TWI820480B - Method and apparatus for precipitation prediction via image retrieving - Google Patents

Abstract

A precipitation forecast method and apparatus for localized areas is described therein. The apparatus provides real-time meteorological information and near-term precipitation forecast in the vicinity of the user’s current location. Precipitation is defined as the accumulated rainfall reaching 0.1 mm or above. Although the resolution of weather forecast models has improved significantly over the years, accurate weather forecasts of localized areas remain a difficult challenge. There are many inherent variables that limit the weather forecast model’s accuracy, such as the non-exact solutions of non-linear partial differential equations, the use of the data interpolation method, etc. The near-vicinity precipitation forecasting method consists of the following steps: (1) image(s) retrieval of the sky; (2) analysis based on the retrieved image(s); and (3) a precipitation forecast of the local area. The image retrieval step involves taking at least one or a plurality of images of the sky via the image retrieval unit, where the elevation angle is between 60 degrees to 120 degrees and the wide angle is between 45 degrees to 180 degrees. The image analysis step consists of an analysis system that is connected to the image retrieval unit to analyze the cloud cover, cloud type, and cloud color. The precipitation forecast step consists of cross-referencing the analysis with a database to calculate the precipitation probability of the local area. By retrieving images of the sky, a real-time precipitation forecast of the user’s location can be obtained. This is an important reference to prepare in advance for any possible weather changes.

Description

Short-range rainfall probability prediction method and device

The invention is a prediction method and device for predicting future rainfall trends by using a system to identify clouds after capturing close-range images.

"Rain" (rain) is a natural phenomenon that most easily affects life. The areas of influence include: the need to carry rain gear, the arrangement of itinerary, the choice of transportation, and the safety of outdoor activities (thunderstorms or heavy rain will Affecting the safety of activities), whether clothes can be dried, etc. Therefore, whether it will rain in their own area is one of the things that everyone pays attention to every day.

Currently, many apps or websites provide real-time weather information and rainfall forecasts. However, a common problem with such weather apps or websites is that the information they provide still cannot reach the accuracy of a small area. They are all based on regional , towns and other units mainly. One of the reasons is that most current weather forecasts use numerical weather prediction models for analysis. Although the grid point resolution continues to improve, it does not provide accurate predictions for a small area. Forecast (the small area can be between 1 km square and 50 km square). At present, the forecast for such small areas is supplemented by data interpolation. Another reason for prediction errors is that nonlinear partial differential equations that simulate the operation of the atmosphere cannot calculate complete solutions and can only obtain approximate solutions. Furthermore, the local terrain, thermal conditions at that time, moisture conditions in the sky at that time, etc. will all affect the subsequent weather development, so the results will be completely different from the original weather and rainfall forecast. For example, when clouds with a high probability of rainfall in a certain area pass through an area, the weather forecast system will say that there will be rainfall in this area based on this situation. However, in fact, part of the cloud layer may have a hollow shape. There will be no rainfall in the area. As shown in Figure 3, Taipei has been completely enveloped, but due to topography, there is no rainfall in the white area pointed by the red arrow in the picture.

In view of the above problems, the present invention provides a small-area and short-range rainfall prediction method and device, thereby allowing users to make real-time rainfall predictions based on their location.

The invention provides a rainfall prediction method, which includes the following steps: a sky image acquisition step, an image analysis step and a rainfall judgment step. Among them, the sky image capture step: capture at least one image towards the sky through an image capture unit device at an angle between 60 degrees and 120 degrees with the ground, and the wide-angle angle of the lens of the image capture unit is 45 degrees ~180 degrees; the image analysis step is to analyze the cloud amount, cloud type or cloud color contained in the captured image by a system, which is connected to the signal of the device with the image capture unit. The rain probability judgment step is to compare the cloud amount, cloud type or cloud color after the above analysis with a stored data in the system, and obtain a rain probability value from the stored data. Through the above steps, the purpose of real-time prediction of rainfall probability is achieved.

Wherein, the image capturing unit can capture a plurality of images at fixed time intervals within a time interval. This can be used to observe the movement direction and changes of clouds as an evaluation item. one.

It can further connect multiple image capture units to capture images at the same time. This can increase the scope of judgment.

