Search

CN-122023357-A - Method for inverting aerosol optical thickness by utilizing night airglow image

CN122023357ACN 122023357 ACN122023357 ACN 122023357ACN-122023357-A

Abstract

The invention discloses a method for inverting aerosol optical thickness by utilizing night airglow images, and belongs to the technical fields of image processing and atmospheric remote sensing. Aiming at the problems of high cost, large uncertainty and difficult business in night aerosol observation of the existing laser radar and starlight photometer, the method classifies night airglow gray level images by utilizing a lightweight convolutional neural network introduced with a compression-activation attention module, screens out clear night sky images, registers and superimposes multiple frames of clear sky images to identify and extract the brightness of the North stars, and reflects the optical thickness of the night aerosol by utilizing the ratio of the brightness of the North stars to the standard brightness of the clear sky based on the beer-lambert law. The invention realizes low-cost and high-time-resolution automatic night aerosol monitoring, expands the new application of the airglow imaging data, breaks through the limitation of the traditional observation means on the night continuous monitoring capability, and has remarkable practicability and popularization value.

Inventors

  • FAN XUEHUA
  • CHEN HONGBIN
  • XIA XIANGAO
  • DUAN MINZHENG

Assignees

  • 中国科学院大气物理研究所

Dates

Publication Date
20260512
Application Date
20260202

Claims (10)

  1. 1. A method for inverting the optical thickness of an aerosol using night-time airglow images, comprising the steps of: s1, classifying continuously acquired night airglow gray images through a trained lightweight convolutional neural network model, and screening out clear night sky type images; S2, registering and superposing the screened multi-frame clear night sky type images to enhance and identify the position of the North star in the images; S3, extracting reference brightness and actual brightness information of the stars under the condition of sunny night sky from the image processed by the S2; S4, calculating the optical thickness of the aerosol at night from the ratio of the brightness of the non-aerosol after attenuation correction to the reference brightness of the North star under the condition of sunny night sky based on the beer-Lambert law.
  2. 2. The method of claim 1, wherein the step of determining the position of the substrate comprises, In S1, the lightweight convolutional neural network model takes MobileNetV network as a main trunk and introduces a compression-activation attention module, and outputs classification results of various night sky states including clear night sky for an input single-channel night airglow gray level image.
  3. 3. The method of claim 2, wherein the step of determining the position of the substrate comprises, The training process of the lightweight convolutional neural network model comprises the steps of dividing an airglow image data set after labeling into a training set, a verification set and a test set, training the model by adopting a cross entropy loss function with label smoothing and a AdamW optimizer, and uniformly processing an input image into a single-channel preset resolution in training.
  4. 4. The method of claim 1, wherein the step of determining the position of the substrate comprises, In S1, when screening the images of the clear night sky category, only selecting the images with the model classification confidence higher than a preset threshold for subsequent processing.
  5. 5. The method of claim 1, wherein the step of determining the position of the substrate comprises, In S2, registering and superposing the multi-frame images comprises registering and aligning the multi-frame images which are continuous in time by taking an image area containing the North stars as a template, superposing the registered multi-frame images so as to stably highlight the North stars in the superposed images and weaken other mobile stars.
  6. 6. The method of claim 1, wherein the step of determining the position of the substrate comprises, In S3, the correction of non-aerosol attenuation includes: According to the geographic position and time of image shooting, calculating the total transmittance of water vapor, oxygen and atmospheric molecules by adopting a radiation transmission model; And performing non-aerosol attenuation correction on the extracted brightness of the North Polaris.
  7. 7. The method of claim 1, wherein the step of determining the position of the substrate comprises, In S3, determining a reference brightness of the polar star under a clear night sky condition includes: And selecting a front-end percentage image with the highest brightness of the North stars from the historical clear night sky images, calculating the brightness average value of the front-end percentage image, and taking the value after non-aerosol attenuation correction as the reference brightness.
  8. 8. The method of claim 1, wherein the step of determining the position of the substrate comprises, Before S1, the method further comprises a step of performing redundancy elimination pretreatment on the continuously acquired night airglow gray images, wherein the redundancy elimination pretreatment extracts images according to fixed time intervals to form a treatment data set.
  9. 9. A computer terminal device is characterized by comprising one or more processors; A memory coupled to the processor for storing one or more programs; when executed by the one or more processors, causes the one or more processors to implement the steps of the method of any of claims 1-8.
  10. 10. A computer readable storage medium, characterized in that the computer program, when executed by a processor, implements the steps of the method according to any of claims 1-8.

