Meteorological radar precipitation pattern prediction method and apparatus

• US 6,128,578 A
• Filed: 11/04/1997
• Issued: 10/03/2000
• Est. Priority Date: 12/26/1996
• Status: Expired due to Term

• Current Assignee: Nippon Telegraph and Telephone Corporation
• Original Assignee: Nippon Telegraph and Telephone Corporation
• Inventors: Sakaino, Hidetomo, Suzuki, Satoshi, Horikoshi, Tsutomu
• Primary Examiner(s): McElheny, Jr., Donald E.
• Application Number:
  US08/964,470
• Time in Patent Office: 1,064 Days
• Field of Search: 702/2, 702/3, 702/4, 706/931, 342/26
• US Class Current: 702/3
• CPC Class Codes: G01S 13/951 ground based
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: Y02A 90/10 Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

Meteorological radar precipitation pattern prediction methods and apparatus are provided, by which operational costs can be reduced, a series of physical processes in a precipitation region and topographical influences can be considered. The main method has the steps of inputting radar images of a precipitation region storing the input radar images as time-series two-dimensional images; calculating various image-feature quantities with respect to two or more of the stored two-dimensional past images; calculating a spatial-temporal transition of the amount of precipitation using an advection-diffusion equation system which indicates various physical effects relating to precipitation phenomena, with the various image-feature quantities as initial values; predicting transition of a pattern with respect to the precipitation region based on results of the calculation; and outputting predicted results as time-series images. Preferably, the advection-diffusion equation system includes time, advection, diffusion, source, sink, and dispersion terms, and a gray-level at each pixel is supplied to the equation as a variable corresponding to an amount of precipitation. If the advection-diffusion equation system includes an advection term which is a product of an advection vector and the first derivative with respect to a gray-level of a relevant pixel in a precipitation pattern, the amount of precipitation is varied according to transition of advection vectors.

Claims

1. A meteorological radar precipitation pattern prediction apparatus comprising:

image input means for inputting radar images of a precipitation region;

image storage means for storing the input radar images as time-series two-dimensional images;

image processing means for calculating various image-feature quantities with respect to two or more two-dimensional past images stored in the image storage means;

calculation means for calculating a spatial-temporal transition of the amount of precipitation using an advection-diffusion equation system which indicates various physical effects relating to precipitation phenomena, with the various image-feature quantities as initial values;

prediction means for predicting transition of a pattern with respect to the precipitation region based on results calculated by the calculation means; and

output means for outputting predicted results as time-series images.

2. An apparatus as claimed in claim 1, wherein the advection-diffusion equation system includes diffusion coefficients and evaluated advection vectors, and the apparatus further comprises:

topographical effect introducing means for setting the diffusion coefficients and the advection vectors in accordance with topographical features of the precipitation region.

3. An apparatus as claimed in claim 1, wherein the advection-diffusion equation system includes an advection term which is a product of an advection vector and the first derivative with respect to a gray-level of a relevant pixel.

4. An apparatus as claimed in claim 1, wherein the image processing means extracts a region having gray-levels greater than a fixed level from each input two-dimensional image and counts the number of pixels belonging:

to the extracted region, and determines the counted number of pixels as an area of a precipitation region in the relevant frame.

5. An apparatus as claimed in claim 1, wherein the advection-diffusion equation system includes time, advection, diffusion, source, sink, and dispersion terms, and a gray-level at each pixel is supplied to the equation as a variable corresponding to an amount of precipitation.

6. An apparatus as claimed in claim 5, wherein diffusion coefficients included in the diffusion term are set in accordance with a rate of change with respect to one-dimensional area transition.

7. An apparatus as claimed in claim 6, wherein one-dimensional non-linear area transition is predicted using extrapolation by approximately fitting a predetermined non-linear function up to a predetermined prediction time, whereby the rate of change with respect to one-dimensional area transition is obtained and corresponding diffusion coefficients are changed according to the obtained rate.

8. An apparatus as claimed in claim 5, wherein data relating to positive and negative regions which are determined by calculating difference between two of the two-dimensional past images are supplied to the source and sink terms of the equation, respectively.

9. An apparatus as claimed in claim 1, wherein the image-feature quantities include an edge-gradient which is the first derivative of an gray-level.

