solar films with different light transmittance.

[0037] In an optional embodiment of the present application, the light source emission unit is arranged at a traffic collection site and is provided with a housing. The laser diode is arranged inside the housing and has a glass mask. The filter is pasted on the glass mask with glass glue, and the thickness of the glass glue is uniform and there are no air bubbles to eliminate the interference phenomenon.

[0038] It is worth noting that lasers of different bands are used to obtain high-quality images, especially in low-light or night conditions. The laser diode is usually configured to operate in a band invisible to the human eye (such as near-infrared) to reduce visual interference to drivers or other personnel. Among them, the wavelength of the near-infrared light laser diode is between 780nm and 850nm. At the same time, the precision of the laser system helps to improve the accuracy of image capture and reduce interference from other light sources.

[0039] The laser diode emits multi-wavelength light, and the light of different bands is processed by a filter and a beam splitter respectively. The polarizing filter further optimizes the collection of optical signals, enabling the multi-modal image acquisition unit to capture high-quality signals inside the vehicle, especially the detailed information under various environmental factors. This multi-wavelength, polarized optical imaging method can effectively overcome the interference of window films, especially the impact of one-way perspective films on traditional monitoring methods. For further illustration, for solar films with different light transmittances, the configuration of the multi-spectral elements is adjusted according to the following relationship for selection and matching: To adapt to solar films with different light transmittances in order to adjust the configuration of the multi-spectral elements, Let T(λ) be the light transmittance of the solar film at wavelength λ, I(λ) be the light intensity emitted by the laser diode, and I f (λ) be the light intensity after passing through the filter, then there is the relationship as shown in formula (1): ; For the beam splitter, it divides the light into multiple wavelength components, and the light intensity of each component is expressed as in formula (2): ; where I s (λ) is the light intensity of each component; λ i is the efficiency of the beam splitter; Let P(θ,λ) be the transmittance of the polarizing filter at angle θ and wavelength λ, then the light intensity finally used for imaging is obtained according to the relationship in formula (3): ; where, I p (θ,λ) is the light intensity for imaging, and I s (λ) is the light intensity of each component.

[0040] Figure 3 is a chart of the spectral light transmittances of several different common automotive window tintings. It can be seen from the chart that the region from 720 nm to 780 nm shows the highest spectral transmittance, thus becoming the best choice for seeing through automotive windows.

[0041] On this basis, we select a laser diode configured in a band invisible to the human eye (such as 780 nm) to avoid glare or other visual interference to the driver.

[0042] The filter is selected with a bandwidth of 10 nm and a center wavelength of 780 nm to ensure efficient filtering.

[0043] The efficiency λ of the beam splitter i is set to 70% to ensure effective separation of light of different bands.

[0044] The transmittance of the polarizing filter is 90% at an angle θ = 45°.

[0045] In the solution of this application, the multi-modal image acquisition unit uses a high-resolution CCD camera and an NIR camera to capture images in the visible light and near-infrared bands respectively.

[0046] Each camera is equipped with an independent filter and polarizer. The filter helps to select different wavelength ranges, and the polarizer helps to remove light pollution generated by non-target surfaces (such as reflected light), thereby enhancing the contrast and clarity of the image to ensure high-quality images are captured.

[0047] On this basis, multiple cameras can obtain polarization images at different polarization angles. This is because the reflected light from the object surface has polarization characteristics. By processing the images at different polarization angles, the in-vehicle target light and ambient light can be effectively separated, thereby reducing background interference and improving the recognizability of the target. The obtained polarization images are subjected to polarization filtering processing to separate the in-vehicle target light and ambient light, and multiple spectral images of different bands are obtained; specifically, image signals of different bands inside the vehicle can be captured, including visible light, near-infrared (NIR), and other possible bands. The multiple fusion of these data enables more accurate image processing.

[0048] The real-time image processing unit uses the polarization through-window imaging algorithm to perform fusion calculations on the obtained multiple spectral images of different bands, extracts the polarization difference information between the target and the background, and obtains the perspective image of the window by reconstructing the target polarization information.

[0049] The multi-modal image acquisition unit can simultaneously obtain images of multiple bands and fuse these image data to obtain richer information. During the fusion process, the real-time image processing unit will adopt the polarization through-window imaging algorithm, and this algorithm is processed through the following steps: First, it is necessary to align (register) the images from different cameras and different bands. This step ensures that the image data at different bands and different polarization angles can be compared and fused in the same coordinate system. Common registration methods include feature point-based matching, gray value-based optimization, etc.

