CN-121999080-A - Multispectral image-based water nitrate treatment and evaluation method

CN121999080ACN 121999080 ACN121999080 ACN 121999080ACN-121999080-A

Abstract

The invention discloses a water nitrate treatment evaluation method based on multispectral images, which particularly relates to the field of image processing and multispectral image analysis, and comprises the steps of performing acquisition operation on multispectral images before and after treatment, extracting water surface textures from the multispectral images, and performing edge detection on the extracted water surface textures to obtain disturbance boundary points; recording the positions of the disturbance boundary points at different acquisition moments point by point to form a disturbance boundary point sequence; according to the invention, by constructing the fluid coordinate grid based on the water disturbance boundary points, the disturbance tracks and the flow direction field to replace the traditional image matching and image correction system of the fixed reference object, the reliable pixel corresponding relation can be established under the condition that the reference object is destroyed in the multi-temporal multi-spectral image before and after treatment, so that the problem that the nitrate treatment effect cannot be reliably evaluated through the multi-spectral image in the prior art is solved.

Inventors

ZHAO JINXIANG
LI SHIYANG
TANG LIANG
WU MINGHONG

Assignees

上海大学

Dates

Publication Date: 20260508
Application Date: 20251222

Claims (8)

1. A water nitrate treatment evaluation method based on multispectral images is characterized by comprising the following steps: S1, performing acquisition operation on multispectral images before and after treatment, extracting water surface textures from the multispectral images, and performing edge detection on the extracted water surface textures to obtain disturbance boundary points; S2, calculating position difference values of adjacent moments point by point for the disturbance boundary point sequence to form a disturbance displacement sequence; s3, calculating tangential directions of disturbance tracks section by section, and performing direction statistics on the tangential directions according to a space region to form a flow direction field; s4, respectively performing coordinate mapping on the multispectral images before and after treatment according to a fluid coordinate grid to form two groups of coordinate mapping point sets, performing point-by-point comparison on the disturbance track forms of the two groups of coordinate mapping point sets, calculating form difference values of each comparison point, and performing position rearrangement on the coordinate mapping point sets according to the form difference values to form corresponding position point sets; S5, calculating a spectrum difference curve pixel by pixel for the multispectral image adjusted by the corresponding position point set, performing region stability screening on the spectrum difference curve, screening out spectrum difference results of a non-stable region, and performing spatial aggregation on the screened spectrum difference results to form a nitrate change graph.
2. The method for treating and evaluating water nitrate based on multispectral images according to claim 1, wherein in S1, multispectral image data are obtained by performing synchronous acquisition operation on multispectral images before and after treatment, channel-by-channel reading is performed on the multispectral image data, each spectrum channel is sequentially locked according to a spectrum channel index, reflection intensity values of water surface areas are read pixel by pixel in each spectrum channel, the read water surface reflection intensity values are summarized to generate a channel reflection set of water surface reflection, and the channel reflection set is rearranged according to a channel sequence to form a channel reflection sequence; Obtaining pixel difference distribution by performing inter-channel difference calculation on the channel reflection sequence, performing neighborhood difference change calculation on the pixel difference distribution, performing point-by-point calculation on the difference change of adjacent pixels to obtain a difference change amount, and reconstructing the gradient amount into a water surface texture according to pixel indexes; Performing difference value calculation on neighborhood pixels of each pixel in the water surface texture, performing point-by-point comparison on each difference value and an adjacent difference value to obtain a difference value mutation position, constructing a mutation point set according to pixel indexes by the difference value mutation position, performing point-by-point expansion on adjacent mutation points in the mutation point set to form a continuous mutation chain, and calibrating pixels in the continuous mutation chain as disturbance boundary points; and reading the disturbance boundary points at point-by-point positions at different acquisition moments to obtain the space coordinates of the disturbance boundary points, performing time index binding on the space coordinates, writing corresponding indexes according to the acquisition moments to form a time sequence record, and sequentially sorting the time sequence record to form a disturbance boundary point sequence and outputting the disturbance boundary point sequence.
3. The method for treating and evaluating water nitrate based on multispectral image according to claim 2, wherein in S2, the method further comprises the steps of performing point-by-point reading on a disturbance boundary point sequence to obtain disturbance boundary point coordinates at adjacent moments, performing subtraction operation on transverse coordinates and longitudinal coordinates of adjacent disturbance boundary points to obtain transverse differences and longitudinal differences, and recording the obtained transverse differences and longitudinal differences according to a position index sequence to form a position difference record; Writing the position difference records according to the acquisition time execution sequence of the disturbance boundary point sequence, and keeping the transverse difference value and the longitudinal difference value corresponding to each moment consistent with the time sequence of the original disturbance boundary point sequence to form a position difference value set organized according to time; respectively writing