JP2012013507A - Light rain detection method, light rain detection device, and light rain detection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable quantification of an amount of light rain.SOLUTION: Captured image data of a wave pattern on a surface of water is fed to an objective function derived with constraints: a dispersion relation of approximation of the wave pattern; and diffusion. The objective function is solved as a minimization problem for velocity and angular frequency, so that a velocity (flow) is found. Then, an average flow is calculated, and a table showing a relationship between the average flow and the degree of the light rain is referred. This enables quantification of a degree of the light rain, while suppressing influence of disturbance caused by wind or the like in the image data.

Description

  The present invention relates to a technique for detecting the degree of light rain by image analysis.

  The weather information is observed utilizing the characteristics of various sensors, cameras, and satellites. Among them, information on rain is one of the most needed in daily life. In the AMeDAS network, a certain amount of rainfall and rain area are observed by a rain gauge, and it is possible to understand the state of rain in a wider area than the precipitation radar.

Supervised by Digital Image Processing Editorial Board, “Digital Image Processing”, 2nd edition, Foundation for Promotion of Image Information Education, March 2006 "Dispersion relation", [online], [Search June 23, 2010], Internet <URL: http://en.wikipedia.org/wiki/Dispersion_relation>

  However, when the amount of precipitation is less than several millimeters per hour, such as so-called light rain, sensing is not performed and it is considered as an error or is visually observed. Therefore, it has not been quantified and a simple observation method has been demanded.

  On the other hand, in image processing, various analysis methods have been shown for the surface (see Non-Patent Document 1), but ripple analysis between light rain and the water surface has not been shown. In computer graphics, a simulation using a wave equation or the like was shown, but model parameters were not estimated from images, and the aim was exclusively to generate images. In the case of image processing, a wind ripple is generated in the real environment, so it is necessary to distinguish it from a ripple caused by light rain.

  The present invention has been made in view of the above, and an object thereof is to quantify the amount of light rain.

  According to a first aspect of the present invention, there is provided a light rain detection method comprising: inputting image data obtained by capturing ripples; storing the image data in a storage unit; reading the image data from the storage unit; Substituting into the objective function derived as the constraint condition and solving the objective function as a minimization problem, calculating the optical flow of the ripples, calculating the average value of the optical flow, and based on the average value, And a step of determining the degree.

  According to a second aspect of the present invention, there is provided a light rain detection apparatus comprising: an input unit that inputs image data obtained by capturing ripples; a storage unit that stores the image data; and a dispersion that approximates ripples by reading the image data from the storage unit. Substituting into the objective function derived from the relational expression, diffusion as a constraint, and solving the objective function as a minimization problem, an analysis means for estimating the optical flow of ripples, and calculating the average value of the optical flow, Determining means for determining the degree of light rain based on the value.

  A light rain detection program according to a third aspect of the present invention causes a computer to execute the light rain detection method.

  According to the present invention, the amount of light rain can be quantified.

It is a functional block diagram which shows the structure of the light rain detection apparatus in this Embodiment. It is a schematic diagram which shows the mode of the ripple which the light rain which fell to the puddle makes. It is a figure which compares the result of the flow estimation when there exists disturbance by a wind etc. with the conventional method. It is a graph which shows the relationship between an average flow and the extent of light rain.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a functional block diagram showing the configuration of the light rain detection apparatus in the present embodiment. The light rain detection apparatus 100 shown in the figure includes an image input unit 110, a data storage unit 120, a wavefront shape analysis unit 130, a collation determination unit 140, and a display unit 150. Each part with which the light rain detection apparatus 100 is provided is good also as what is comprised by the computer provided with the arithmetic processing unit, the memory | storage device, etc., and the process of each part is performed by a program. This program is stored in a storage device included in the light rain detection apparatus 100, and can be recorded on a recording medium such as a magnetic disk, an optical disk, or a semiconductor memory, or provided through a network.

  The image input unit 110 inputs time-series image data obtained by photographing the spread of ripples on the water surface with a camera and stores the data in the data storage unit 120. The spread of ripples on the surface of the water is caused by light rain falling in a fixed container installed outdoors.

  The wavefront shape analysis unit 130 derives an objective function using a dispersion relational expression approximating a ripple and diffusion as a constraint condition, and inputs two continuous image data read from the data storage unit 120 into the objective function, and calculates the velocity and angularity. Solve for frequency as a minimization problem to find velocity (flow).

  The collation determination unit 140 calculates the average of the flow obtained by the wavefront shape analysis unit 130, and quantifies the amount of light rain by referring to the calculated average flow with a table accumulated in advance.

  Display unit 150 displays the determination result of collation determination unit 140.

  Note that the process of the light rain detection device 100 is performed in the order of the image input unit 110, the data storage unit 120, the wavefront shape analysis unit 130, the collation determination unit 140, and the display unit 150 described above.

  Next, details of the processing of the wavefront shape analysis unit 130 will be described.

  FIG. 2 is a schematic diagram showing a ripple pattern created by light rain falling in a puddle when there is no wind. When the rain 210 falls into the puddle 200, a plurality of ripples 220 are formed. Adjacent ones of the ripples 220 form a complex pattern one after another.

  In order to detect such a pattern in image processing, the simplest method is to apply a spatial first derivative for each image (Non-Patent Document 1), or to create a time difference image from a time-series image. is there.

