Pingge Jiang

Research and Work Experience

Drexel University, Data Fusion Lab, ECE Department

Research Assistant, PhD

• Worked as a researcher for Center for Sleep and Circadian Neurobiology, University of Pennsylvania. The research topic was mice sleeping signal analysis. I served for designing the mice tracking algorithm based on overnight monitoring videos to get perfect mice location and generate the sleeping signal. The data was used for M.D.s for further analysis.
• Passed Ph.D. candidacy exam of ECE Department of Drexel University. I was focusing on radiological image registration and segmentation algorithms. My research topic was state-of-art CT and MRI registration.
• Worked on developing high-performance image registration algorithm. A map-reduce framework (Hadoop) was innovatively introduced to radiological image registration for large CT and MRI images. B-spline surface refinement was used as the deformable registration method.
• Worked on 3D patient tumor tracking based on multiple 2d x-ray scans. A multi-atlas segmentation is used for tumor locating, then we do image registration and gantry projection to get motion vectors. The tracking results are used for automatic real-time gantry refinement.
• Worked on sparse B-spline image registration for radiological images. An Octree-based multi-grid B-spline algorithm is introduced by focusing on detecting high-variation areas of images to decouple discontinuous motions. This algorithm is performing an accurate, efficient and low-cost registration process.
• Worked on sparse multi-grid B-spline registration with fully convolutional neural network

Alibaba Group, LBS team, Business Platform

Intern - Senior Software Engineer in Machine Learning

• Worked on customer analysis and sales prediction based on Alibaba big data. Participated project: A systematic model for Hema Supermarket candidate location scoring.

Purdue University, TASI Lab, ECE Department

Research Assistant, MS

• Severed as the key researcher for the naturalistic driving data analysis project sponsored by Toyota, Inc. Being the researcher in task 3, our goal was using large-scale naturalistic diving data to realize pedestrian detection, pedestrian tracking and pedestrian behavior analysis.
• We hired 110 drivers collecting daily driving data for us. I developed most of the software and GUIs used for data collection and processing. More than 10 undergraduates were working on the data by using them.
• Developed the pedestrian detection algorithm for large-scale naturalistic driving data. The first stage was driving environmental categorization. Then we used feature based or constraints based algorithms for pedestrian appearance probability estimation for different categories.
• The second algorithm we developed for better pedestrian detection performance was using HOG as pedestrian descriptor. We created a huge database as training data and then used machine learning method (ELM & SVM) to fuse multiple scores to determine pedestrian locations.
• A robust pedestrian tracking algorithm was developed, which involved with pedestrian detection, pedestrian segmentation, feature matching, feature fusion and motion learning. The algorithm was approved to be robust that the performance for complex real-life scenarios is better than other existing algorithms
• Master’s thesis: A new comprehensive approach for pedestrian tracking and status analysis. Pedestrian walking speed, pedestrian-vehicle relationship, collision probability are analyzed based on developed pedestrian detection and tracking algorithms.

Patents

• Du, E.Y., Yang, K., Jiang, P., Sherony, R. and Takahashi, H., Toyota Motor Engineering & Manufacturing North America, Inc., Toyota Jidosha Kabushiki Kaisha, Indiana University Research and Technology Corporation, 2015. System and method of alerting a driver that visual perception of pedestrian may be difficult. U.S. Patent 9,070,023.

Qualification

• Experienced programmer in C/C++, Matlab, Java, Python, Linux, software implementation and Human-Computer GUI design.
• Extensive knowledge in math background, digital image/video processing, computer vision, 2D/3D medical image segmentation/registration, pattern recognition, biometrics, object detection and tracking, classifier, neural networks, algorithm design and machine learning.