a transverse difference value and a longitudinal difference value of a first acquisition time arranged according to the acquisition time sequence in the position difference value set into a transverse accumulated value and a longitudinal accumulated value of the acquisition time; And at the subsequent acquisition time, performing addition operation on the transverse difference value of the current acquisition time and the transverse accumulated value written in the last acquisition time in the accumulated value set formed according to the acquisition time sequence to obtain the transverse accumulated value of the current acquisition time, performing addition operation on the longitudinal difference value of the current acquisition time and the longitudinal accumulated value written in the last acquisition time in the accumulated value set formed according to the acquisition time sequence to obtain the longitudinal accumulated value of the current acquisition time, and writing the transverse accumulated value and the longitudinal accumulated value of the current acquisition time into the accumulated value set formed according to the acquisition time sequence.
4. The method for water nitrate management evaluation based on multispectral image according to claim 3, wherein in S2, in the process of performing time-by-time accumulation on the position difference value set, before writing the transverse accumulated value and the longitudinal accumulated value at the current acquisition time into the accumulated value set formed according to the acquisition time sequence, judging whether the difference between the transverse accumulated value at the current acquisition time and the transverse accumulated value written at the last acquisition time in the set and the difference between the longitudinal accumulated value at the current acquisition time and the longitudinal accumulated value written at the last acquisition time in the set exceed a continuous variation allowable range or not: if the horizontal accumulated value and the vertical accumulated value do not exceed the continuous variation allowable range, writing the horizontal accumulated value and the vertical accumulated value at the current acquisition time into the accumulated value set according to the acquisition time sequence; If any one of the two accumulation values exceeds the allowable range of continuous change, respectively setting the transverse accumulation value and the longitudinal accumulation value to be written in at the current acquisition time as the transverse accumulation value and the longitudinal accumulation value written in at the last acquisition time in the set, and writing the transverse accumulation value and the longitudinal accumulation value into the accumulation value set according to the acquisition time sequence; After the accumulated values at all the acquisition moments are written, arranging the transverse accumulated values and the longitudinal accumulated values which are arranged according to the acquisition time sequence in the accumulated value set to form a disturbance displacement sequence; and (3) performing point-by-point connection on the disturbance displacement sequence, connecting the transverse accumulated values and the longitudinal accumulated values of adjacent acquisition moments in the disturbance displacement sequence according to the acquisition time sequence to form a continuous point sequence, and judging whether position breakage occurs between adjacent connection segments: If no position fracture occurs, connecting the continuous point rows according to the acquisition time sequence to form a connecting result, and arranging the connecting result into a disturbance track according to the acquisition time sequence; If position fracture occurs, resetting a transverse accumulated value and a longitudinal accumulated value corresponding to the acquisition time when the position fracture occurs as the transverse accumulated value and the longitudinal accumulated value of the previous acquisition time arranged according to the acquisition time sequence, performing one-time correction on the disturbance displacement sequence based on the reset transverse accumulated value and the reset longitudinal accumulated value, performing point-by-point connection according to the corrected disturbance displacement sequence, and finishing according to the acquisition time sequence to generate a disturbance track.
5. The method for treating and evaluating water nitrate based on multispectral images of claim 4, wherein in S3, the method further comprises the steps of reading disturbance tracks section by section, calculating coordinate differences of space coordinates corresponding to adjacent acquisition moments in the disturbance tracks to obtain transverse coordinate differences and longitudinal coordinate differences, calculating the ratio of the transverse coordinate differences to the longitudinal coordinate differences to obtain direction ratios, and writing the direction ratios into a sequence of direction ratio according to the acquisition time sequence; Dividing the direction ratio of each section in the direction ratio sequence into a transverse direction component and a longitudinal direction component, carrying out proportion normalization on the transverse direction component and the longitudinal direction component to obtain a normalized direction component, and writing the normalized direction component into the tangential direction according to the acquisition time sequence; Executing direction consistency comparison on tangential directions in the same area, and judging whether the direction difference value between any tangential direction in the area and other tangential directions in the same area exceeds a preset direction deviation threshold value or not: If the direction difference values among all tangential directions in the region do not exceed a preset direction deviation threshold value, taking a first tangential direction arranged according to the acquisition time sequence in the region as a region direction statistical result; If a tangential direction with a direction difference value exceeding a preset direction deviation threshold exists, performing proportional correction on the tangential direction with the direction difference value exceeding the threshold, and taking the corrected tangential direction as a direction statistical result of the area; writing the direction statistics results of all the areas into the