有条件地将值分配给相邻的栅格像元吗？

我有一个值栅格：

m <- matrix(c(2,4,5,5,2,8,7,3,1,6,
         5,7,5,7,1,6,7,2,6,3,
         4,7,3,4,5,3,7,9,3,8,
         9,3,6,8,3,4,7,3,7,8,
         3,3,7,7,5,3,2,8,9,8,
         7,6,2,6,5,2,2,7,7,7,
         4,7,2,5,7,7,7,3,3,5,
         7,6,7,5,9,6,5,2,3,2,
         4,9,2,5,5,8,3,3,1,2,
         5,2,6,5,1,5,3,7,7,2),nrow=10, ncol=10, byrow = T)
r <- raster(m)
extent(r) <- matrix(c(0, 0, 10, 10), nrow=2)
plot(r)
text(r)

从该栅格中，如何根据此插图为当前像元的8个相邻像元分配值（或更改值）？我从此代码行在当前单元格中放置了一个红点：

points(xFromCol(r, col=5), yFromRow(r, row=5),col="red",pch=16)

在这里，预期结果将是：

其中当前像元的值（即值栅格中的5）被替换为0。

总体而言，必须按以下方式计算8个相邻像元的新值：

新值=红色矩形中包含的单元格值的平均值*当前单元格（红色点）与相邻单元格之间的距离（即，对角线相邻单元格的sqrt（2）或其他情况下的1）

更新资料

当相邻像元的边界超出栅格限制时，我需要计算尊重条件的相邻像元的新值。不遵守条件的相邻单元将等于“ NA”。

例如，如果使用[row，col]表示法的参考位置是c（1,1）而不是c（5,5），则只能计算右下角的新值。因此，预期结果将是：

     [,1] [,2] [,3]       
[1,] NA   NA   NA         
[2,] NA   0    NA         
[3,] NA   NA   New_value

例如，如果参考位置为c（3,1），则只能计算右上角，右上角和右下角的新值。因此，预期结果将是：

     [,1] [,2] [,3]       
[1,] NA   NA   New_value         
[2,] NA   0    New_value         
[3,] NA   NA   New_value

这是我通过使用函数进行的首次尝试，focal但是我很难编写自动代码。

选择相邻的单元格

mat_perc <- matrix(c(1,1,1,1,1,
                     1,1,1,1,1,
                     1,1,0,1,1,
                     1,1,1,1,1,
                     1,1,1,1,1), nrow=5, ncol=5, byrow = T)
cell_perc <- adjacent(r, cellFromRowCol(r, 5, 5), directions=mat_perc, pairs=FALSE, sorted=TRUE, include=TRUE)
r_perc <- rasterFromCells(r, cell_perc)
r_perc <- setValues(r_perc,extract(r, cell_perc))
plot(r_perc)
text(r_perc)

如果相邻的单元格位于当前单元格的左上角

focal_m <- matrix(c(1,1,NA,1,1,NA,NA,NA,NA), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)*sqrt(2)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的上中角

focal_m <- matrix(c(1,1,1,1,1,1,NA,NA,NA), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的左上角

focal_m <- matrix(c(NA,1,1,NA,1,1,NA,NA,NA), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)*sqrt(2)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的左上角

focal_m <- matrix(c(1,1,NA,1,1,NA,1,1,NA), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的右上角

focal_m <- matrix(c(NA,1,1,NA,1,1,NA,1,1), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的左下角

focal_m <- matrix(c(NA,NA,NA,1,1,NA,1,1,NA), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)*sqrt(2)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的下中角

focal_m <- matrix(c(NA,NA,NA,1,1,1,1,1,1), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

如果相邻的单元格位于当前单元格的右下角

focal_m <- matrix(c(NA,NA,NA,NA,1,1,NA,1,1), nrow=3, ncol=3, byrow = T)
focal_function <- function(x) mean(x,na.rm=T)*sqrt(2)
test <- as.matrix(focal(r_perc, focal_m, focal_function))

AssignValuesToAdjacentRasterCells下面的函数返回一个新的RasterLayer对象，该对象具有从原始栅格输入中分配的所需值。该功能检查从参考位置开始的相邻像元是否在光栅范围内。如果超出一定范围，它还会显示消息。如果您需要移动参考位置，则只需编写一个将输入位置更改为c（i，j）的迭代即可。