  In the case of light rain, the falling light rain is transparent in the image or is often not reflected due to a lack of camera resolution, so the pattern obtained from the image data input by the image input unit 110 is a pattern on the water surface. In other words, it is considered to be a pattern of ripples 220 formed by rain 210 colliding with the puddle 200.

The pattern of the ripples 220 may be an attenuation model in which the amplitude of the wave at the center of the concentric circle that is the collision point of the rain 210 is large, and the amplitude attenuates in a vibrational manner around the periphery. At this time, when the ripple 220 is regarded as one characteristic of the wave, it can be approximated by a dispersion relational expression (see Non-Patent Document 2) of the following expression (1).

  Here, ω, g, and k are angular frequency, gravitational acceleration, and wave number, respectively. In other words, this relational expression shows that there is a physical relationship between the wavelength of the wave and the (phase) speed, that the longer the wavelength, the slower the speed, and the shorter the wavelength, the faster the speed. In other words, the attenuation is within the range where the dispersion relational expression holds. Therefore, a feature that can be applied to this dispersion relational expression from the image data can be regarded as the ripple 220.

  Next, the apparent speed of the ripple is estimated by optical flow. At this time, Expression (1) is set as a constraint condition. In addition, the relation between the velocity and the angular velocity and the constraint condition for the ripples to diffuse are given.

Two continuous image data are read from the data storage unit 120. Let the luminance of the pixel of the image data be I (x, y, t). (X, y) represents a position, and t represents time. By minimizing the objective function of the following expression (2) with respect to the velocity (u, v) and the angular frequency ω = (ω _x , ω _y ), the velocity on the water surface can be obtained. λ ₁ , λ ₂ , and λ ₃ are parameters representing the degree of contribution of Equation (1), and are 1.0 here. Angular frequency is the product of velocity and wave number. The obtained speed is called an optical flow (flow).

  FIG. 3 is a diagram illustrating a flow estimation result when there is a disturbance due to wind or the like.

  In the image data 300 of FIG. 3, in addition to ripples due to rain, disturbances are generated on the water surface by the wind, and the water surface is complicated. When motion is detected by the conventional flow method (Non-Patent Document 1), the motion is disturbed as shown by the flow detection result indicated by reference numeral 310.

  On the other hand, when a motion is detected by the method according to the present embodiment, a ripple characteristic spreading from one point is obtained as in the flow detection result indicated by reference numeral 320.

  Next, details of the processing of the collation determination unit 140 will be described.

  FIG. 4 is a graph showing the relationship between the average flow and the degree of light rain.

  The collation determination unit 140 calculates the average flow by summing the flows estimated by the wavefront shape analysis unit 130 and dividing the sum by the size of the image data (for example, 100 × 100). The collation determination unit 140 holds a table indicating the relationship between the average flow based on the relationship shown in FIG. 4 and the degree of light rain. By referring to the table, the magnitude of the calculated average flow can be determined for the degree of light rain. For example, it is quantified as 1-5.

  It is also possible to use a viscous fluid other than water. When the viscous fluid is put in a container installed outdoors and the flow is obtained by the above method, the equation (2) can be applied as it is by adjusting the viscosity coefficient (0.0001 or the like) of the viscous fluid. By using a viscous fluid instead of water, it is effective for limiting the place where there is vibration and the magnitude of light rain to be detected to a certain level or more.

  Moreover, the influence of a wind can be relieved by surrounding a container with a wall. In this case, the camera is disposed outside the container.

  As described above, according to the present embodiment, image data obtained by capturing ripples on the water surface is input to a dispersion relational expression that approximates ripples, and an objective function that is derived using diffusion as a constraint, and the minimum is obtained with respect to velocity and angular frequency. By calculating the speed (flow) by solving as a problem, calculating the average of the flow, and referring to a table showing the relationship between the average flow and the degree of light rain, the influence of wind disturbance in the image data is suppressed. However, the degree of light rain can be quantified.

DESCRIPTION OF SYMBOLS 100 ... Light rain detection apparatus 110 ... Image input part 120 ... Data storage part 130 ... Wavefront shape analysis part 140 ... Collation determination part 150 ... Display part

Claims (5)

Inputting image data in which ripples are photographed and storing the image data in a storage means;
Reading the image data from the storage means, substituting the dispersion relational expression approximating the ripple, and the objective function derived from diffusion as a constraint, and estimating the optical flow of the ripple by solving the objective function as a minimization problem When,
Calculating an average value of the optical flow and determining a degree of light rain based on the average value;
A method for detecting light rain, comprising: The ripples are observed on the surface of the viscous fluid placed in the container,
The light rain detection method according to claim 1, wherein the analyzing step estimates an optical flow in consideration of a viscosity coefficient of the viscous fluid. An input means for inputting image data obtained by photographing a ripple,
Storage means for storing the image data;
Analysis of reading the image data from the storage means, substituting into a dispersion relational expression approximating ripples, and an objective function derived from diffusion as a constraint, and estimating the optical flow of ripples by solving the objective function as a minimization problem Means,
A determination means for calculating an average value of the optical flow and determining a degree of light rain based on the average value;
A light rain detection apparatus comprising: The ripples are observed on the surface of the viscous fluid placed in the container,
4. The light rain detection apparatus according to claim 3, wherein the analysis means estimates an optical flow in consideration of a viscosity coefficient of the viscous fluid.   A light rain detection program for causing a computer to execute the light rain detection method according to claim 1.

Images