flow direction field according to the area sequence; Taking the direction statistical result of each region in the flow direction field as a direction connecting line in the direction constraint execution region to form a direction constraint line, and executing segment connection continuity judgment between adjacent direction constraint lines: If the direction difference value between the adjacent direction constraint lines exceeds the preset direction continuity threshold value, performing one-time direction correction on the direction statistical result of the area, re-performing the direction connection in the area according to the corrected direction statistical result, and then writing the direction constraint lines; and performing segment-by-segment connection on the direction constraint lines according to the image space sequence, and performing index arrangement on the connection lines between the direction constraint lines according to the space coordinates to form a fluid coordinate grid.
6. The method for treating and evaluating water nitrate based on multispectral image according to claim 5, wherein in S4, pixel-by-pixel reading is performed on multispectral images before and after treatment, grid indexes of pixels in a fluid coordinate grid are written, and binding is performed on the grid indexes and space coordinates of corresponding pixels to form coordinate mapping point sets before and after treatment; the method comprises the steps of performing point-by-point reading on mapping points with the same grid index in a coordinate mapping point set before and after treatment, performing difference value calculation on transverse coordinates and longitudinal coordinates between corresponding points to form a morphology difference value, and performing point-by-point comparison on the morphology difference value to judge whether the morphology difference value exceeds a preset difference value threshold value or not: If the preset difference threshold is not exceeded, the space coordinates of the mapping points are subjected to one-time coordinate correction, and the corrected space coordinates are written into the original indexes to form initial corresponding point records; and carrying out in-region coordinate continuity detection on the initial corresponding point record, carrying out difference value calculation on the space coordinates of the corresponding points in the same region, and judging whether the difference value exceeds a continuity threshold value: If the continuity threshold value is not exceeded, the corresponding points in the area are written into the corresponding point set of the area according to the grid sequence, and if the continuity threshold value is exceeded, the corresponding points with the difference value exceeding the continuity threshold value in the area are subjected to one-time position correction, and the corrected corresponding points are written into the corresponding point set of the area according to the grid sequence; And performing index sorting on all the region corresponding point sets according to the spatial sequence of the fluid coordinate grid to form corresponding position point sets.
7. The method of claim 6, wherein S5 further comprises reading the pre-treatment multispectral image and the post-treatment multispectral image adjusted by the corresponding position point sets pixel by pixel, writing the reflection intensity values of the pre-treatment pixels on each spectrum channel at the same corresponding position point into a reflection sequence of the pre-treatment channel according to the index sequence of the spectrum channel, and writing the reflection intensity values of the post-treatment pixels on each spectrum channel at the same corresponding position point into a reflection sequence of the post-treatment channel according to the index sequence of the spectrum channel; performing channel-by-channel difference value calculation on the channel reflection sequence before treatment and the channel reflection sequence after treatment, and recording the channel difference value according to the spectrum channel index to form an initial spectrum difference curve; Performing point-by-point reading in the initial spectrum difference curve, performing difference calculation on spectrum differences of adjacent pixels in the same area according to the pixel arrangement sequence to obtain a spectrum difference value, and judging whether the spectrum difference value exceeds a preset stability threshold value or not: If the spectrum difference of the adjacent pixels in the region does not exceed the preset stability threshold value, writing the spectrum difference of the adjacent pixels in the region into a region stability difference record according to the region sequence; if the spectrum difference exceeds the preset stability threshold, performing one-time difference correction on the spectrum difference of the pixel, and writing the corrected spectrum difference into a region stable difference record according to the region sequence; and sequentially sorting all the region stability difference records to form a stability spectrum difference set.
8. The method for water nitrate management and evaluation based on multispectral image of claim 7, wherein in S5, further comprising performing point-by-point connection of the spectrum differences in the same spatial neighborhood according to the spatial coordinate order by performing spatial reading on the stable spectrum difference set, performing difference value calculation on the spectrum differences between adjacent connection points, and judging whether the difference value between the adjacent connection points exceeds a preset aggregation threshold value: if the preset aggregation threshold value is not exceeded, writing adjacent connecting line points into the initial aggregation record according to the space coordinate sequence; if the spectrum difference of the connecting point exceeds the preset aggregation threshold value, performing one-time aggregation correction on the spectrum difference of the connecting point, and writing the corrected connecting point into the initial aggregation record according to the sequence of the space coordinates; And performing spatial sequence arrangement on all initial aggregation records to form an aggregation spectrum difference record, and performing index arrangement on the aggregation spectrum difference record according to the image spatial sequence to form a nitrate change chart.