Wherein, the step of determining the rain probability further uses the system to capture through the Internet and the system further performs analysis based on the date information of the captured image.

The system further captures air pressure, temperature, absolute humidity, relative humidity, dew point temperature, wind direction and wind speed information detected at the location of the device with the image capture unit.

The system further captures information on the high-altitude humidity distribution around the ground.

The image capture unit can capture ultraviolet images or infrared images, and the system assists in determining cloud attributes and analyzing air pollution levels based on the captured ultraviolet images or infrared images.

The device with the image capture unit further includes a cloud height measurement unit, and when capturing the image, the device with the image capture unit marks the orientation in the image file.

A short-range rainfall probability prediction device includes: an image capturing unit, an access unit, a computing unit and a display unit. The image capturing unit is used to capture an image of the sky above the earth. The access unit includes a database that stores condition data corresponding to a plurality of rain probability values.

The computing unit is used to perform image analysis, analyze the cloud amount, cloud type, and cloud color of the captured image; and to judge the probability of rain, and compare the cloud amount, cloud type, and cloud color after the above analysis with those in the database. Compare the data and obtain the rain probability value based on the stored data. The display unit is used to display the current rain probability value or the rain machine for the next 1 hour or 2 hours. rate value.

With the above device, users can make judgments after capturing the image. This device can be an ordinary smart phone or a specially designed device.

10:Image capture unit

20:Access unit

30:Arithmetic unit

40:Display unit

Figure 1 is a flow chart of an embodiment of the present invention.

FIG. 2 is a schematic block diagram of a system according to an embodiment of the present invention.

Figure 3 shows the weather radar echo map.

Figure 4 is a photo of the cloud state.

Figure 5 shows the photo data content of Figure 4 (note of shooting time).

Figure 6 is a screenshot of checking the air pollution status on the website at the same time as Figure 4.

7 to 14 are schematic diagrams showing a rainy state according to an embodiment of the present invention.

Let us first understand here that the so-called rain probability refers to the possibility of rainfall at a certain location. The definition of rainfall is when the cumulative rainfall reaches 0.1mm or more, it is considered to have rainfall. Therefore drizzle is considered rainfall. Please refer to a rain probability prediction method shown in Figure 1, which includes the following steps: a sky image acquisition step S1, an image analysis step S2 and a stored data comparison S3.

Among them, the sky image capturing step S1: capture at least one image toward the sky through an image capturing unit at an angle between 60 degrees and 120 degrees with the ground, and the image capturing unit The wide-angle angle of Yuanzhi lens is between 45 degrees and 180 degrees. The image described is an image captured from the ground to the sky. It can be captured through a colored filter, or it can be captured using visible or invisible near-infrared light, ultraviolet light, and infrared thermal energy. image. Furthermore, the captured images may also include detection of ultraviolet light levels. The distance between each cloud and the shooting device can also be detected, and then converted into the height of the cloud layer. In addition, the captured images may also be several photos taken at different times, or a panoramic photo composed of several photos.

Image analysis step S2: A system analyzes the cloud amount, cloud type and cloud color of the captured image to determine which of the following conditions it belongs to.

(1) Low-level thin clouds: The rainfall pattern tends to be drizzle and light rain because the water droplets are usually small.

(2) Clouds with thicker middle layers: They may contain relatively large water droplets, forming heavy rain, and even ice crystals, which may cause heavy rain, such as hail, snow, and rain plus snow.

(3) When the altostratus clouds are cirrus, cirrocumulus, or cirrostratus, there will be no rainfall.

(4) Cloud color: When it appears dark, it means it contains a high amount of water and the probability of rainfall is high.

Rain probability determination step S3: Compare the cloud amount, cloud type and cloud color after the above analysis with a stored data in the system, and obtain a rain probability value from the stored data.

The stored data includes: the probability of rainfall corresponding to various cloud shapes, cloud colors, and cloud amounts. Furthermore, it can be combined with images detected by near-infrared light and ultraviolet light, or infrared heat energy distribution conditions, or air pollution conditions. influence coefficient, etc.

Among them, the image capturing unit can capture a plurality of images at fixed time intervals within a time interval, thereby obtaining the movement variation of the cloud. Or connect multiple Taiwanese movies Like capture units to capture a wider range of effects simultaneously.

