Intelligent traffic multi-angle night vision snapshot camera device and camera system

Abstract

The invention discloses an intelligent traffic multi-angle night vision snapshot camera device and system, and belongs to the technical field of image processing, and the device comprises a multi-mode sensing layer which is connected to a traffic information collection end; the anti-interference physical layer comprises a lens anti-interference module arranged at the traffic information acquisition end; the edge computing layer comprises an embedded NPU (Network Processing Unit) chip, and the chip realizes multi-stage distributed processing through a multi-stage processing mechanism; and the cloud collaboration layer realizes real-time synchronization and storage of data through a distributed server cluster in combination with a network communication module, and supports access of multiple terminal devices at the same time. In the implementation process of the technical scheme of the invention, multi-dimensional data acquisition is performed on a traffic scene through the multi-modal sensing layer, and distributed processing is performed on the acquired image through the edge calculation layer after acquisition is completed, so that multi-dimensional analysis of the traffic image is realized.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, specifically to a multi-angle night vision capture imaging device and camera system for smart transportation. Background Art

[0002] Thanks to the development of image processing technology, smart transportation is gradually moving towards automation and precision. The combination of image processing technology and smart transportation can realize functions such as identification of various illegal behaviors and traffic flow management, and is the current mainstream development direction of smart transportation.

[0003] The coordinated use of smart transportation and image processing first requires image acquisition of the target area, capturing images or videos through a camera device, and then performing real-time analysis through an image processing algorithm to identify traffic violations. This method has become the current mainstream solution and has been widely used. In order to improve the accuracy of capture and subsequent identification, the currently commonly used method is to use high-precision cameras and multi-angle shooting technology, combined with night vision functions to ensure clear image capture under various lighting conditions, and combined with corresponding image processing algorithms to perform multi-dimensional analysis of the captured images, thereby accurately identifying various types of traffic violations and improving the efficiency and accuracy of traffic management.

[0004] However, there are still some problems in the application of existing technologies. For example, in adverse weather conditions such as rain and fog, the image clarity decreases, resulting in reduced recognition accuracy. Especially in low-light environments at night, the capture ability of the camera device is limited and it is easily interfered with. Although infrared methods are used to enhance night vision effects in existing technologies, due to the serious color distortion of infrared images and insufficient detail capture, it is difficult to meet the needs of complex traffic scenes, resulting in large recognition errors in actual applications, affecting the overall effect of traffic management.

[0005] Therefore, it is necessary to provide smart traffic multi-angle night vision capture imaging device and camera system to solve the above problems.

[0006] It should be noted that the above information disclosed in this Background section is only for understanding the background technology of the present application concept, and therefore, it may contain information that does not constitute prior art. Summary of the Invention

[0007] Based on the above-mentioned problems existing in the prior art, the problem to be solved by this application is: to provide a smart traffic multi-angle night vision capture imaging device and camera system, and to improve the imaging effect by optimizing the images of scenes with different visibility such as rainy and foggy weather.

[0008] The technical solution adopted by this application to solve the technical problem is: a smart traffic multi-angle night vision capture imaging system, including: A multimodal perception layer connected to a traffic information collection terminal for collecting multi-dimensional data on traffic scenes; An anti-interference physical layer, which includes a lens anti-interference module provided at the traffic information collection terminal, and is used to reduce the impact of environmental factors on imaging in unusual weather conditions; The edge computing layer includes an embedded NPU chip that implements multi-level distributed processing through a multi-level processing mechanism; The cloud collaboration layer realizes real-time synchronization and storage of data through a distributed server cluster combined with a network communication module, and supports access from multiple terminal devices at the same time.

[0009] During the implementation of the technical solution of this application, multi-dimensional data of traffic scenes is collected through the multimodal perception layer, and after the collection is completed, the collected images are distributedly processed through the edge computing layer, thereby realizing multi-dimensional analysis of traffic images.

