TWI650529B - Mountain flood detection method - Google Patents

Abstract

A mountain flood detection method for detecting mountain floods via an image processing technology includes: capturing an image of a water surface in an area of a river by an image capturing device to generate a water surface image; and graying the water surface image, A gray-scale image is generated, and an identification area is selected for analysis in the gray-scale image; the pixel of the identification area is divided into a water body pixel and a non-water body pixel by a group algorithm; and the water body in the identification area is calculated. A percentage of the pixel, and determine whether the percentage is not less than a reference value. If the judgment result is yes, it indicates that a mountain flood occurs.

Description

Mountain flood detection method

The invention relates to a mountain flood detection method, in particular to a method for detecting mountain floods via image processing technology.

According to the intensive rains during the rainy season and the typhoon, it is easy to cause flooding in the mountainous areas. When the mountain area contains a lot of sand and gravel, it will cause a large amount of floods to entrain the soil and gravel to form a rock flow, which will wash the slopes and rivers and threaten the people. Life and property safety, therefore, relevant disaster prevention units (such as: Soil and Water Conservation Bureau) set up relevant monitoring instruments for key disaster areas of earth and rock flow, such as: cable detectors or ground sound detectors, etc., so that when the earth and rock flow occurs, Immediate warnings are issued to inform the people to evacuate.

In particular, the cable detector is placed above a specific elevation on the riverbed and across the river. When the earth-rock flow occurs and exceeds this specific elevation, the cable of the cable detector can be destroyed by the earth-rock flow carried by the flood. To provide warnings; the ground sound detectors are used to detect the vibration characteristics caused by the earth and stone flow, supplemented by time-frequency analysis technology to determine whether it is caused by earth and rock flow. However, the method of using the wire rope detector must be used when the earth and stone flow reaches a certain scale (when the cable is broken), so that the warning can not be issued early, and the cable is a one-time detection instrument, which must be used after use. Reset, it is inconvenient for areas with frequent disasters. In addition, although the use of geoacoustic detectors can solve this problem, it is still missing by the earth and stone flow, or by the external force and misjudgment.

Therefore, the present invention proposes to monitor photography by means of a non-contact method The machine detects the occurrence of floods in the mountains, where the surveillance camera can use Closed Circuit Television (CCTV), which is erected by the Soil and Water Conservation Bureau in the mountains. Surveillance cameras, such as: downstream water level measurement and urban flooding height measurement, examples can be considered as: Republic of China No. 101140546 "water level measurement method" patent case, the patent case using surveillance cameras Obtain the image and analyze the water level in the image. However, the patent case must rely on a flat wall or pier as a water level judgment block to convert the pixel size to the actual size, so it is not suitable for use in a rugged mountainous environment to determine the height of the water level in the mountain river.

In view of this, the present invention provides a mountain flood detection method to provide a flood prevention application for river water levels.

In order to solve the above problems, an object of the present invention is to provide a mountain flood detection method capable of detecting the occurrence of mountain floods through image processing technology.

The invention provides a mountain flood detection method, which uses an image capturing device to shoot a water surface of a river in a mountainous area to generate a water surface image. The image capturing device is a closed circuit television installed in the mountain area, and the photographing rate is One water surface image per minute; grayscale processing on the water surface image to generate a grayscale image, and selecting an identification region for analysis in the grayscale image; and pixel of the identification region by a clustering algorithm Dividing into a water body pixel and a non-water body pixel; calculating a percentage of the water body pixel in the identification area, and determining whether the percentage is not less than a reference value, and if the judgment result is yes, it indicates that a mountain flood occurs.

Accordingly, the mountain flood detection method of the present invention can detect whether a mountain flood occurs by using an image processing technology, and is easily damaged by the earth and rock flow than the conventional contact type, thereby reducing the installation cost.

The clustering algorithm sets the grayscale values of any two pixels of the identification region as the initial group center of the two groups, and calculates the grayscale values of the remaining pixels in the identified region. The central gap of each of the initial group centers is assigned to the group with the smallest center gap to complete the grouping, and an average value of the grayscale values of the respective pixels of the two groups is calculated, so that each of them generates a better group center, and judges Whether the center difference between each of the preferred group centers and the respective initial group centers is less than a set value, and if the determination result is yes, each of the optimal group centers is generated, and the optimal group centers of the two groups are averaged. An optimal group average is generated, and the average of the N surface images is taken as a threshold value. In this way, compared with the conventional contact type, it is easy to be destroyed by the earth and stone flow, the installation cost can be reduced, and the efficiency of the overall disaster prevention warning can be improved by increasing the data acquisition rate.

Wherein, if the center difference between each of the preferred group centers and the center of the respective initial group centers is not less than a set value, each of the preferred group centers is used as the initial group center of each group and regrouped, if the number of regrouping If the value exceeds a range, each of the preferred group centers is used as the center of each of the optimal groups. In this way, compared with the conventional contact type, it is easy to be destroyed by the earth and stone flow, the installation cost can be reduced, and the efficiency of the overall disaster prevention warning can be improved by increasing the data acquisition rate.

