It’s possible that the values in the Sobel filter need to be played with to obtain these more blended regions.The second method tested was the Canny filter..From looking at the results of the Canny filter to the left, it can immediately be seen that Canny filter picks up considerably more edges..The disadvantage is that considerably more noise is picked up as well..This is expected, considering that the Canny algorithm takes the union of pixels that pass a magnitude threshold function and adjacent pixels that pass a somewhat lower threshold function..The Canny algorithm is more complex than a simple Sobel filter..It also depends on a Gaussian smoothing to be performed beforehand..The use of a median filter may be prohibiting optimal results.The Laplacian of Gaussian filter returned the best resultsThe final filter that was tested was the Laplacian of Gaussian filter..This filter is perfect for edge detection when scale is a factor..No matter what filter we use, we see that most of the prominent roads are captured..The marking of the noise that comes from small crop shadows or a possible house adjacent to a road are filter dependent..These are small details that are on a different scale as compared to the roads..This characteristic of the farm images makes them ideal for a Laplacian of Gaussian filter.. More details