[0010] Furthermore, the multimodal perception layer includes an imaging module, an environmental monitoring module and a network communication module. The imaging module is used to image the target object. The imaging module includes a visible light imaging unit and an infrared imaging unit. The visible light imaging unit is used to collect the target object under normal lighting conditions, and the infrared imaging unit is used to capture the target object at night or in low light environments, as well as to capture the target object in scenes with low visibility in rainy and foggy weather.

[0011] Furthermore, the multimodal perception layer also includes an imaging optimization module, which performs image fusion processing through collaborative imaging of multi-band spectral data, and the multi-band spectral data includes short-wave visible light image data and long-wave near-infrared image data.

[0012] Furthermore, the image fusion processing is implemented by the weighted pixel superposition method, in which the short-wave visible light image data and the long-wave near-infrared image data are pixel-superimposed according to the weights, and the distribution of weights is dynamically adjusted according to the data monitored by the environmental monitoring module. Among them, the weights are obtained by the LSTM network, and the environmental parameters collected by the environmental monitoring module are input into the LSTM network. The time series characteristics of each parameter are extracted through the network, and then the optimal weight of the imaging effect under different environmental parameters is judged according to the time series.

[0013] Furthermore, after respectively obtaining the corresponding weights, the short-wave visible light image data and the long-wave near-infrared image data are pixel-wise superimposed according to the weights to generate a fused image. The specific fusion process is: first, the short-wave visible light image data and the long-wave near-infrared image data are pre-processed according to the corresponding weights to obtain image data after weight adjustment, and the size of the image data after weight adjustment remains consistent. Then, the superposition ratio is determined through pixel-by-pixel comparative analysis. After calculating the superposition ratio, the short-wave visible light image data and the long-wave visible light image data are pixel-wise fused according to the superposition ratio to generate a fused superimposed image.

[0014] Furthermore, the overlay ratio is determined by adopting a pixel-based difference analysis method, which includes: respectively obtaining the proportion of effective pixels in the short-wave visible light image data and the long-wave near-infrared image, where the effective pixel refers to the pixel with clear information characteristics in the image, and then calculating the overlay ratio according to the proportion of effective pixels. The specific calculation formula of the overlay ratio is: overlay ratio = (short-wave effective pixel number / total pixel number)*short-wave weight + (long-wave effective pixel number / total pixel number)*long-wave weight.

[0015] Furthermore, the multi-level processing mechanism of the edge computing layer is as follows: the first-level processing includes real-time image acquisition at the acquisition end and recording of timestamp information, and then resizing, color correction and noise suppression of the acquired image; the second-level processing uses the NPU chip for feature extraction and target recognition, as well as image fusion processing; the third-level processing recognizes the fused image and generates recognition results, and then generates an analysis report by comparing the recognition results with the database, where the database contains historical traffic data and historical environmental parameters, and automatically fills in the current data during each recognition process.

[0016] Furthermore, the cloud-based collaborative layer also includes a collaborative early warning module, which is used to conduct timely collaborative early warnings through real-time analysis and identification results, and send early warning notifications to relevant management departments through the network communication module to provide real-time early warnings for the target area.

[0017] Smart traffic multi-angle night vision capture imaging device, the device includes: A multi-physical quantity sensor array, wherein the multi-physical quantity sensor array is used to collect environmental parameters of a target area; An embedded NPU chip is connected to a multi-physical quantity sensor array and a lens group. The chip is used to receive and process data collected by the multi-physical quantity sensor array and dual-band image information collected by the lens group in real time. The embedded NPU chip performs imaging optimization according to environmental parameters and combines with the LSTM network for pixel superposition to generate a fused image. The specific fusion process is as follows: first, the short-wave visible light image data and the long-wave near-infrared image data are pre-processed according to the corresponding weights to obtain image data after weight adjustment, and the size of the image data after weight adjustment is kept consistent to achieve size alignment, thereby facilitating pixel fusion processing. Then, the superposition ratio is determined by pixel-by-pixel comparative analysis. The pixel-by-pixel comparative analysis is performed to determine the superposition ratio using pixel-based difference analysis. The proportion of effective pixels in the short-wave visible light image data and the long-wave near-infrared image is obtained respectively. An effective pixel refers to a pixel with clear information characteristics in the image. Then, the superposition ratio is calculated according to the proportion of effective pixels. The specific calculation formula for the superposition ratio is: superposition ratio = (short-wave effective pixel number / total pixel number) * short-wave weight + (long-wave effective pixel number / total pixel number) * long-wave weight; The annular air nozzle is set around the lens and is connected to a centralized air pump. The annular air nozzle is connected to the embedded NPU chip. After the embedded NPU chip analyzes the environmental parameters, it determines the weather conditions and sends an execution signal to the centralized air pump to spray air, forming an annular airflow barrier. Therefore, in unconventional weather conditions, dust and water droplets are not easily attached to the surface of the lens group, ensuring stable imaging quality. A polarizer loading unit includes a polarizer that automatically adjusts the polarization angle according to changes in ambient light to eliminate imaging effects caused by glare and reflection.