Wherein, the formula of the threshold T can be as follows: Where T is the threshold value, A _i and B _i are the best group centers of the two groups of the i-th water surface image, respectively, and C _i is the optimal group average value of the i-th water surface image, i is the The time series of the N water surface images captured by the image capturing device in the same image area, and i=1, 2, 3, . . . , N. In this way, compared with the conventional contact type, it is easy to be destroyed by the earth and stone flow, the installation cost can be reduced, and the efficiency of the overall disaster prevention warning can be improved by increasing the data acquisition rate.

Wherein, the reference value is as follows: Where E=1 represents mountain flooding, E=0 represents mountain flooding, P is the percentage of water pixels in the identified area, and P ₀ is the water pixel in the identified area during non-rainfall. Percentage. In this way, the river flow before the rain can be accurately determined to improve the overall disaster prevention warning effect.

Wherein, if the percentage is less than a reference value, it means that the mountain flood has not occurred. In this way, compared with the conventional contact type, it is easy to be destroyed by the earth and stone flow, the installation cost can be reduced, and the efficiency of the overall disaster prevention warning can be improved by increasing the data acquisition rate.

〔this invention〕

S1‧‧‧ capture steps

S2‧‧‧ grayscale steps

S3‧‧‧ grouping steps

S4‧‧‧ alignment steps

A‧‧‧Best Group Center

B‧‧‧Best Group Center

C‧‧‧Best clustering average

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

Fig. 2 is a diagram showing the relative frequency of gray scale values of a water surface image according to an embodiment of the present invention.

The above and other objects, features and advantages of the present invention will become more <RTIgt; (Pixels) is the smallest unit of image composition used to represent the resolution of the image. For example, if the resolution of the image is 1024×768, it means that the image has 1024×768 pixels. It will be understood by those of ordinary skill in the art to which the present invention pertains.

The "Color Level" as used throughout the present invention refers to the degree of gradation of the color component or brightness exhibited by the pixel, for example, the red (R), green (G), and blue (B) components of the color image. The range of the gradation ranges from 0 to 255; alternatively, the range of the gradation of the grayscale image (Luminance) may range from 0 to 255, which can be understood by those of ordinary skill in the art to which the present invention pertains.

"Relative Frequency" as used throughout the present invention refers to the ratio of the number of pixels of each grayscale value in the determination region to the total number of pixels in the determination region. Values are understood by those of ordinary skill in the art to which the present invention pertains.

Referring to FIG. 1 , a flowchart of a method for detecting a mountain flood detection method according to the present invention may include a step S1, a gray step S2, a group step S3, and an alignment step. S4.

The capturing step S1 can take an image capturing device (such as a camera) to shoot a water surface of a river in a mountainous area to generate a water surface image, which can be a color image or a black and white image. Limitation, in the present embodiment, the water surface image is a color image and is described as a follow-up. In addition, because the mountainous area is susceptible to weather conditions such as sunlight, cloud cover or fog, the image of the water surface captured by the image capture device is blurred, or the degree of brightness is different, which affects the subsequent image processing analysis. Therefore, by shooting the same image Several water surface images during the day or days of the day, and subsequent steps for each of the water surface images to improve the accuracy of the judgment. The number of the water surface images is preferably 100 or more, but is not limited thereto.

In addition, by adjusting the photographing rate of the image capturing device, it is known that the water surface image is changed to one water surface image every minute every 10 minutes, so that the water surface image capturing rate can be improved, and the overall height is improved. The effectiveness of disaster prevention warnings.

The grayscale step S2 is based on the color gradation of the red, green and blue components of each pixel of the water image, and the average of the tone of the water image is converted to a brightness range of 0 to 255, so that a grayscale image is generated. And selecting a Region of Interest (ROI) in the grayscale image to perform a subsequent image processing step, and the image size of the identified region is not greater than the image size of the grayscale image. For example, the image may be selected. A trapezoidal range across the channel serves as an identification area. In this way, the time required for subsequent image processing can be reduced, and the system operation complexity can be reduced. However, when the color tone range of the original color tone of the water surface image is the brightness of 0 to 255, the gray leveling step S2 may be omitted, which can be understood by those having ordinary knowledge in the technical field of the present invention. Narration.

The grouping step S3 can analyze the identification area using a K-means algorithm (K-means). Since the water surface is brighter when the water is flowing, the gray scale value is relatively high, and the gray scale value of the bare soil and the gravel after being wetted by rain or river water is relatively low, so the Each pixel in the identification area is divided into a water body pixel and a non-water body pixel (such as bare soil or gravel) for subsequent comparison analysis.

