Image analysis workflow

  • Images grouped into batches by cropping system and date to maximize uniformity.
  • When multiple images available, images selected for uniformity of exposure and color
  • Occasional images possessed atypical color - in these cases, color balance was adjusted on a copy of the image, and both the original and copy were analyzed. (After review of false color images, the adjusted copy was retained only if classification was improved.) Note: We have fewer instances of this since we began using a DSLR camera in 2009 .
  • Pixels sampled using a custom macro in ImageJ.
  • Thresholds determined graphically for each batch of images (using R ). We used several custom functions in R to facilitate threshold determination.
  • Images were batch processed using a shell script that called the C++ image analysis program and compiled the output.
  • False color images reviewed to assure correct classification. Thresholds adjusted and batch processing repeated if needed.

Image Analysis Workflow

Original Image

In this image, the soil has a reddish color not present in the other images in this group.

Color Adjusted Image

Here we have adjusted the color in the image to reduce the red color in the soil. We used GIMP ( but any good image editor should work.

Original Image - Evaluation

Leaves 13%; Residue 26%; Bare Soil 57%; Ambiguous 4%

Color Adjusted Image - Evaluation

Leaves 25%; Residue 25%; Bare Soil 47%; Ambiguous 2%