[0018] Furthermore, the model of the embedded NPU chip is Hailo-8.

[0019] The beneficial effects of the present application are: the smart traffic multi-angle night vision capture imaging device and camera system provided by the present application collects multi-dimensional data of traffic scenes through a multimodal perception layer, and after the collection is completed, the collected images are distributedly processed through the edge computing layer, thereby realizing multi-dimensional analysis of traffic images, and in the subsequent image fusion processing process, the weighted pixel superposition method is adopted to superimpose the short-wave visible light image data and the long-wave near-infrared image data according to the weights. The distribution of weights is dynamically adjusted according to the data monitored by the environmental monitoring module. The fused image can maintain a high degree of detail restoration and environmental adaptability under different lighting conditions, effectively solving the imaging difficulty of traditional methods in complex lighting environments.

[0020] In addition to the above-described purposes, features and advantages, the present application also has other purposes, features and advantages. The present application will be further described in detail below with reference to the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings that constitute part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation on this application. In the drawings: Figure 1 This is a schematic diagram of the overall structure of the smart traffic multi-angle night vision capture imaging system in this application; Figure 2 This is a connection diagram of the smart traffic multi-angle night vision capture imaging device in this application; Among them, the reference numerals in the figures are: 101. Multi-physical quantity sensor array; 102. Embedded NPU chip; 103. Lens assembly; 104. Ring air nozzle; 105. Polarizer loading unit. DETAILED DESCRIPTION

