A rainfall detection method based on video image

Abstract

The invention discloses a rainfall detection method based on video image, including such steps as collecting water surface gray image sequence with single camera at fixed time interval, dividing the water surface gray image sequence into training sample image and to-be-measured image, dividing training sample image into no rain, light rain, moderate rain and heavy rain according to rainfall, and marking; enhancing the image texture by adopting the method based on the phase consistency, and performing the gray difference between two frames of images at the adjacent time after the enhancement toobtain a residual image; carrying out fast Fourier transform on the residual image to obtain the residual amplitude spectrum, normalizing the residual amplitude spectrum, and extracting four features; training the SVM classifier according to the extracted residual amplitude spectrum features; using the generated SVM classifier to identify the rainfall type of the image to be measured. The invention has the advantages of small calculation amount, low cost and intuitive and easy verification.

Description

technical field [0001] The invention relates to a rain detection method, in particular to a rain detection method based on video images. Background technique [0002] As a weather phenomenon, rainfall is an important research object of global hydrothermal cycle and climate change. The abnormal temporal and spatial distribution of rainfall is an important factor that induces various natural disasters. At present, most domestic observations and inferences of rainfall phenomena use water metering cylinders, rain gauges, and rain gauges to obtain macroscopic measurement data such as the degree of precipitation, the total amount of precipitation, and the frequency of each actual precipitation. Traditional rainfall data and data are generally Obtained by rain gauges or weather radar, the most widely used is the mechanical tipping bucket rain gauge. These methods often need to be combined with manual measurement to distinguish precipitation intensity, which has a low degree of aut...

AI Technical Summary

Problems solved by technology

The prior art discloses a method for detecting rainfall using images, collecting linear array images, and realizing automatic observation of precipitation phenomena by acquiring image information of precipitation particles, but it needs to be analyzed and calculated in combination with the actual shape model of precipitation particles, and the amount of calculation is large in actual work , the system cost is higher

Images