Converting contours in an image to lines in shapefile using Python

Question

I have processed a screenshot from Google Maps using OpenCV and identified some contours for the red road sections, and I want to project these contours to a shapefile. I have searched but couldn't find an answer that followed the same purpose. Here is the screenshot and the contours I found:

I also have the X, Y coordinates of each contour in pixels like the following:

[array([[[271, 485]],
 
        [[271, 488]],
 
        [[272, 489]],
 
        [[272, 491]],
 
        [[273, 492]],
 
        [[274, 492]],
 
        [[273, 492]],
 
        [[272, 491]],
 
        [[272, 489]],
 
        [[271, 488]]], dtype=int32)]

I have been trying to find a way to convert my X, Y coordinates to spatial longitude and latitude and then draw the contour with another library like fiona. I have seen map projections using pyproj geopandas gdal but couldn't apply any of them to my case. Can you point me in the right direction?

Answer 1

Try the below code and see if it helps:

# Import required modules
import cv2
import os
import shapefile
import numpy as np

# Import additional modules
import cv2 as cv
import pcv

# Read the image file
img = cv2.imread('/content/cropped.png')

# Check if image file exists and path is correct
if not os.path.exists(img):
    print("Image file not found or path is incorrect")
    exit()

# Create a new shapefile with lines as the geometry type
w = shapefile.Writer('output_shapefile.shp', shapeType=shapefile.POLYLINE)

# Create fields for the shapefile attributes
w.field('ID', 'N')

# Define RGB min and max values for red-amber color range
rgb_min_red_amber = np.array([0, 25, 25], np.uint8)
rgb_max_red_amber = np.array([30, 255, 255], np.uint8)

# Extract the contours from the image
rgb_min = rgb_min_red_amber.copy()
rgb_max = rgb_max_red_amber.copy()

# Convert image from BGR to HSV color space
hsv_img = cv.cvtColor(img, cv.COLOR_BGR2HSV)

# Extract pixels within specified color range
fiber = cv.inRange(hsv_img, rgb_min, rgb_max)

# Skeletonize the extracted pixels
fiber_skel = pcv.morphology.skeletonize(fiber)

# Extract contours from skeletonized image
contours, hierarchy = cv.findContours(fiber_skel, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)


# Check if any contours were extracted
if len(contours) == 0:
    print("No contours were extracted from the image")
    exit()

# Iterate over the contours and add them as lines to the shapefile
for i, cnt in enumerate(contours):
    # Remove single-dimensional entries from contour array
    cnt = cnt.reshape(1, -1, 2)

    # Add contour to shapefile as a line
    w.line(cnt.tolist())

    # Record ID of contour in shapefile
    w.record(i)

# Save the shapefile
w.close()

This code uses the cv2.findContours() method to extract the contours from the image. It then creates a new shapefile using the shapefile module and adds the contours as lines to the shapefile. Finally, it saves the shapefile.