About me: research and engineering practice

What I work on, and how

I use images to read what is impractical to measure directly: the temperature and speed of airflow, and the body weight and lameness of dairy cows. I have also worked on funded crop imaging projects. I usually start by pinning down the question, then choose the sensing setup, build the model, check it against scale or velocity-sensor readings, and integrate it into a system.

Papers are in publications, funded work in research grants, and systems and tools in projects.

Technology worth building does not stop at a model or a demo. It carries through from sensing and data processing to deployment and day-to-day operation.

Education and certifications

Lab website: ISSP & MES LAB

Research topics

Reading airflow temperature from schlieren color

Filters laser-printed on low-cost projector film carry scratches that haze the image, so we dehaze it before a neural network predicts temperature from color. Published in IEEE Transactions on Computational Imaging.

Predicting airflow speed from schlieren video

A time-series network given one image region only followed the speed trend; with the region of interest split into six areas, it tracked a reference velocity sensor more closely. Published in IEEE Sensors Journal.

What shapes color in schlieren images

With a thermal imager as reference, causal decomposition examines how airflow and temperature affect schlieren color, identifying temperature change as a key factor. Best Student Paper Award (2nd place), Automation 2024.

Estimating dairy cow weight from depth images

Among back, hip and side depth features, back features with Gaussian process regression predicted and generalized best, beating farm estimates from body length and heart girth. Published in Biosystems Engineering.

Grading lameness from walking video

Lameness shows over a walk, not in one frame. I accumulate segmented frames into gait and back-height energy images, classify each, and fuse the outputs into one of four classes. A desktop app reads live camera streams.

Early warning for dairy cow health (in progress)

A Taiwan NSTC project I work on (since August 2026) aims to assess cow lameness, heat stress and calving risk from visible and thermal images, with retrieval-augmented generation and AI agents for decision support.

Imaging crop disease and waterlogging stress

Funded work on inspecting Pachira plants with visible and thermal imaging on a conveyor sorter, and on grading soybean waterlogging resilience, where adding image features improved prediction by about 10%.