[0022] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0024] Example 1: Figure 1 As shown, the present application provides a smart traffic multi-angle night vision capture imaging system, which is applied to smart traffic, mainly involving capturing and identifying pedestrians, non-motor vehicles, motor vehicles and other targets, analyzing the captured images or video information, judging whether there is any illegal behavior, and transmitting it to the traffic management department in real time, thereby realizing intelligent traffic management, and the scenarios targeted by this embodiment are low-light environments at night and traffic monitoring under complex weather conditions, and improving the capture accuracy and imaging quality from the hardware and software levels, so that the subsequent illegal behavior analysis is more accurate, such as Figure 1 As shown, the system includes: A multimodal perception layer, which is connected to the traffic information collection terminal through integration or plug-in. The traffic information collection terminal is used to collect multi-dimensional data of traffic scenes, including various images and videos in traffic scenes, and collect data in different scenarios through the multimodal perception layer; Among them, the multimodal perception layer includes an imaging module and an environmental monitoring module. The imaging module is used to image the target object. The imaging module includes at least two imaging units, such as a visible light imaging unit and an infrared imaging unit. The visible light imaging unit is used to capture the target object under normal lighting conditions, such as a clear daytime environment. The infrared imaging unit is used to capture the target object at night or in a low-light environment, as well as in rainy and foggy weather with low visibility. It should be noted that the visible light imaging unit and the infrared imaging unit can operate independently or together; The environmental monitoring module is used to monitor the surrounding environment in real time and adaptively switch the imaging module based on the monitoring results. For example, according to parameters such as ambient temperature, humidity, and light intensity, the parameter settings of the imaging unit are dynamically adjusted to optimize the imaging effect and ensure that the appropriate imaging unit can be used in different environments. In this embodiment, the environmental monitoring module is a multi-physical quantity sensor array, such as a BME688 meteorological sensor, a speed sensor, and an OPC-N3 aerosol sensor. The multi-physical quantity sensor array comprehensively monitors environmental parameters such as temperature, humidity, and particulate matter concentration in the target area. The multimodal perception layer also includes a network communication module, which is responsible for data transmission and instruction issuance between modules to ensure real-time information sharing. It also receives external information through this module. For example, it can capture local weather data through the network communication module and analyze it according to time periods to predict future weather change trends, adjust the imaging parameters of each imaging unit in advance, and optimize the capture effect. The multimodal perception layer also includes an imaging optimization module, which uses multi-band spectral collaborative imaging. The number of band spectra is related to the number of imaging units. The advantages of different imaging units are combined to fuse the image. Specifically, taking the visible light imaging unit and the infrared imaging unit as an example, the visible light imaging unit and the infrared imaging unit have different imaging characteristics. In this embodiment, the imaging data of different channels are fused according to the different bands to achieve dual-band collaborative imaging. The imaging principle of the visible light imaging unit is mainly achieved through short-wave visible light, while the imaging principle of the infrared imaging unit is mainly achieved through long-wave near-infrared. Externally, during the fusion processing, the weighted pixel superposition method is adopted to superimpose the short-wave visible light image data and the long-wave near-infrared image data according to the weight. The distribution of weights is dynamically adjusted according to the data monitored by the environmental monitoring module. For example, in low-light environments (night and rainy and foggy weather), the weight of infrared data is increased to improve image clarity; while in sufficient light (daytime under normal weather), the weight of infrared data is reduced and the performance of visible light data is enhanced to ensure rich image details. The specific value of the weight is dynamically calculated by the imaging optimization module based on real-time environmental parameters without manual calibration. In this embodiment, the weights are obtained using an LSTM network. The environmental parameters collected by the environmental monitoring module are input into the LSTM network. The network extracts the temporal characteristics of each parameter, and then judges the optimal weights of the imaging effects under different environmental parameters according to the temporal sequence, thereby eliminating the environmental judgment process of the target area. In addition, since the existing lighting environment determination method usually has only two lighting levels (sufficient lighting and insufficient lighting), the weight values ​​obtained through the environmental parameters can be adjusted steplessly, and the superposition weights of image data in different bands are directly matched with the environmental parameters, resulting in better imaging effects. The LSTM network is a deep learning model based on long short-term memory, which can capture the dynamic changes of environmental parameters according to the input parameters. The network