基于多边形坐标对3D振幅矩阵进行子集化的技术求助

Hey there! Let's sort out this amplitude matrix subsetting problem for your bioacoustics work. Here's a straightforward, step-by-step solution that should get you the result you need:

Step 1: Create a grid of all time-frequency points #

First, we need to map every point in your s$amp matrix to its corresponding time and frequency coordinates. We'll use expand.grid to generate all combinations of the time and frequency vectors from your spectro object:

# Generate all time-frequency coordinate pairs
grid_points <- expand.grid(time = s$time, freq = s$freq)

Step 2: Identify points inside the polygon #

Next, we'll check which of these grid points fall inside (or on the boundary of) your con.max polygon. The base R function grDevices::point.in.polygon does this perfectly—it returns 1 for points inside the polygon, 2 for points on the boundary, and 3 for points outside. We'll use this to create a logical vector marking valid points:

# Check each point's position relative to the polygon
inside_status <- point.in.polygon(
  x = grid_points$time,
  y = grid_points$freq,
  pol.x = con.max$x,
  pol.y = con.max$y
)

# Flag points that are inside or on the polygon boundary
is_inside <- inside_status %in% c(1, 2)

Step 3: Reshape the logical vector to match your amplitude matrix #

The is_inside vector is a 1D list, but we need it to match the 2D structure of s$amp. Since s$amp has rows corresponding to frequencies and columns corresponding to times (the default from seewave::spectro), we'll reshape the vector into a matrix with the same dimensions:

# Convert the logical vector to a matrix matching s$amp's dimensions
inside_matrix <- matrix(is_inside, nrow = length(s$freq), ncol = length(s$time))

Step 4: Subset the amplitude matrix #

Finally, we'll create your new amplitude matrix by keeping only the values inside the polygon and setting everything else to NA (which acts as "blank" for plotting):

# Create the subset matrix: keep inside values, set outside to NA
amp_subset <- s$amp
amp_subset[!inside_matrix] <- NA

Verify the result #

You can confirm this works by overlaying the subset matrix on your original spectrogram:

# Plot the original spectrogram
spectro(tico, grid=F, osc=F, scale=F)
# Add the subset amplitude data
image(x = s$time, y = s$freq, z = amp_subset, add = TRUE, col = seewave::spectro.colors(100))
# Redraw the polygon to check alignment
polygon(x = con.max$x, y = con.max$y, lwd = 2)

A quick check: If you're unsure about the dimensions of s$amp, run dim(s$amp) to confirm rows correspond to frequencies and columns to times. If your setup differs, adjust the nrow/ncol arguments in the matrix() call accordingly.

内容的提问来源于stack exchange，提问作者Pedro Rocha