Description

Multispectral image-based water nitrate treatment and evaluation method Technical Field The invention relates to the technical field of image processing and multispectral image analysis, in particular to a water nitrate treatment evaluation method based on multispectral images. Background In the current water nitrate treatment evaluation technology based on multispectral images, a mainstream method generally relies on shorelines, riverbed textures, fixed structures and other relatively stable space reference targets, basic image correction is completed through geometric registration and radiometric calibration, and then multispectral images in different time phases are unified to the same space reference frame by combining image matching algorithms such as feature point extraction, region correspondence, deformation model fitting and the like so as to support the comparative analysis of nitrate concentration distribution before and after treatment; However, in actual governance engineering, nitrate reduction is often accompanied by strong intervention measures such as substrate sludge dredging, ecological base laying, river channel shaping, slope protection strengthening and the like, and these operations systematically remodel the micro-topography and the sediment structure of the river bed, so that the original bottom texture is destroyed or replaced in a large area, and the distribution from the composition of substrate particles, the reflection roughness to the turbidity field can be changed radically; Therefore, the evaluation of the treatment effect of the multispectral image essentially depends on accurate image matching and image correction to ensure the spatial comparability of images at different times, but the nitrate treatment process can obviously change the river bed structure, so that the traditional registration and correction system based on bottom textures and fixed ground objects loses physical basis, and finally the multi-time-phase image cannot establish reliable corresponding relation, thereby seriously weakening the credibility of nitrate treatment evaluation, which is the technical problem to be solved currently. Disclosure of Invention In order to overcome the defects in the prior art, the embodiment of the invention provides a water nitrate treatment evaluation method based on multispectral images, which is used for constructing a fluid coordinate grid based on water disturbance boundary points, disturbance tracks and flow direction fields to replace the traditional image matching and image correction system of a fixed reference object, so that reliable pixel corresponding relations can be established for multispectral images before and after treatment under the condition that the reference object is destroyed, and the problem that the nitrate treatment effect proposed in the background art cannot be reliably evaluated through the multispectral images is solved. In order to achieve the purpose, the invention provides the following technical scheme that the water nitrate treatment evaluation method based on multispectral images comprises the following steps: S1, performing acquisition operation on multispectral images before and after treatment, extracting water surface textures from the multispectral images, and performing edge detection on the extracted water surface textures to obtain disturbance boundary points; S2, calculating position difference values of adjacent moments point by point for the disturbance boundary point sequence to form a disturbance displacement sequence; s3, calculating tangential directions of disturbance tracks section by section, and performing direction statistics on the tangential directions according to a space region to form a flow direction field; s4, respectively performing coordinate mapping on the multispectral images before and after treatment according to a fluid coordinate grid to form two groups of coordinate mapping point sets, performing point-by-point comparison on the disturbance track forms of the two groups of coordinate mapping point sets, calculating form difference values of each comparison point, and performing position rearrangement on the coordinate mapping point sets according to the form difference values to form corresponding position point sets; S5, calculating a spectrum difference curve pixel by pixel for the multispectral image adjusted by the corresponding position point set, performing region stability screening on the spectrum difference curve, screening out spectrum difference results of a non-stable region, and performing spatial aggregation on the screened spectrum difference results to form a nitrate change graph. In a preferred embodiment, in S1, the method further includes obtaining multispectral image data by performing synchronous acquisition operation on multispectral images before and after treatment, performing channel-by-channel reading on the multispectral image data, sequentially locking each spectrum chann