10. An apparatus as claimed in claim 9, wherein the image processing means divides the pattern of the precipitation region into one or more regions having a large edge-gradient and one or more regions having a small edge-gradient with respect to a predetermined value.

11. An apparatus as claimed in claim 5, wherein the image processing means divides the pattern of the precipitation region into one or more regions having a large edge-gradient and one or more regions having a small edge-gradient with respect to a predetermined value, calculates the center of gravity of each region having the large edge-gradient, and weights an amount of sinking relating to the sink term according to a distance from the center of gravity, and in proportion to the distance from the center of gravity, a larger dispersion coefficient is supplied to the dispersion term.

12. An apparatus as claimed in claim 10, wherein the image processing means sets advection vectors supplied to the advection-diffusion equation system by extracting a shift of the center of gravity of each region having the large edge-gradient.

13. An apparatus as claimed in claim 12, wherein if advection in the relevant precipitation region is strong, the image processing means also extracts a shift of the center of gravity of each region having the small edge-gradient so as to set the advection vectors.

14. An apparatus as claimed in claim 3, wherein the image processing means determines the advection vectors of the pattern of the precipitation region by dividing each frame into plural small blocks and calculating a direction and a distance relating to points having the highest similarity between two frames of the images based on the Cross Correlation method, and evaluates overall advection vectors throughout the frame by iterating an operation using a moving average filter.

15. An apparatus as claimed in claim 3, wherein the image processing means evaluates the advection vectors in the image from a precipitation pattern by extracting contours of the precipitation pattern in two of the images;

assuming displacements between the contours as advection vectors; and

propagating the advection vectors by iterating an operation using a moving average filter.

16. An apparatus as claimed in claim 3, wherein the image processing means evaluates the advection vectors in the image from a precipitation pattern by extracting contours of the precipitation pattern in two of the images;

assuming displacements between the contours as advection vectors; and

substituting the advection vectors as initial values of velocity vectors of a predetermined fluid equation and iterating time integral of the equation whereby velocity vectors are interpolated throughout the image, and the image processing means inputs evaluated velocity vectors based on the interpolation into the advection-diffusion equation system.

17. An apparatus as claimed in claim 15 or 16, wherein in the detection of the contours and the displacements therebetween:

image processing operations including binarization;

labeling;

elimination of isolated points; and

erosion/dilation are previously performed for the pattern of the precipitation region;

a discrete contour is detected using a contour-searching algorithm; and

regarding each contour point on the discrete contour, a displacement of the contour is calculated based on a distance, in the direction of a normal with respect to the contour in a past frame, from the contour point of the contour in the past frame to an intersection on a corresponding contour in the present frame.

18. An apparatus as claimed in claim 17, wherein the normal of the contour is defined using the central difference method.

19. An apparatus as claimed in claim 16, wherein the predetermined fluid equation is the Navier-Stokes equation.

20. An apparatus as claimed in claim 16, wherein the predetermined fluid equation is the Burgers equation.

21. An apparatus as claimed in claim 1, wherein the advection-diffusion equation system is transformed into a discrete form and is solved using, any one of the finite difference method, the finite element method, and the spectral method.

22. An apparatus as claimed in claim 2, wherein the topographical effect introducing means sets the advection vectors supplied to the advection-diffusion equation system in accordance with a topographical feature of the precipitation region.

23. An apparatus as claimed in claim 2, wherein the topographical effect introducing means sets the diffusion coefficients in accordance with a topographical feature of the precipitation region.

24. An apparatus as claimed in claim 1, wherein the advection-diffusion equation system includes dispersion coefficients and the topographical effect introducing means sets the dispersion coefficients in accordance with a topographical feature of the precipitation region.

25. An apparatus as claimed in claim 24, wherein if the pattern of the precipitation region passes through both ocean and land regions, larger dispersion coefficients are set for the ocean region in comparison with those set for the land region.

26. An apparatus as claimed in claim 1, wherein the calculation means iterates time integral of the advection-diffusion equation system with a time span shorter than a relevant observation period of the radar.