Furthermore, the rain probability determination step S3 may further combine the information of air pressure, temperature, absolute humidity, relative humidity, dew point temperature, wind direction and wind speed of the location detected by the device to perform corresponding determination.

Furthermore, the rain probability determination step S3 may further combine a real-time data captured through the Internet and the date information of the captured image for analysis and determination.

Furthermore, the device with the image capture unit further includes a cloud height measurement unit, and after capturing the image, the device with the image capture unit marks the orientation in the image file to facilitate the determination of cloud movement. direction. In addition, a static electricity measuring device can be further combined to measure the amount of static electricity on the ground. When the measured amount of static electricity on the ground is higher than the normal value, there will be a chance of thunderstorms.

Referring to Figure 2, the present invention provides another embodiment: a rain probability prediction device. The device includes: an image capturing unit 10, an access unit 20, a computing unit 30 and a display unit 40.

The image capture unit 10 is used to capture an image of the sky above the earth, which can be further combined with filters of different colors, or other spectra can be used to capture the image.

The access unit 20 includes a database that stores condition data corresponding to a plurality of rain probability values.

The computing unit 30 is used to perform image analysis, analyze the cloud amount, cloud type, and cloud color of the captured image; and to determine the probability of rain, and compare the cloud amount, cloud type, and cloud color after the above analysis with the database. Compare the data and obtain the rain probability value from the stored data. A display unit 40 is used to display the rain probability value. For its presentation, please refer to Figure 7 to Figure 14 shown.

The device may further include a cloud height measurement unit, an electrostatic quantity detection unit, an ultraviolet quantity detection unit, an infrared detection unit, and an infrared thermal energy detection unit.

In order to make the content of the above methods and devices clearer, further explanation is given below: The system analyzes the cloud attributes, cloud amount, cloud thickness, cloud color and sky brightness of the captured images. The so-called cloud genus refers to which cloud species or mixed cloud species the cloud belongs to. Cloud types are mainly divided into ten types. This includes cumulus, cumulonimbus, altocumulus, stratocumulus, stratus, nimbostratus, altostratus, cirrus, cirrocumulus and cirrostratus. After the cloud genus is confirmed, if it can be further subdivided, it will be further subdivided into cloud species (species), such as fibrous clouds, hook-shaped clouds, dense cirrus clouds, etc., and variations (varieties) such as chaotic clouds, spine clouds, and wavy clouds. wait. In addition, if the supplementary features of clouds such as anvil clouds, mammary clouds, etc. and accessory clouds such as frieze clouds, fragmentary clouds, etc. can be further subdivided, they will be further classified. .

Cloud cover is the proportion of the sky that is filled with clouds. Divide the sky into eight equal parts to estimate the proportion of cloud cover in the entire sky.

The vertical extent of a cloud refers to the thickness of the cloud itself.

The cloud color refers to the degree of black and gray clouds. Generally speaking, thicker clouds mean they contain more water than thin clouds. Because sunlight cannot penetrate thicker cloud bases, the color of the cloud bottoms will appear grayer or darker.

The sky brightness refers to how bright the overall sky is.

Use this device to determine the probability of rain: compare the above analyzed cloud attributes, cloud amount, cloud thickness, cloud color and sky brightness with a stored data in the system, and use various matching algorithms. ) or artificial intelligence algorithm (artificial intelligence algorithm, perform cross comparison and correlation analysis. In addition, please cooperate with Figure 7 and Figure 14. If the clouds are scattered in different positions in the sky, in order to more clearly display which cloud is likely to rain, the device can also specifically mark the clouds in the picture that are more likely to rain (as shown in the picture) box in the formula).

Furthermore, the system further determines the position of the sun in the sky based on time and photos to facilitate analysis of the shooting position and angle.

Furthermore, the system can also distinguish special atmospheric optical phenomena (Atmospheric optics) such as sun halo. This information can also be used as judgment information for weather forecasts. For example, when you see a solar halo, it means there are cirrostratus clouds in the sky. Cirrostratus clouds are made of ice crystals, and it won't rain at the moment. But it may be a sign before a front comes. If the clouds become thicker and lower next, there is a good chance of rain in the future.