In detail, please refer to FIG. 2 together, the grouping step S3 can divide the identification area into two groups, and set the gray level value of any two pixels of the identification area as an initial of each of the two groups. Cluster centers calculate the difference between the grayscale values of the remaining pixels in the identified area and the centers of the initial clusters, and assign them to the group with the smallest gap to complete the grouping. Then, calculating an average value of the grayscale values of the respective pixels of the two groups to generate a preferred group center, and determining whether the center difference between each of the preferred group centers and the respective initial group centers is less than a set value If the result of the determination is yes, then an optimal group center A, B is generated; if the judgment result is no, each of the preferred group centers is used as the initial group center of each group and regrouped, if regrouped If the number of times exceeds a range of values, each of the preferred group centers is used as each of the optimal group centers A, B. The set value can be set to 0.01, and the range value can be set to 100 times, but not limited thereto. Subsequently, an average of the best group centers A, B of the two groups is taken to produce an optimal group average C.

Performing the grayscale step S2 and the grouping step S3 for each of the water surface images, respectively generating the optimal group average value C for each of the water surface images, and averaging the N water surface images with each of the optimal group average values C The value is used as a threshold T. The formula for the threshold T can be as follows: Where T is the threshold value, A _i and B _i are respectively the best group centers of the two groups of the i-th water surface image, C _i is the optimal group average value of the i-th water surface image, and i is the image The timing numbers of the N water surface images captured by the device in the same image area, and i=1, 2, 3, . . . , N.

Determining whether the grayscale value of each pixel in the identification area is not less than the threshold value, and if the determination result is yes, setting the pixel as the water body pixel; if the determination result is no, setting the pixel as the non-aqueous pixel .

The comparison step S4 calculates a percentage of the water body pixels in the identification area, and determines whether the percentage is not less than a reference value. If the judgment result is yes, it indicates that a mountain flood occurs; if the judgment result is no, the mountain area is indicated. The flood did not occur and the formula can be as shown in the following formula (2): Among them, E=1 represents mountain flooding, E=0 represents mountain flooding, P is the percentage of water pixels in the identified area, and P ₀ is the water pixel in the identification area during non-rainfall. percentage. In addition, the reference value may be twice the percentage of the water body pixels in the identification area generated during the non-rainfall period, and the “non-rainfall period” refers to the non-rainfall period before the rain event occurs, preferably The day before the rain event occurred, so that the river flow before rainfall can be accurately determined to improve the overall disaster prevention warning.

In summary, the mountain flood detection method of the present invention can detect whether a mountain flood occurs by using an image processing technology, and is easily damaged by the earth and stone flow compared with the conventional contact type, thereby reducing the installation cost.

While the invention has been described in connection with the preferred embodiments described above, it is not intended to limit the scope of the invention. The technical scope of the invention is protected, and therefore the scope of the invention is defined by the scope of the appended claims.

Claims (6)

A mountain flood detection method includes: taking an image capturing device to shoot a water surface of a river in a mountainous area to generate a water surface image, the image capturing device is a closed circuit television installed in the mountain area, and the photographing rate is One water surface image per minute; grayscale processing on the water surface image to generate a grayscale image, and selecting an identification region for analysis in the grayscale image; and pixel of the identification region by a clustering algorithm Dividing into a water body pixel and a non-water body pixel; calculating a percentage of the water body pixel in the identification area, and determining whether the percentage is not less than a reference value, and if the judgment result is yes, it indicates that a mountain flood occurs. The method for detecting mountain floods according to claim 1, wherein the clustering algorithm sets a grayscale value of any two pixels of the identification region as an initial group center of the two groups, and calculates the identification region. The grayscale value of the remaining pixels is different from the center of each of the initial group centers, and is assigned to the group with the smallest center gap to complete the grouping, and an average value of the grayscale values of the respective pixels of the two groups is calculated to generate each a preferred group center, and determining whether the center difference between each of the preferred group centers and the respective initial group centers is less than a set value, and if the determination result is yes, each generating an optimal group center for the two groups The best group centers are averaged to produce an optimal group average, and the average of the N group images is used as the threshold for the best group average. The method for detecting mountain floods as described in claim 2, wherein if the center of each of the preferred group centers and the center of the respective initial group centers is not less than a set value, each of the preferred group centers is used as each The initial group center of the group is regrouped. If the number of regroupings exceeds a range of values, each of the preferred group centers is used as the best group center. For example, the mountain flood detection method described in claim 2, wherein the formula of the threshold value T can be expressed as follows: Where T is the threshold value, A _i and B _i are respectively the best group centers of the two groups of the i-th water surface image, C _i is the optimal group average value of the i-th water surface image, and i is the image The timing numbers of the N water surface images captured by the device in the same image area, and i=1, 2, 3, . . . , N. For example, the mountain flood detection method described in claim 1 is wherein the reference value is as follows: Among them, E=1 represents mountain flooding, E=0 represents mountain flooding, P is the percentage of water pixels in the identified area, and P ₀ is the water pixel in the identification area during non-rainfall. percentage. For example, in the mountain flood detection method described in claim 1, wherein if the percentage is less than a reference value, it indicates that the mountain flood has not occurred.