27. An apparatus as claimed in claim 10, wherein the image processing means determines advection vectors supplied to the advection-diffusion equation system based on a precipitation pattern, and in the determination:

first initial advection vectors are determined by extracting a change of the center of gravity of each region having the large edge-gradient;

second initial advection vectors are determined by dividing each frame into plural small blocks and calculating a direction and a distance relating to points having the highest similarity between two frames of the images based on the cross correlation method;

third initial advection vectors are determined by extracting contours of the precipitation pattern in two of the images and assuming displacements between the contours as the advection vectors; and

an operation using a moving average filter is iterated with respect to the first, second and third initial advection vectors so as to evaluate overall advection vectors throughout the frame.

40. An apparatus as claimed in any one of claims 1, 28, 31, and 35, wherein the two or more two-dimensional past images are selected according to a sampling rate and a moving velocity of a relevant object.

28. A meteorological radar precipitation pattern prediction apparatus comprising:

image input means for inputting radar images of a precipitation region;

image storage means for storing the input radar images as time-series two-dimensional images;

image processing means for outputting area-transition information of a precipitation region with respect to two or more two-dimensional past images stored in the image storage means as one-dimensional time-series signals;

function fitting means for fitting a non-linear function into the one-dimensional time-series signals by using the least square method;

prediction means for predicting future area-transition based on results of the function fitting; and

output means for outputting a pattern transition of the precipitation region as time-series images based on predicted results.

29. An apparatus as claimed in claim 28, wherein the non-linear function is a quadratic function.

30. An apparatus as claimed in claim 28, wherein the prediction means predicts a maximum value at a switching point from increase to decrease of the area and/or a minimum value at a switching point from decrease to increase of the area.

31. A meteorological radar precipitation pattern prediction apparatus comprising:

image input for inputting radar images of a precipitation region;

image storage means for storing the input radar images as time-series two-dimensional images;

image transition detecting means for detecting an intensity transition between two or more two-dimensional past images stored in the image storage means;

image generating and vanishing means for generating or vanishing a local precipitation-region image according to the detected intensity transition by using an image processing method; and

prediction means for predicting transition of a pattern with respect to the precipitation region based on results image-processed by the image generating and vanishing means; and

output means for outputting predicted results as time-series images.

32. An apparatus as claimed in claim 31, wherein the image transition detecting means detects the intensity transition using an evaluation equation in which differences between the frames are summed.

33. An apparatus as claimed in claim 31, wherein the image generating and vanishing means iterates convolution integral calculation and inverse integral calculation thereof with respect to a Gaussian function and image data of the precipitation region.

34. An apparatus as claimed in claim 33, wherein the image generating and vanishing means provides intensity information for maintaining gray-levels of the image at a fixed level for each iteration.

35. A meteorological radar precipitation pattern prediction apparatus comprising:

image input means for inputting radar images of a precipitation region;

image storage means for storing the input radar images as time-series two-dimensional images;

image processing means for detecting and outputting intensity and area transitions between two or more two-dimensional past images stored in the image storage means as one-dimensional time-series signals;

precipitation-region growth and decay predicting means for predicting transition with respect to growth and decay of the precipitation region based on the one-dimensional time-series signals; and

output means for outputting a pattern transition of the precipitation region as time-series images based on predicted results.

36. An apparatus as claimed in claim 35, wherein the image processing means divides each two-dimensional image into plural small blocks, and detects the area and intensity transition for each divided block.

37. An apparatus as claimed in claim 35, wherein the image processing means calculates the one-dimensional time-series signals by performing time-series calculation relating to each of area and intensity differences between images which are a predetermined number of frames apart from each other.

38. An apparatus as claimed in claim 37, wherein the image processing means determines a function of the intensity transition by calculating the first derivative with respect to each calculated difference.

39. An apparatus as claimed in claim 35, wherein the precipitation-region growth and decay predicting means calculates cross-correlation values relating to the area and intensity transitions by using a cross-correlation function and evaluates a time lag between both transitions.

41. A meteorological radar precipitation pattern prediction method comprising the steps of:

inputting radar images of a precipitation region;

storing the input radar images as time-series two-dimensional images;

calculating various image-feature quantities with respect to two or more of the stored two-dimensional past images;

calculating a spatial-temporal transition of the amount of precipitation using an advection-diffusion equation system which indicates various physical effects relating to precipitation phenomena, with the various image-feature quantities as initial values;

predicting transition of a pattern with respect to the precipitation region based on results of the calculation; and

outputting predicted results as time-series images.

42. A method as claimed in claim 41, wherein the advection-diffusion equation system includes diffusion coefficients and evaluated advection vectors, and the method further comprises the step of setting the diffusion coefficients and the advection vectors in accordance with topographical features of the precipitation region.