Description

Method for inverting aerosol optical thickness by utilizing night airglow image Technical Field The invention belongs to the technical field of image processing and atmospheric remote sensing, and particularly relates to a method for inverting aerosol optical thickness by utilizing night airglow images. Background The observation of the optical thickness of the aerosol at night plays an important role in understanding the physical and chemical processes of the atmosphere, improving the forecasting capability of the air pollution mode and accurately evaluating the average daily air quality. Assimilating the aerosol optical thickness data into numerical prediction mode helps to improve the prediction accuracy, however, practical verification shows that the prediction error of the mode increases in the prediction period of 12 hours and 36 hours, which is directly because aerosol observation data for night period is proved to be lack as constraint. Further, the PM 2.5 concentration widely used in ambient air quality assessment is an average value of 24 hours, and significant errors are necessarily introduced if the daily PM 2.5 concentration is estimated by means of only the aerosol optical thickness during the day. Therefore, the acquisition of the night aerosol optical thickness data is important to improve the daily PM 2.5 concentration estimation precision and further improve the reliability of air quality estimation and prediction. Currently, the main technical means for acquiring the characteristics of night aerosol include laser radar observation and remote sensing by a starlight or moon photometer. The lidar is able to obtain a vertical profile of the attenuated backscatter coefficient of the aerosol, but in inverting the critical parameter of the extinction coefficient of the aerosol, an empirical value called the lidar ratio must be preset. The laser radars of different regions and different types of aerosols have obvious ratio difference, and even in the same place, the laser radars can change day and night, and the inherent uncertainty severely restricts the accuracy and the reliability of the characteristics of the laser radars, such as the inversion of the optical thickness of the aerosols, and the like. On the other hand, although the starlight photometer can realize inversion of the optical thickness of the aerosol at night, the instrument is required to be provided with a large-aperture optical system to collect extremely weak starlight, so that the whole system has high technical complexity and high manufacturing cost, and the starlight photometer is difficult to carry out large-scale business network deployment observation. The moon-based observation method has the advantages that the moon surface reflectivity changes along with the changes of the moon phase and the observation geometric conditions, the calibration is extremely complex due to the non-lambertian characteristics, the inversion result precision is seriously dependent on the instrument calibration coefficient, the existing moon observation instrument still has obvious technical defects, and the existing moon observation instrument needs to be subjected to contrast correction by depending on other observation means. In summary, in the prior art, the inversion accuracy is affected due to uncertain key parameters, or the business popularization and application are hindered due to over-high cost and technical threshold, so that the actual problem of lack of the night aerosol optical thickness observation data is jointly caused, and further improvement of the air pollution mode forecasting capability and realization of all-weather air quality accurate assessment are limited. Disclosure of Invention In order to solve the technical problems, the invention provides a method for inverting the optical thickness of aerosol by utilizing night airglow images, so as to solve the problems in the prior art. In a first aspect, to achieve the above object, the present invention provides a method for inverting an aerosol optical thickness using a night time airglow image, comprising the steps of: s1, classifying continuously acquired night airglow gray images through a trained lightweight convolutional neural network model, and screening out clear night sky type images; S2, registering and superposing the screened multi-frame clear night sky type images to enhance and identify the position of the North star in the images; S3, extracting reference brightness and actual brightness information of the stars under the condition of sunny night sky from the image processed by the S2, and then carrying out non-aerosol attenuation correction on the extracted reference brightness and actual brightness of the stars. S4, calculating the optical thickness of the aerosol at night from the ratio of the brightness of the non-aerosol after attenuation correction to the reference brightness of the North star under the condition of sunny night sky based on the beer-L