Furthermore, the system retrieves meteorological information related to public meteorological observation and forecast data through the Internet. This weather information includes the Meteorological Bureau's regional forecast for your location, the latest surface weather map (see the configuration of high and low pressures, front positions, etc.), the latest upper-altitude weather map (see the configuration of trough lines, etc.), and the latest radar echo map (overall The range and moving direction of cloud and rain areas), the latest satellite cloud images (looking at the moving direction of clouds, overall cloud thickness and overall cloud height, etc.) can be used as data for further analysis. This means that when making rain judgments, you have more information about the current and future trends of the overall weather system. For example: According to the weather map of the Meteorological Bureau, it is shown that a plum rain front will pass through your location in the future. Then you can further confirm whether there is a possibility of immediate rain in the area based on the cloud changes in the photos. Generally, the cloud changes before and after the Meiyu front passes can be roughly divided into the following: cirrocumulus → cirrocumulus → cirrostratus → fortress altocumulus + altocumulus → altostratus → nimbostratus + cumulus. Therefore, it is possible to judge the rainy time by observing changes in clouds.

The system further captures the GPS coordinates, altitude, time, date, ground air pressure, ground temperature, ground absolute humidity, ground dew point temperature, ground relative humidity, ground wind direction and ground wind speed information of the location of the device with the image capture unit. . This ground weather information can be used to determine current and future weather conditions. For example, if the ground temperature is very close to the ground dew point temperature, it means that the moisture content of the air may be close to saturation. So the sky may be about to rain.

The system further produces humidity distribution information above the location. This distribution is not a humidity distribution map calculated from satellite data (satellite water vapor imagery), but uses the attenuation of base station transmission to determine the water vapor in the surrounding environment to produce a small-scale humidity distribution map in this area. If there are enough base stations, this kind of humidity distribution map will be more precise and real-time than that calculated by satellite.

The image can be taken at regular intervals to determine the changes in the clouds above the location, the direction of the clouds and the speed of the clouds in the sky.

The device further includes a cloud height measurement unit.

Among them, the device may further include a cloud height measurement unit, a ground power distribution measurement unit, a red light filter, an ultraviolet light filter, and a near infrared light filter. filter), infrared thermal filter (thermal infrared filter), ultraviolet detection unit, and air pollution status analysis unit.

The ground power distribution condition measuring unit can detect the amount of positive charge on the ground. If the amount of positive charge on the ground tends to increase, it means that the voltage between the ground and the bottom of the cloud is increasing. This phenomenon may represent the development of thunderstorm cells at the location.

The red light filter and the near-infrared filter can respectively use red light and near-infrared light to capture images, thereby obtaining clearer cloud shapes. Red and near-infrared light have relatively long wavelengths. Generally in the sky, because the diameter of particles in the air including O ₂ , NO ₂ , etc. is much smaller than the wavelength of light, the main scattering mechanism is Rayleigh scattering. The intensity of Rayleigh scattering is inversely proportional to the fourth power of the wavelength of visible light (the degree of scattering is ~λ-4), so blue and purple have the strongest scattering, while red light and near-infrared light are less likely to be scattered by particles in the air. Therefore, photos taken with red light and near-infrared light will be slightly clearer than visible light photos, so it is better to distinguish cloud shapes in the sky. If there is a slight haze, the photos taken with red light and near-infrared light will be clearer. However, the effect of this method will be limited in the evening or when there is sunset.

The air pollution condition analysis unit may be a method that captures visible light images, near-infrared light images, and ultraviolet light images respectively to analyze phenomena caused by air pollution factors. This phenomenon is specific to larger aerosol particles, such as pollen, dust or smog. The scattering mechanism at this time is called Mie scattering. When taking pictures of a polluted sky with different bands, you can judge the degree of air pollution based on the degree of fog in different bands, and predict whether the rain will be dirty when it rains. For example, at sunrise or sunset, if the sky is redder, it means the concentration of aerosol particles in the air is higher than usual. Furthermore, the presence of red or orange in the sky is another indicator of air pollution. In addition, inferring whether the inversion layer exists based on the horizontal development of the cloud top can also help determine whether there is air pollution in the air. Basically, ultraviolet light and visible light have shorter wavelengths than infrared light, so when the air is dirty, it is easier to block these wavelength bands, making the photos blurry. Infrared light has a relatively long wavelength, so it is easier to bypass these air pollution problems. The photo will be clearer in comparison. Through the different sharpness of the photos taken in different wavebands, it can be inferred whether the air is dirty or not. If the photos taken by this ultraviolet light, visible light, and infrared light are blurry, it means that the air should be a little dirty.