has temporal characteristics and will not lose temporal information during the analysis process, which facilitates subsequent image processing and analysis, ensuring that high-quality imaging results can be obtained under different environmental conditions. After respectively obtaining the corresponding weights, the short-wave visible light image data and the long-wave near-infrared image data are pixel-superimposed according to the weights to generate a fused image. The specific fusion process is: first, the short-wave visible light image data and the long-wave near-infrared image data are pre-processed according to the corresponding weights to obtain the image data after weight adjustment, and the size of the image data after weight adjustment is kept consistent to achieve size alignment, thereby facilitating pixel fusion processing. Then, the superposition ratio is determined by pixel-by-pixel comparative analysis. The pixel-by-pixel comparative analysis is performed to determine the superposition ratio by adopting pixel-based difference analysis. The proportion of effective pixels in the short-wave visible light image data and the long-wave near-infrared image is obtained respectively. The effective pixel refers to the pixel with clear information characteristics in the image. Then, the superposition ratio is calculated according to the proportion of effective pixels. The specific calculation formula for the overlay ratio is: Overlay ratio = (shortwave effective pixel count / total pixel count) * shortwave weight + (longwave effective pixel count / total pixel count) * longwave weight. For example, when the shortwave effective pixel count is 5000, the total pixel count is 10000, the shortwave weight is 0.7, the longwave effective pixel count is 3000, and the longwave weight is 0.3, the overlay ratio = (5000 / 10000) * 0.7 + (3000 / 10000) * 0.3 = 0.35 + 0.21 = 0.56, after calculating the stacking ratio, the shortwave and longwave image data are pixel-wise fused according to the ratio to generate a fused stacked image. Taking the stacking ratio of 0.56 as an example, the pixel value of the shortwave image is multiplied by 0.56, and the pixel value of the longwave image is multiplied by 0.44. The two are then added to obtain the pixel value of the fused image. The final fused image not only retains the detail information of the shortwave image, but also integrates the environmental characteristics of the longwave image, thereby improving the overall clarity and contrast of the image. In this way, the fused image can maintain a high degree of detail restoration and environmental adaptability under different lighting conditions, effectively solving the imaging difficulty of traditional methods in complex lighting environments. Compared with the traditional pixel-level stacking method, due to the introduction of the weight adjustment mechanism and the dynamic optimization of the weights through the LSTM model, the stacking result is always related to the current environmental parameters, thereby ensuring the consistency of the fused image in different environments. The system also includes an anti-interference physical layer, which includes a lens anti-interference module arranged at the traffic information collection end. The lens anti-interference module is used to reduce the impact of environmental factors on imaging under unconventional weather conditions. The lens anti-interference module includes an annular air nozzle, which is arranged around the lens and connected to a centralized air pump. The centralized air pump sprays air to form an annular airflow barrier. Therefore, under unconventional weather conditions, dust and water droplets are not easily attached to the lens surface, ensuring stable imaging quality; in addition, the lens anti-interference module also includes a polarizer loading unit, which can automatically adjust the polarization angle according to changes in ambient light to eliminate the negative effects of glare and reflection. The specific implementation principle can be referred to the existing technology and will not be described in detail in this embodiment; The system also includes an edge computing layer, which includes an embedded NPU chip, such as Hailo-8. The chip has the characteristics of high efficiency and low power consumption, can perform lightweight model deployment, process image data in real time, reduce latency, and realize multi-level distributed processing through a multi-level processing mechanism. Specifically, the first-level processing includes real-time image acquisition at the acquisition end and recording timestamp information, and then resizing, color correction and noise suppression of the acquired image. The second-level processing uses the NPU chip to perform feature extraction and target recognition, as well as image fusion processing in the aforementioned process. The third-level processing will recognize the fused image and generate recognition results, and then generate an analysis report by comparing the recognition results with the database. The database contains historical traffic data and historical environmental parameters, and automatically supplements the current data during each recognition process to achieve dynamic data updates and continuous optimization of the model, ensuring the high accuracy and stability of the system in different environments. The system also includes a cloud-based collaborative layer, which uses a distributed server cluster combined with a network communication module to achieve real-time synchronization and storage of data, and simultaneously supports access by multiple terminal devices. Image data and recognition results acquired in different target areas can be uploaded to the cloud in real time. After cloud-based big data analysis, data docking and sharing can be achieved, making the dynamic update of the database more accurate, improving decision-making support capabilities, and ensuring the real-time and efficient nature of traffic management.

