Yangming Cao

Project Overview:

• Developed an efficient and accurate blood oxygen detection algorithm using PPG technology to provide technical support for the medical health field

Responsibilities:

• Requirement Analysis and Algorithm Framework Construction
• Analyzed medical equipment requirements in depth and constructed an algorithm framework based on PPG principles
• Signal Processing and Algorithm Optimization
• Applied the Lambert-Beer law to optimize signal acquisition and processing processes; introduced the AMPD algorithm to improve the accuracy of signal peak detection
• Noise Reduction Technology Application
• Integrated Denoising RNN to effectively reduce interference from electromyographic signals and other noise
• Algorithm Stability and Accuracy Improvement
• Used RNN and LSTM technologies to enhance algorithm stability and robustness in variable environments

Achievements:

• Implemented a PPG-based blood oxygen detection algorithm, improving detection accuracy

Project Overview

• The development project of multispectral target recognition algorithms aims to enhance the accuracy and robustness of target detection through advanced image processing technology to adapt to diverse practical application scenarios

Responsibilities:

• Algorithm Design and Feature Engineering: Designed feature extraction processes based on representation learning and performed precise segmentation of multispectral images using the Felzenszwalb method
• Technical Research and Innovation: Conducted in-depth research on the latest research results in the field, quickly reproducing and improving algorithms to ensure the cutting-edge nature and applicability of technical solutions
• Algorithm Optimization and Performance Improvement:Identified and resolved technical bottlenecks of existing algorithms, applied hypothesis testing theory to improve target detection algorithms, reduced ill-conditioned matrix problems, and minimized reliance on prior knowledge
• Model Architecture and Adjustment: Adopted YOLOv5 as the backbone network, introduced channel attention mechanisms, and customized and adjusted the model architecture for spectral image data
• Stability and Generalization Verification: Implemented cross-validation strategies, using data partitioning and enhancement techniques to ensure the algorithm's stability and generalization in different scenarios
• Comprehensive Testing and Quality Assurance: Responsible for the overall testing process of the project, including testing in simulated real environments, to ensure the algorithm's performance in complex scenarios and meet strict acceptance standards

Achievements:

• Successfully developed and optimized the multispectral target recognition algorithm, significantly improving recognition accuracy and algorithm performance
• The introduction of channel attention mechanisms and cross-validation strategies enhanced the algorithm's stability and generalization
• The optimized algorithm improved recognition accuracy by 20% and response speed by 50% in complex scenarios
• The project was successfully implemented and passed the bidding process, gaining market recognition and commercial value, laying a technical foundation for subsequent target recognition projects and promoting the company's technical development and market competitiveness

Project Overview:

• Participated in the development of a dedicated multispectral mobile app, focusing on enhancing user experience and image processing efficiency

Responsibilities:

• Collaborated with cross-departmental teams to deeply analyze business requirements and design an intuitive and user-friendly UI interface
• Applied Material Design principles, utilizing Android-specific UI components such as Fragments, and custom elements like rounded corners and pop-ups
• Optimized auto-exposure and HDR algorithms to meet image processing needs under different lighting conditions
• Conducted native Android development using Java and Kotlin, improving the execution efficiency of C++ algorithms through JNI technology
• Implemented front-end and back-end data communication based on the HTTP protocol, ensuring real-time data synchronization and stability

Achievements:

• Successfully led and completed project development, and the app was launched and put into use, providing an excellent user experience
• The project became an important tool for the company to demonstrate technical capabilities, laying a solid foundation for subsequent product development

Project Overview:

• Constructed and maintained a big data platform for storing, processing, and analyzing spectral data during a two-month project, ensuring high efficiency, stability, and scalability

Responsibilities:

• Analyzed business requirements in depth and designed the MySQL database based on entity relationships, optimizing table structures for efficient storage and management of spectral data
• Developed backend architecture using the Spring Boot framework, achieving data processing and algorithm calls through modular design, ensuring system stability and scalability
• Implemented front-end and back-end communication using HTTP and FTP protocols, defining API interfaces and data transmission formats for efficient communication
• Enhanced performance by implementing the Felzenszwalb image algorithm in C/C++ and integrating it into the backend via JNI, significantly improving image processing performance
• Responsible for platform deployment and maintenance, configuring servers and networks to ensure stable platform operation and data security

Achievements:

• Achieved research and development goals on schedule, successfully built a spectral big data platform with cloud computing capabilities; provided a technical foundation and template for subsequent company development, accelerating the development process of new modules; significantly improved image processing speed through the locally integrated image algorithm via JNI