43. A method as claimed in claim 41, wherein the advection-diffusion equation system includes an advection term which is a product of an advection vector and the first derivative with respect to a gray-level of a relevant pixel.

44. A method as claimed in claim 41, wherein the step of calculating the image-feature quantities includes extracting a region having gray-levels greater than a fixed level from each input two-dimensional image and counting the number of pixels belonging to the extracted region, and determining the counted number of pixels as an area of a precipitation region in the relevant frame.

45. A method as claimed in claim 41, wherein the advection-diffusion equation system includes time, advection, diffusion, source, sink, and dispersion terms, and a gray-level at each pixel is supplied to the equation as a variable corresponding to an amount of precipitation.

46. A method as claimed in claim 45, wherein diffusion coefficients included in the diffusion term are set in accordance with a rate of change with respect to one-dimensional area transition.

47. A method as claimed in claim 46, wherein one-dimensional non-linear area transition is predicted using extrapolation by approximately fitting a predetermined non-linear function up to a predetermined prediction time, whereby the rate of change with respect to one-dimensional area transition is obtained and corresponding diffusion coefficients are changed according to the obtained rate.

48. A method as claimed in claim 45, wherein data relating to positive and negative regions which are determined by calculating difference between two of the two-dimensional past images are supplied to the source and sink terms of the equation, respectively.

49. A method as claimed in claim 41, wherein the image-feature quantities include an edge-gradient which is the first derivative of an gray-level.

50. A method as claimed in claim 49, wherein the step of calculating the image-feature quantities includes dividing the pattern of the precipitation region into one or more regions having a large edge-gradient and one or more regions having a small edge-gradient with respect to a predetermined value.

51. A method as claimed in claim 45, wherein the step of calculating the image-feature quantities includes dividing the pattern of the precipitation region into one or more regions having a large edge-gradient and one or more regions having a small edge-gradient with respect to a predetermined value, calculating the center of gravity of each region having the large edge-gradient, and weighting an amount of sinking relating to the sink term according to a distance from the center of gravity, and in proportion to the distance from the center of gravity, a larger dispersion coefficient is supplied to the dispersion term.

52. A method as claimed in claim 50, wherein advection vectors supplied to the advection-diffusion equation system are set by extracting a shift of the center of gravity of each region having the large edge-gradient.

53. A method as claimed in claim 52, wherein if advection in the relevant precipitation region is strong, a shift of the center of gravity of each region having the small edge-gradient is also extracted so as to set the advection vectors.

54. A method as claimed in claim 43, wherein the step of calculating the image-feature quantities includes determining the advection vectors of the pattern of the precipitation region by dividing each frame into plural small blocks and calculating a direction and a distance relating to points having the highest similarity between two frames of the images based on the Cross Correlation method, and evaluating overall advection vectors throughout the frame by iterating an operation using a moving average filter.

55. A method as claimed in claim 43, wherein the step of calculating the image-feature quantities includes evaluating the advection vectors in the image from a precipitation pattern by extracting contours of the precipitation pattern in two of the images;

assuming displacements between the contours as advection vectors; and

by propagating the advection vectors by iterating an operation using a moving average filter.

56. A method as claimed in claim 43, wherein the step of calculating the image-feature quantities includes evaluating the advection vectors in the image from a precipitation pattern by extracting contours of the precipitation pattern in two of the images;

assuming displacements between the contours as advection vectors; and

by substituting the advection vectors as initial values of velocity vectors of a predetermined fluid equation and iterating time integral of the equation whereby velocity vectors are interpolated throughout the image, and the step further includes inputs evaluated velocity vectors based on the interpolation into the advection-diffusion equation system.

57. A method as claimed in claim 55 or 56, wherein in the detection of the contours and the displacements therebetween:

image processing operations including binarization;

libeling;

elimination of isolated points; and

erosion/dilation are previously performed for the pattern of the precipitation region;

a discrete contour is detected using a contour-searching algorithm; and

regarding each contour point on the discrete contour, a displacement of the contour is calculated based on a distance, in the direction of a normal with respect to the contour in a past frame, from the contour point of the contour in the past frame to an intersection on a corresponding contour in the present frame.