[0050] During the image fusion process, the real-time image processing unit will extract the key information in the images of different bands. For example, the visible light image can provide information about the color and shape of the in-vehicle objects, the near-infrared image can reveal hidden targets under low-light conditions, and the polarization information helps to distinguish the window reflected light from the light of the internal target. The processing unit will extract the features of the background and the target through techniques such as image segmentation and edge detection.

[0051] Based on images with different polarization angles, the system can calculate the polarization differences between the target object and the background. This polarization difference information usually reflects the optical properties of the object's surface, such as reflectivity, refractive index, etc. By analyzing these differences, the system can accurately identify the target object, especially in complex environments, eliminating the interference of background light and window reflection light.

[0052] During the fusion process of multi-band and multi-polarization images, the real-time image processing unit performs weighted averaging according to the weights and characteristics of images in different bands. Through this fusion, information in multiple bands can be retained simultaneously, enhancing the visual effect of the image. Especially in the reconstruction of window perspective, the system can obtain a clearer perspective image.

[0053] The overcrowding detection unit first preprocesses the obtained perspective image of the window, including resizing the image, normalizing, etc., to meet the input requirements of the model. Use a pre-trained model to extract features from the image; specifically, a pre-trained convolutional neural network (CNN) model such as ResNet or VGG can be used to extract features from the image. These features are the key information in the image that helps identify the target. Apply object detection algorithms to identify the people in the image; apply object detection algorithms such as YOLO (You Only Look Once) or SSD (Single Shot MultiBox Detector) to identify the people in the image. These algorithms can directly predict the category and location of the target on the image. To improve the detection accuracy, the system also applies the non-maximum suppression (NMS) algorithm to reduce the targets of repeated detection and ensures that only targets with high confidence are detected by setting an appropriate confidence threshold.

[0054] During the object detection process, each detected target (such as a person) will output a bounding box and a corresponding confidence score. The bounding box is represented as (x 1 , y 1 x 2 , y 2 ), where (x 1 , y 1 ) are the coordinates of the upper left corner of the bounding box, and (x 2 , y 2 ) are the coordinates of the lower right corner. The confidence score represents the reliability of the detection result, usually between 0 and 1.

[0055] For each detected person, the algorithm will output a bounding box and a corresponding confidence score. By calculating the number of bounding boxes, the total number of people in the image can be obtained.

[0056] Suppose there are n detected people in an image, and each person i corresponds to a bounding box and a confidence score Ci If the confidence threshold is set to T (in an alternative embodiment of the present application, T = 0.7), only when C i ≥ T, the target is counted into the total number of people.

[0057] Therefore, the total number of people N can be calculated by formula (5): where [·] is the indicator function, which takes the value of 1 when the condition is true and 0 otherwise.

[0058] To improve the accuracy, it may be necessary to post-process the detection results, such as non-maximum suppression (NMS) to eliminate duplicate detected objects, and adjust the threshold to filter out low-confidence detection results.

[0059] Finally, the system will output a value containing the number of all detected people in the image.

[0060] In the solution of the present application, a further improvement is that the multi-modal image acquisition unit is further provided with an ordinary camera for capturing the license plate information of the vehicle. According to the obtained vehicle license plate information, the overcrowding detection unit retrieves vehicle model information, emergency contact numbers, etc., and obtains the load capacity of the vehicle for comparison with the number of people recognized through the perspective image of the vehicle window.

[0061] Suppose the license plate number captured by the ordinary camera is P, and the vehicle model information can be retrieved according to the license plate number.

[0062] According to the license plate number P, the system can query the database or API interface to obtain the vehicle model information of the vehicle, and then obtain the standard load capacity L of the vehicle model. For example, if the vehicle model corresponding to the license plate number is a 5-seater sedan, then the standard load capacity L = 5.

[0063] By comparing the number of people N recognized through the perspective image with the standard load capacity L of the vehicle, it can be determined whether there is overcrowding. If N > L, it is considered that the vehicle is overcrowded.