数据输入

# Load packages
library("raster")

# Load matrix data
m <- matrix(c(2,4,5,5,2,8,7,3,1,6,
              5,7,5,7,1,6,7,2,6,3,
              4,7,3,4,5,3,7,9,3,8,
              9,3,6,8,3,4,7,3,7,8,
              3,3,7,7,5,3,2,8,9,8,
              7,6,2,6,5,2,2,7,7,7,
              4,7,2,5,7,7,7,3,3,5,
              7,6,7,5,9,6,5,2,3,2,
              4,9,2,5,5,8,3,3,1,2,
              5,2,6,5,1,5,3,7,7,2), nrow=10, ncol=10, byrow = TRUE)

# Convert matrix to RasterLayer object
r <- raster(m)

# Assign extent to raster
extent(r) <- matrix(c(0, 0, 10, 10), nrow=2)

# Plot original raster
plot(r)
text(r)
points(xFromCol(r, col=5), yFromRow(r, row=5), col="red", pch=16)

功能

# Function to assigning values to the adjacent raster cells based on conditions
# Input raster: RasterLayer object
# Input position: two-dimension vector (e.g. c(5,5))

AssignValuesToAdjacentRasterCells <- function(raster, position) {

  # Reference position
  rowPosition = position[1]
  colPosition = position[2]

  # Adjacent cells positions
  adjacentBelow1 = rowPosition + 1
  adjacentBelow2 = rowPosition + 2
  adjacentUpper1 = rowPosition - 1
  adjacentUpper2 = rowPosition - 2
  adjacentLeft1 = colPosition - 1 
  adjacentLeft2 = colPosition - 2 
  adjacentRight1 = colPosition + 1
  adjacentRight2 = colPosition + 2

  # Check if adjacent cells positions are out of raster positions limits
  belowBound1 = adjacentBelow1 <= nrow(raster)
  belowBound2 = adjacentBelow2 <= nrow(raster)
  upperBound1 = adjacentUpper1 > 0
  upperBound2 = adjacentUpper2 > 0
  leftBound1 = adjacentLeft1 > 0 
  leftBound2 = adjacentLeft2 > 0 
  rightBound1 = adjacentRight1 <= ncol(raster)
  rightBound2 = adjacentRight2 <= ncol(raster) 

  if(upperBound2 & leftBound2) {

    val1 = mean(r[adjacentUpper2:adjacentUpper1, adjacentLeft2:adjacentLeft1]) * sqrt(2)

  } else {

    val1 = NA

  }

  if(upperBound2 & leftBound1 & rightBound1) {

    val2 = mean(r[adjacentUpper1:adjacentUpper2, adjacentLeft1:adjacentRight1])

  } else {

    val2 = NA

  }

  if(upperBound2 & rightBound2) {

    val3 = mean(r[adjacentUpper2:adjacentUpper1, adjacentRight1:adjacentRight2]) * sqrt(2)

  } else {

    val3 = NA

  }

  if(upperBound1 & belowBound1 & leftBound2) {

    val4 = mean(r[adjacentUpper1:adjacentBelow1, adjacentLeft2:adjacentLeft1])

  } else {

    val4 = NA

  }

  val5 = 0

  if(upperBound1 & belowBound1 & rightBound2) {

    val6 = mean(r[adjacentUpper1:adjacentBelow1, adjacentRight1:adjacentRight2])

  } else {

    val6 = NA

  }

  if(belowBound2 & leftBound2) {

    val7 = mean(r[adjacentBelow1:adjacentBelow2, adjacentLeft2:adjacentLeft1]) * sqrt(2)

  } else {

    val7 = NA

  }

  if(belowBound2 & leftBound1 & rightBound1) {

    val8 = mean(r[adjacentBelow1:adjacentBelow2, adjacentLeft1:adjacentRight1])

  } else {

    val8 = NA

  }

  if(belowBound2 & rightBound2) {

    val9 = mean(r[adjacentBelow1:adjacentBelow2, adjacentRight1:adjacentRight2]) * sqrt(2)