[0025] The cloud-based collaborative layer also includes a collaborative early warning module, which is used to conduct timely collaborative early warnings through real-time analysis and identification results, and send early warning notifications to relevant management departments through the network communication module to provide real-time early warnings for target areas. For example, when traffic accidents or facility damage occur, each department will receive early warning information in a timely manner.

[0026] Example 2: Figure 2 As shown, this application also proposes a smart traffic multi-angle night vision capture device, which includes: A multi-physical quantity sensor array 101, which is used to collect environmental parameters of a target area; The embedded NPU chip 102 is connected to the multi-physical quantity sensor array 101 and is connected to the lens group 103. The chip is used to receive and process the data collected by the multi-physical quantity sensor array 101 and the dual-band image information collected by the lens group 103 in real time. The embedded NPU chip 102 performs imaging optimization according to environmental parameters and combines the LSTM network to perform pixel superposition to generate a fused image. The specific fusion process is: first, the short-wave visible light image data and the long-wave near-infrared image data are pre-processed according to the corresponding weights to obtain the image data after weight adjustment, and the size of the image data after weight adjustment is kept consistent to achieve size alignment, thereby facilitating pixel fusion processing. Then, the superposition ratio is determined by pixel-by-pixel comparative analysis. The pixel-by-pixel comparative analysis determines the superposition ratio by adopting pixel-based difference analysis to obtain the proportion of effective pixels in the short-wave visible light image data and the long-wave near-infrared image respectively. The effective pixel refers to the pixel with clear information characteristics in the image. Then, the superposition ratio is calculated according to the proportion of effective pixels. The specific calculation formula of the superposition ratio is: Superposition ratio = (Short-wave effective pixel points / total pixel points)*short-wave weight + (long-wave effective pixel points / total pixel points)*long-wave weight; An annular air nozzle 104 is provided around the lens and is connected to a centralized air pump. The annular air nozzle 104 is connected to the embedded NPU chip 102. After the embedded NPU chip 102 analyzes environmental parameters, it determines the weather conditions and sends an execution signal to the centralized air pump to spray air, forming an annular airflow barrier. Therefore, under unconventional weather conditions, dust and water droplets are not easily attached to the surface of the lens group 103, ensuring stable imaging quality. The polarizer loading unit 105 includes a polarizer that automatically adjusts the polarization angle according to changes in ambient light, thereby eliminating imaging effects caused by glare and reflection.

[0027] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. Intelligent traffic multi-angle night vision capture imaging system, characterized by: include: A multimodal perception layer connected to a traffic information collection terminal for collecting multi-dimensional data on traffic scenes; An anti-interference physical layer, which includes a lens anti-interference module provided at the traffic information collection terminal, and is used to reduce the impact of environmental factors on imaging in unusual weather conditions; The edge computing layer includes an embedded NPU chip that implements multi-level distributed processing through a multi-level processing mechanism; The cloud collaboration layer realizes real-time synchronization and storage of data through a distributed server cluster combined with a network communication module, and supports access from multiple terminal devices at the same time.

2. The intelligent traffic multi-angle night vision capture imaging system according to claim 1 is characterized by: The multimodal perception layer includes an imaging module, an environmental monitoring module and a network communication module. The imaging module is used to image the target object. The imaging module includes a visible light imaging unit and an infrared imaging unit. The visible light imaging unit is used to collect the target object under normal lighting conditions, and the infrared imaging unit is used to capture the target object at night or in low-light environments, as well as in rainy and foggy weather with low visibility.

3. The intelligent traffic multi-angle night vision capture imaging system according to claim 2 is characterized by: The multimodal perception layer also includes an imaging optimization module, which performs image fusion processing on the image through collaborative imaging of multi-band spectral data, where the multi-band spectral data includes short-wave visible light image data and long-wave near-infrared image data.

4. The intelligent traffic multi-angle night vision capture imaging system according to claim 3 is characterized by: The image fusion processing is achieved by using the weighted pixel superposition method. The short-wave visible light image data and the long-wave near-infrared image data are pixel-superimposed according to the weights. The distribution of weights is dynamically adjusted according to the data monitored by the environmental monitoring module. The weights are obtained by using the LSTM network. The environmental parameters collected by the environmental monitoring module are input into the LSTM network. The time series characteristics of each parameter are extracted through the network, and then the optimal weight of the imaging effect under different environmental parameters is determined according to the time series.

5. The intelligent traffic multi-angle night vision capture imaging system according to claim 4 is characterized by: After obtaining the corresponding weights respectively, the short-wave visible light image data and the long-wave near-infrared image data are pixel-wise superimposed according to the weights to generate a fused image. The specific fusion process is: first, the short-wave visible light image data and the long-wave near-infrared image data are pre-processed according to the corresponding weights to obtain image data after weight adjustment, and the size of the image data after weight adjustment remains consistent. Then, the superposition ratio is determined through pixel-by-pixel comparison and analysis. After calculating the superposition ratio, the short-wave visible light image data and the long-wave visible light image data are pixel-wise fused according to the superposition ratio to generate a fused superimposed image.

