Introduction to Image Processing for Environmental Science

Introduction to Image Processing for Environmental Science

Introduction to Image Processing for Environmental Science

Default Expiration: Does not expire

Earners of this badge have completed a hands-on image processing module spanning classical computer vision and modern AI methods, applied to real environmental science images. Using Python in Google Colab with LLM-assisted coding, participants progressed from pixel-level image manipulation through threshold-based segmentation and morphological measurement, to YOLO deep learning object detection and segmentation, and finally to CNN and Random Forest image classification with data augmentation. Real-world datasets included oyster shell photographs from UMCES field collection and IFCB (Imaging FlowCytobot) phytoplankton images, grounding every technique in an applied environmental science context.

About this Credential

Learning Objectives:

  1. Explain how digital images are represented as numerical arrays and manipulate pixel values directly using NumPy to create grayscale and RGB images.
  2. Apply classical computer vision techniques using OpenCV — including channel separation, pixel intensity profiling, threshold-based masking, contour detection, bounding box fitting, and ellipse fitting — to segment objects and extract quantitative morphological measurements.
  3. Recognize the limitations of threshold-based segmentation on complex environmental images and articulate why AI-based methods are needed.
  4. Run YOLO (You Only Look Once) inference using pretrained models, interpret detection and segmentation outputs, and fine-tune a YOLO model on a custom annotated environmental science dataset.
  5. Prepare image datasets for classification: loading, resizing, class-balanced splitting, and applying data augmentation (rotation, blurring) to address class imbalance.
  6. Build, train, and evaluate a Convolutional Neural Network (CNN) classifier in TensorFlow/Keras, interpreting training loss curves, F1 scores, recall, and confusion matrices.
  7. Extract statistical image features and train a Random Forest classifier; compare CNN and Random Forest performance across balanced and imbalanced dataset conditions.
How to add this type of credential to LinkedIn:
  • Copy the specific credential link for your earned record from your My Credentials page.
  • In LinkedIn, choose Add profile section > Licenses & Certifications.
  • Use this badge title as the certification name and “University of Maryland Center for Environmental Science (UMCES)” as the issuing organization.
  • Paste your unique credential link into “Credential URL”.