If the device can obtain air pollution meteorological information while connected to the Internet, it can be combined with the pictures taken for a more complete analysis of air pollution conditions. For example, Figure 4 was taken in Taipei City using the visible light band, and you can see that there is a bit of haze in the sky. According to the air quality conditions of the Meteorological Bureau at that time, it is true that the air above Taipei City was a little bad. The sky above Taipei City appears dark yellow. Figure 5 is the archive time of Figure 4, and Figure 6 is a screenshot of the air quality data at this time.

The ultraviolet detection unit can detect the amount of ultraviolet light detected on the ground, thereby determining whether the ultraviolet rays are currently strong or not and assisting in determining cloud type.

The infrared thermal energy filter can be a method that uses infrared thermal energy photos to determine the cloud type, cloud amount, cloud thickness, and cloud color of clouds in the sky when there is no sunlight.

Furthermore, a measuring unit for measuring the distribution of electrostatic energy on the ground can be further combined, which can be used to detect the amount of positive charge on the ground. When the amount of positive charge on the ground tends to increase, it means that there is a trend of increasing voltage between the ground and the bottom of the cloud. This phenomenon may represent the development of thunderstorm cells at the location and can be used as a basis for auxiliary judgment.

Furthermore, the present invention can determine the air pollution status based on previously captured photos, and combined with the rainfall forecast, can display in the report whether the rainwater contains higher concentrations of pollutants.

In addition, if it is at night or when the sun is completely blocked, near-infrared rays can be used to capture images, and then analysis can be performed based on the captured images.

S1: Sky image acquisition steps

S2: Image analysis steps

S3: Store data for comparison

Claims (3)

A short-range rainfall probability prediction method, which includes the following steps: a sky image capturing step, capturing at least one image toward the sky at an angle between 60 degrees and 120 degrees from the ground through a device of an image capturing unit, and the The wide-angle angle of the lens of the image capture unit is between 45 degrees and 180 degrees. In addition, the image capture unit can capture multiple images at fixed time intervals within a time interval to obtain the moving direction of the clouds. ; An image analysis step, in which a system analyzes the cloud amount, cloud type, cloud color and cloud movement direction contained in the captured image, wherein the system is connected to the device signal with the image capture unit , and the cloud amount is the proportion of the sky filled with clouds; in a rain probability judgment step, the cloud amount, cloud type or cloud color after the above analysis is compared with a stored data in the system, and the corresponding stored data is obtained A rain probability value is obtained, and then it is determined whether to move to the position of the device with the image capture unit based on the moving direction of the cloud. The stored data includes: the rainfall probability corresponding to cloud shape, cloud color and cloud amount. The short-range rainfall probability prediction method as described in claim 1, wherein a plurality of image acquisition units can be further connected to perform image acquisition at the same time. A short-range rainfall probability prediction device, which includes: an image capture unit used to capture an image of the sky above the ground, and the wide-angle angle of the lens of the image capture unit is between 45 degrees and 180 degrees, In addition, the image capture unit can capture a plurality of images at fixed time intervals within a time interval, thereby obtaining the moving direction of the cloud; an access unit includes a database that stores a plurality of strokes The stored data corresponding to the rain probability value. The stored data includes: cloud shape, cloud color, and rainfall corresponding to cloud amount. Probability; a computing unit used to perform image analysis, analyze the cloud amount, cloud type, and cloud color of the captured image; and to judge the probability of rain, and combine the cloud amount, cloud type, and cloud color after the above analysis with Compare the data in the database, obtain a rain probability value from the stored data, and then determine whether to move to the location of the device with the image capture unit based on the moving direction of the cloud, where the cloud amount is The proportion of the sky filled with clouds; a display unit used to display the rain probability value.

Images