[0064] The vehicle overcrowding automatic monitoring system based on the multi-spectral polarization through-window technology provided by the present application combines the multi-spectral element and the multi-modal image acquisition unit, enabling the system to obtain high-quality in-vehicle images under various lighting conditions (including low light and night). The real-time image processing unit uses the polarization through-window imaging algorithm, which can effectively overcome the interference of the window film, especially the influence of the one-way perspective film on the traditional monitoring method, thereby improving the accuracy of image processing. The overcrowding detection unit uses an advanced object detection algorithm, combined with the non-maximum suppression algorithm, to ensure high confidence and low false alarm rate of the detection results. The system has a high degree of automation, reduces the need for manual inspections, and can improve the work efficiency of the traffic management department. At the same time, the system only analyzes the information on the window surface and does not directly peek into the details inside the vehicle, protecting the privacy of passengers.

[0065] It should be noted that the above-described embodiments are only descriptions of the preferred modes of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present application shall fall within the protection scope determined by the claims of the present application. It should be noted that in this text, relational terms such as first and second 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 term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

Claims

1. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology is characterized by: include: Light source emitting unit, using multi-spectral elements to penetrate automobile window glass and solar film; The multimodal image acquisition unit is provided with a plurality of multimodal cameras for acquiring image data of different modalities; A real-time image processing unit processes the image data of different modes acquired by the multi-modal image acquisition unit to obtain a perspective image of the vehicle window; an overcrowding detection unit, which uses a passenger detection algorithm to identify people in the perspective image of the vehicle window; The overcrowding alarm unit is used to automatically trigger an overcrowding alarm when the overcrowding detection unit detects that the number of passengers in the vehicle exceeds the legal or safety limit.

2. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology according to claim 1 is characterized in that: The multi-spectral element includes a laser diode, a filter, a beam splitter and a polarizer; Among them, the laser diode is used to emit light of different wavelengths; the filter is used to selectively pass light of specific wavelengths and block unnecessary wavelengths; the beam splitter is used to separate light of different wavelengths; the polarizer is used to generate polarized light in a specific direction, and the polarization state of light is changed by rotating the polarizer to enhance the detection of solar films with different transmittances.

3. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology according to claim 2 is characterized in that: In order to adapt to solar films with different transmittances and adjust the configuration of the multi-spectral element, let T(λ) be the transmittance of the solar film to wavelength λ, I(λ) be the light intensity emitted by the laser diode, and I f (λ) is the light intensity after passing through the filter, then we have the relationship: ; For the beam splitter, it divides the light into multiple wavelength components, and the light intensity of each component is s (λ) is expressed as: ; where λ i is the efficiency of the beam splitter; Assume P(θ,λ) is the transmittance of the polarizer at angle θ and wavelength λ, then the final light intensity I used for imaging is p (θ,λ) is: .

4. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology according to claim 3 is characterized in that: The multimodal image acquisition unit acquires polarization images at different polarization angles, and performs polarization filtering on the polarization images to separate the target light and the ambient light in the vehicle, thereby obtaining a plurality of spectral images in different bands; The real-time image processing unit uses the polarization window imaging algorithm to fuse and calculate the acquired spectral images of multiple different bands, extract the polarization difference information of the target and the background, and obtain the perspective image of the car window by reconstructing the target polarization information.

5. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology according to claim 4 is characterized in that: The overcrowding detection unit first preprocesses the acquired perspective image of the vehicle window, including adjusting the image size, normalizing, etc., to meet the input requirements of the model; Extract features from images using a pre-trained model; Apply object detection algorithms to identify people in images; For each detected person, the algorithm outputs a bounding box and a corresponding confidence score. By counting the number of bounding boxes, the total number of people in the image is obtained.

6. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology according to claim 5 is characterized in that: Assume that there are n detected persons or pre-set targets in an image, and each person i corresponds to a bounding box and a confidence score C i If the confidence threshold is set to T, only when C i When ≥T, the target will be counted in the total number; The total number of people N is: Where [·] is an indicator function, which takes the value 1 when the condition is true and 0 otherwise.

7. The vehicle interior state automatic monitoring system based on multi-spectral polarization window technology according to claim 1 is characterized in that: The multimodal image acquisition unit is also provided with a common camera for capturing the vehicle's license plate information. Based on the acquired vehicle license plate information, the overcrowding detection unit retrieves the vehicle model information and obtains the number of people carried by the vehicle for comparison with the number of people recognized through the perspective image of the vehicle window.

Images