  } else {

    val9 = NA

  }

  # Build matrix
  mValues = matrix(data = c(val1, val2, val3,
                            val4, val5, val6,
                            val7, val8, val9), nrow = 3, ncol = 3, byrow = TRUE)    

  if(upperBound1) {

    a = adjacentUpper1

  } else {

    # Warning message
    cat(paste("\n Upper bound out of raster limits!"))
    a = rowPosition
    mValues <- mValues[-1,]

  }

  if(belowBound1) {

    b = adjacentBelow1

  } else {

    # Warning message
    cat(paste("\n Below bound out of raster limits!"))
    b = rowPosition
    mValues <- mValues[-3,]

  }

  if(leftBound1) {

    c = adjacentLeft1

  } else {

    # Warning message
    cat(paste("\n Left bound out of raster limits!"))
    c = colPosition
    mValues <- mValues[,-1]

  }

  if(rightBound1) {

    d = adjacentRight1

  } else {

    # Warning
    cat(paste("\n Right bound out of raster limits!"))
    d = colPosition
    mValues <- mValues[,-3]

  }

  # Convert matrix to RasterLayer object
  rValues = raster(mValues)

  # Assign values to raster
  raster[a:b, c:d] = rValues[,]  

  # Assign extent to raster
  extent(raster) <- matrix(c(0, 0, 10, 10), nrow = 2)

  # Return raster with assigned values
  return(raster)      

}

运行示例

# Run function AssignValuesToAdjacentRasterCells

# reference position (1,1)
example1 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(1,1))

# reference position (1,5)
example2 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(1,5))

# reference position (1,10)
example3 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(1,10))

# reference position (5,1)
example4 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(5,1))

# reference position (5,5)
example5 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(5,5))

# reference position (5,10)
example6 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(5,10))

# reference position (10,1)
example7 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(10,1))

# reference position (10,5)
example8 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(10,5))

# reference position (10,10)
example9 <- AssignValuesToAdjacentRasterCells(raster = r, position = c(10,10))

绘图示例

# Plot examples
par(mfrow=(c(3,3)))

plot(example1, main = "Position ref. (1,1)")
text(example1)
points(xFromCol(example1, col=1), yFromRow(example1, row=1), col="red", cex=2.5, lwd=2.5)

plot(example2, main = "Position ref. (1,5)")
text(example2)
points(xFromCol(example2, col=5), yFromRow(example2, row=1), col="red", cex=2.5, lwd=2.5)

plot(example3, main = "Position ref. (1,10)")
text(example3)
points(xFromCol(example3, col=10), yFromRow(example3, row=1), col="red", cex=2.5, lwd=2.5)

plot(example4, main = "Position ref. (5,1)")
text(example4)
points(xFromCol(example4, col=1), yFromRow(example4, row=5), col="red", cex=2.5, lwd=2.5)

plot(example5, main = "Position ref. (5,5)")
text(example5)
points(xFromCol(example5, col=5), yFromRow(example5, row=5), col="red", cex=2.5, lwd=2.5)

plot(example6, main = "Position ref. (5,10)")
text(example6)
points(xFromCol(example6, col=10), yFromRow(example6, row=5), col="red", cex=2.5, lwd=2.5)

plot(example7, main = "Position ref. (10,1)")
text(example7)
points(xFromCol(example7, col=1), yFromRow(example7, row=10), col="red", cex=2.5, lwd=2.5)

plot(example8, main = "Position ref. (10,5)")
text(example8)
points(xFromCol(example8, col=5), yFromRow(example8, row=10), col="red", cex=2.5, lwd=2.5)

plot(example9, main = "Position ref. (10,10)")
text(example9)
points(xFromCol(example9, col=10), yFromRow(example9, row=10), col="red", cex=2.5, lwd=2.5)

图例

注：白细胞意味着NA值

对于小矩阵上的矩阵算子，这是有意义的，并且易于处理。但是，将这样的函数应用于大型栅格时，您可能需要真正重新考虑逻辑。从概念上讲，这在一般应用中并不真正跟踪。您正在谈论的是传统上称为块统计的内容。但是，块统计本质上是从栅格的一个角开始，并用运算符替换指定窗口大小内的值块。通常，这种类型的运算符用于汇总数据。如果您考虑使用条件来计算矩阵的中心值，那将更容易处理。这样，您可以轻松使用聚焦功能。