58. A method as claimed in claim 57, wherein the normal of the contour is defined using the central difference method.

59. A method as claimed in claim 56, wherein the predetermined fluid equation is the Navier-Stokes equation.

60. A method as claimed in claim 56, wherein the predetermined fluid equation is the Burgers equation.

61. A method as claimed in claim 41, wherein the advection-diffusion equation system is transformed into a discrete form and is solved using any one of the finite difference method, the finite element method, and the spectral method.

62. A method as claimed in claim 42, wherein the step of setting the diffusion coefficients and the advection vectors includes setting the advection vectors supplied to the advection-diffusion equation system in accordance with a topographical feature of the precipitation region.

63. A method as claimed in claim 42, wherein the step of setting the diffusion coefficients and the advection vectors includes setting the diffusion coefficients in accordance with a topographical feature of the precipitation region.

64. A method as claimed in claim 41, wherein the advection-diffusion equation system includes dispersion coefficients and the step of setting the diffusion coefficients and the advection vectors includes setting the dispersion coefficients in accordance with a topographical feature of the precipitation region.

65. A method as claimed in claim 64, wherein if the pattern of the precipitation region passes through both ocean and land regions, larger dispersion coefficients are set for the ocean region in comparison with those set for the land region.

66. A method as claimed in claim 41, wherein the step of calculating a spatial-temporal transition using the advection-diffusion equation system includes iterating time integral of the advection-diffusion equation system with a time span shorter than a relevant observation period of the radar.

67. A method as claimed in claim 50, wherein the step of calculating the image-feature quantities includes determining advection vectors supplied to the advection-diffusion equation system based on a precipitation pattern, and in the determination:

first initial advection vectors are determined by extracting a change of the center of gravity of each region having the large edge-gradient;

second initial advection vectors are determined by dividing each frame into plural small blocks and calculating a direction and a distance relating to points having the highest similarity between two frames of the images based on the cross correlation method;

third initial advection vectors are determined by extracting contours of the precipitation pattern in two of the images and assuming displacements between the contours as the advection vectors; and

an operation using a moving average filter is iterated with respect to the first, second and third initial advection vectors so as to evaluate overall advection vectors throughout the frame.

71. A method as claimed in one of claims 41 and 68, wherein the two or more two-dimensional past images are selected according to a sampling rate and a moving velocity of a relevant object.

74. A storage medium storing a computer program for making a computer execute any one of the methods claimed in claims 41, 68, and 73.

68. A meteorological radar precipitation pattern prediction method comprising the steps of:

inputting radar images of a precipitation region;

storing the input radar images as time-series two-dimensional images;

outputting area-transition information of a precipitation region with respect to two or more of the stored two-dimensional past images as one-dimensional time-series signals;

fitting a non-linear function into the one-dimensional time-series signals by using the least square method;

predicting future area-transition based on results of the function fitting; and

outputting a pattern transition of the precipitation region as time-series images based on predicted results.

69. A method as claimed in claim 68, wherein the non-linear function is a quadratic function.

70. A method as claimed in claim 68, wherein the step of predicting future area-transition includes predicting a maximum value at a switching point from increase to decrease of the area and/or a minimum value at a switching point from decrease to increase of the area.

72. A gas-distribution pattern prediction apparatus comprising:

image input means for inputting gas-distribution images of gas distributed in the air;

image storage means for storing the input images as time-series two-dimensional images;

image processing means for calculating various image-feature quantities with respect to two or more two-dimensional past images stared in the image storage means;

calculation means for calculating a spatial-temporal transition of the distributed gas using an advection-diffusion equation system which indicates various physical effects relating to gas discharged into the air, with the various image-feature quantities as initial values;

prediction means for predicting transition of a pattern with respect to the gas distribution based on results calculated by the calculation means; and

output means for outputting predicted results as time-series images.

73. A gas-distribution pattern prediction method comprising the steps of:

inputting gas-distribution images of gas distributed in the air;

storing the input images as time-series two-dimensional images;

calculating various image-feature quantities with respect to two or more of the stored two-dimensional past images;

calculating a spatial-temporal transition of the distributed gas using an advection-diffusion equation system which indicates various physical effects relating to gas discharged into the air, with the various image-feature quantities as initial values;

predicting transition of a pattern with respect to the gas distribution based on results of the calculation; and

outputting predicted results as time-series images.