Skip to content Skip to navigation
See our Campus Ready site for the most up to date information about instruction.Campus ReadyCOVID Help
Mechatronics, Embedded Systems and Automation

Congratulations to Haoyu Niu for winning a $10K grant from Bayer Crop Science

December 26, 2020

We are very pleased to announce that Haoyu Niu has won a $10K grant  from Bayer Crop Science. It is a global competition we feel very proud of this award! This is part of recent effort to explore the potential of IoLT (internet of living things)   Sample projects in ag soil can be seen here.

Title: Scio for nematodes detection

Prepared by Haoyu Niu, Ph.D. student.

Faculty mentor: YangQuan Chen (email (Phone 209-3862958)

Web submission contents.

What is your lab’s research hypothesis?

A low-cost pocket-sized, cutting-edge technology micro-spectrometer, Scio, serves as a novel proximate sensor on early detection of nematodes infestation levels with machine learning algorithms, such as Neural Networks, Support Vector Machines, Random Forest, and so on. The programmable Scio will be used to measure the reflectance of the near-infrared band of the walnut leaves. Nematodes infection levels will be analyzed based on the reflectance from walnut leavces.

What is the rationale for this hypothesis?

Our Mechatronics, Embedded Systems, and Automation (MESA) lab has assessed the performances of Scio in a classification model for nematodes infection levels, which has shown improved classification accuracy for early detection of nematodes infection levels. By using the Neural Networks model, it can classify nematodes levels with an accuracy of 72% so far. The results show that there might be a strong relationship between the near-infrared reflectance and nematodes infection levels.

How has the hypothesis been validated to date? (optional)

The Scio was used to measure near-infrared reflectance from the walnut leaves in 2018. There were 30 sampling trees from 6 different blocks. Each sampling tree was measured three times to reduce the likelihood of errors or anomalous results. Based on the root-lesion nematodes number, the walnut trees were classified into three levels of nematodes infestation. The collected nematodes infestation data were trained using scikit-learn algorithms. The PCA model had the highest accuracy of 72%.

What is your research plan?

Early detection of nematodes infestation is important in the walnut industry for management decision support. To further assess the role of Scio sensor and machine learning algorithms for early detection of nematodes levels, we will design more field experiments. This study will take 1 year to complete. In the future, more data will be collected for training the model at UC Kearny. With more data from the walnut trees, more training data will hypothetically improve the accuracy of the model.

Last updated 12/26/2020 by Professor Chen, YangQuan