请记住，栅格焦点函数正在读取数据块，这些数据块表示基于传递给w参数的矩阵在定义的邻域中的焦点值。结果是每个邻域的向量，并且焦点运算符的结果仅分配给焦点单元，而不分配给整个邻域。可以将其想象为获取围绕一个单元格值的矩阵，对其进行操作，为该单元格分配一个新值，然后移至下一个单元格。

如果确保na.rm = FALSE，则向量将始终代表精确的邻域（即，相同长度的向量），并被强制转换为可在函数内进行操作的矩阵对象。因此，您可以简单地编写一个函数，该函数采用期望向量，强制转换为矩阵，应用邻域表示法逻辑，然后将一个值分配为结果。然后可以将此函数传递给raster :: focal函数。

这是基于简单的强制和对焦点窗口的评估而在每个单元上发生的情况。“ w”对象本质上是与焦点传递w参数相同的矩阵定义。这就是在每次焦点评估中定义子集矢量大小的原因。

w=c(5,5)
x <- runif(w[1]*w[2])
x[25] <- NA
print(x)
( x <- matrix(x, nrow=w[1], ncol=w[2]) ) 
( se <- mean(x, na.rm=TRUE) * sqrt(2) )
ifelse( as.vector(x[(length(as.vector(x)) + 1)/2]) <= se, 1, 0) 

现在创建一个可以应用上述逻辑的功能。在这种情况下，您可以将se对象指定为值，或将其用作条件，例如“ ifelse”，以基于评估值分配值。我添加ifelse语句来说明如何评估邻域的多个条件并应用矩阵位置（邻域表示法）条件。在这个伪函数中，将x强制转换为矩阵是完全不必要的，仅用于说明如何实现。可以将邻域符号条件直接应用于矢量，而无需矩阵强制，因为矢量中的位置将应用于其在焦点窗口中的位置并保持固定。

f.rast <- function(x, dims=c(5,5)) {
  x <- matrix(x, nrow=dims[1], ncol=dims[2]) 
  se <- mean(x, na.rm=TRUE) * sqrt(2)
  ifelse( as.vector(x[(length(as.vector(x)) + 1)/2]) <= se, 1, 0)   
}  

并将其应用于栅格

library(raster)
r <- raster(nrows=100, ncols=100)
  r[] <- runif( ncell(r) )
  plot(r)

( r.class <- focal(r, w = matrix(1, nrow=w[1], ncol=w[2]), fun=f.rast) )
plot(r.class)  

通过使用[row，col]表示法对栅格进行子集化，可以轻松更新栅格值。只需注意，行和列从栅格的左上角开始；r [1,1]是左上像素索引，r [2,1]是r [1,1]下一个。

# the function to update raster cell values
focal_raster_update <- function(r, row, col) {
  # copy the raster to hold the temporary values
  r_copy <- r
  r_copy[row,col] <- 0
  #upper left
  r_copy[row-1,col-1] <- mean(r[(row-2):(row-1),(col-2):(col-1)]) * sqrt(2)
  #upper mid
  r_copy[row-1,col] <- mean(r[(row-2):(row-1),(col-1):(col+1)])
  #upper right
  r_copy[row-1,col+1] <- mean(r[(row-2):(row-1),(col+1):(col+2)]) * sqrt(2)
  #left
  r_copy[row,col-1] <- mean(r[(row-1):(row+1),(col-2):(col-1)])
  #right
  r_copy[row,col+1] <- mean(r[(row-1):(row+1),(col+1):(col+2)])
  #bottom left
  r_copy[row+1,col-1] <- mean(r[(row+1):(row+2),(col-2):(col-1)]) * sqrt(2)
  #bottom mid
  r_copy[row+1,col] <- mean(r[(row+1):(row+2),(col-1):(col+1)])
  #bottom right
  r_copy[row+1,col+1] <- mean(r[(row+1):(row+2),(col+1):(col+2)]) * sqrt(2)
  return(r_copy)
}
col <- 5
row <- 5
r <- focal_raster_update(r,row,col)

dev.set(1)
plot(r)
text(r,digits=2)