6. The intelligent traffic multi-angle night vision capture imaging system according to claim 5, characterized in that: The overlay ratio is determined by a pixel-based difference analysis method, which includes: obtaining the proportion of effective pixels in the short-wave visible light image data and the long-wave near-infrared image respectively, where an effective pixel refers to a pixel with clear information characteristics in the image, and then calculating the overlay ratio according to the proportion of effective pixels. The specific calculation formula for the overlay ratio is: overlay ratio = (short-wave effective pixel number / total pixel number) * short-wave weight + (long-wave effective pixel number / total pixel number) * long-wave weight.

7. The intelligent traffic multi-angle night vision capture imaging system according to claim 6, characterized in that: The multi-level processing mechanism of the edge computing layer is as follows: the first-level processing includes real-time image acquisition at the acquisition end and recording of timestamp information, and then resizing, color correction and noise suppression of the acquired image; the second-level processing uses the NPU chip for feature extraction and target recognition, as well as image fusion processing; the third-level processing recognizes the fused image and generates recognition results, and then generates an analysis report by comparing the recognition results with the database, where the database contains historical traffic data and historical environmental parameters, and automatically fills in the current data during each recognition process.

8. The intelligent traffic multi-angle night vision capture imaging system according to claim 7, characterized in that: The cloud collaboration layer also includes a collaborative early warning module, which is used to conduct timely collaborative early warnings through real-time analysis and identification results, and send early warning notifications to relevant management departments through the network communication module to provide real-time early warnings for the target area.

9. Intelligent traffic multi-angle night vision capture imaging device, characterized by: The device includes: A multi-physical quantity sensor array (101), the multi-physical quantity sensor array (101) is used to collect environmental parameters of a target area; An embedded NPU chip (102) is connected to the multi-physical quantity sensor array (101) and is connected to a lens group (103). The chip is used to receive and process data collected by the multi-physical quantity sensor array (101) and dual-band image information collected by the lens group (103) in real time. The embedded NPU chip (102) performs imaging optimization according to environmental parameters and combines with the LSTM network to perform pixel superposition to generate a fused image. The specific fusion process is as follows: first, the short-wave visible light image data and the long-wave near-infrared image data are pre-processed according to corresponding weights to obtain image data after weight adjustment, and the size of the image data after weight adjustment is kept consistent to achieve size alignment, thereby facilitating pixel fusion processing. Then, the superposition ratio is determined by pixel-by-pixel comparative analysis. The pixel-by-pixel comparative analysis is performed to determine the superposition ratio by using pixel-based difference analysis. The ratio of effective pixels in the short-wave visible light image data and the long-wave near-infrared image is obtained respectively. The effective pixel refers to the pixel with clear information characteristics in the image. Then, the superposition ratio is calculated according to the ratio of effective pixels. The specific calculation formula of the superposition ratio is: Superposition ratio = (Short-wave effective pixel points / total pixel points)*short-wave weight + (long-wave effective pixel points / total pixel points)*long-wave weight; An annular air nozzle (104) is provided around the lens, and the annular air nozzle (104) is connected to a centralized air pump. The annular air nozzle (104) is connected to the embedded NPU chip (102). After the embedded NPU chip (102) analyzes environmental parameters, it determines the weather conditions and sends an execution signal to the centralized air pump to eject airflow, thereby forming an annular airflow barrier. Therefore, under unconventional weather conditions, dust and water droplets are not easily attached to the surface of the lens group (103), thereby ensuring stable imaging quality. A polarizing plate loading unit (105) comprises a polarizing plate capable of automatically adjusting a polarization angle, wherein the polarizing plate is used to automatically adjust the polarization angle according to changes in ambient light, thereby eliminating imaging effects caused by glare and reflection.

10. The intelligent traffic multi-angle night vision capture imaging device according to claim 9, characterized in that: The model of the embedded NPU chip (102) is Hailo-8.

Images