Qi Haoran祁皓然

• Average Overall Score:均分: 88.05/100
• Core Courses:核心课程: Signals and Systems (93), Linear Algebra (92), Embedded System Design (A), Digital Image Processing (A), Modern Control Theory (88)信号与系统 (93), 线性代数 (92), 嵌入式系统 (A), 数字图像处理 (A), 现代控制理论 (88)

Published in University Chemistry (Peer-reviewed Journal, Hosted by Peking University) 成果: 发表于北京大学主办的《大学化学》期刊

• Authored a first-author paper (DOI: 10.12461/PKU.DXHX202404036) detailing a novel vision-based concentration detection device.作为第一作者撰写论文(DOI: 10.12461/PKU.DXHX202404036)，详细介绍了一种基于视觉的创新浓度检测装置。
• Engineered a Raspberry Pi + OpenCV vision system; built a regression between solution color (HSV) and concentration based on Lambert–Beer law, achieving a correlation coefficient R² > 0.99.开发了树莓派+OpenCV视觉系统；依据朗伯–比尔定律建立颜色(HSV)与浓度的回归关系，相关系数R² > 0.99。
• Devised a custom 3D-printed black-box environment and disabled camera auto-white-balance to eliminate light interference, ensuring data integrity and measurement repeatability.设计了定制的3D打印黑箱环境，并禁用相机自动白平衡以消除光线干扰，确保了数据完整性和测量可重复性。

Paper submitted to the Chinese Journal of Applied Ecology (A National Core Journal, CSCD Indexed, Under Review) 成果: 论文已投稿至《应用生态学报》(国家核心期刊, CSCD索引, 在投)

• Co-developed the YOLOv8-MAT-2H model by integrating MSBlock and ADown modules to enhance feature extraction for small, complex targets in ecological monitoring.合作开发了YOLOv8-MAT-2H模型，通过集成MSBlock和ADown模块增强了对生态监测中小而复杂目标的特征提取能力。
• Optimized the model to improve mAP@0.5:0.95 by 2.57% while simultaneously reducing model parameters by 20%, ideal for edge device deployment.优化模型使mAP@0.5:0.95提升2.57%，同时将模型参数减少20%，适合边缘设备部署。
• Spearheaded strategies to handle environmental occlusions in real-world river settings, contributing directly to the system's final operational accuracy of nearly 80% for the Beijing Water Authority.主导制定了应对真实河流环境中遮挡问题的策略，为系统最终达到接近80%的业务准确率(服务于北京市水务局)做出了直接贡献。

• Contributed to a project developing a fiber optic biosensor that improved the detection limit by 15 times compared to traditional ELISA technology.参与开发了一种光纤生物传感器，其检测限比传统ELISA技术提高了15倍。
• Executed data processing algorithm optimization and participated in sensor design and validation, helping to reduce detection time from hours to just 5 minutes.负责数据处理算法优化及传感器设计验证，帮助将检测时间从数小时缩短至仅5分钟。

Information & Networking Track,信息化与网络化赛道， 1st Prize (North China Region)华北赛区一等奖

• Led a team to design and deploy a complete industrial network system for a simulated factory environment.带领团队为模拟工厂环境设计并部署了一套完整的工业网络系统。
• Architected the network topology from abstract requirements, defining the IP schema, VLAN segmentation, and firewall access control policies.根据抽象需求构建网络拓扑，定义了IP规划、VLAN分段和防火墙访问控制策略。
• Engineered and commissioned the entire system, configuring Siemens S7 PLCs and industrial switches via TIA Portal to ensure robust, redundant communication.工程实现并调试了整个系统，通过TIA Portal配置西门子S7 PLC和工业交换机，确保了稳定、冗余的通信。

Quadruped Robot Collaborative Challenge,四足机器人协同挑战赛， 2nd Prize (National Final)全国总决赛二等奖

• Directed the team and developed main functional packages using ROS for a quadruped robot to perform autonomous navigation, object manipulation, and animal recognition.指导团队并使用ROS为主功能包进行开发，使四足机器人能够执行自主导航、物体操纵和动物识别。
• Engineered an innovative solution to overcome hardware limitations of the Raspberry Pi by implementing a "position-then-detect" strategy, capturing and processing a single frame to bypass real-time video latency.设计了一种创新方案，通过实施“先定位后检测”策略，捕捉和处理单帧图像，绕过实时视频延迟，从而克服了树莓派的硬件限制。
• Trained and deployed a YOLOv8 model that achieved 91.5% accuracy in identifying six types of animals under variable lighting conditions through targeted data augmentation.训练并部署了一个YOLOv8模型，通过有针对性的数据增强，在多变的光照条件下识别六种动物的准确率达到91.5%。

Tourism Prediction Model 旅游预测模型， National Second Prize全国二等奖

• Architected and fused Random Forest and XGBoost models on 20 years of meteorological data to predict regional weather phenomena.基于20年气象数据，融合Random Forest与XGBoost预测区域性气象现象。
• Modeled and validated a Growing Degree Days (GDD) model to forecast flowering periods for ecological and tourism applications.建立并验证GDD模型预测植物花期，为生态与旅游提供量化依据。

Monocular Vision Measurement Device 单目视觉测量装置， Second Prize (Beijing Region)二等奖（北京赛区）

• Engineered an STM32F407-based embedded vision system to measure object distance and size with a single camera, achieving distance error < ±2cm and size error < ±0.5cm.基于STM32F407实现嵌入式视觉系统，单目测距与测尺寸，距离误差< ±2cm、尺寸误差< ±0.5cm。
• Applied non-linear fitting for distance calibration and perspective transform to correct distortion, improving accuracy.采用非线性拟合进行距离标定并使用透视变换校正畸变，提升准确度。
• Integrated a lightweight YOLOv5s for robust detection in complex scenes, demonstrating embedded AI capability.集成轻量YOLOv5s实现复杂场景鲁棒检测，体现嵌入式AI能力。

Project: An intelligent monitoring system for industrial centrifuges, 项目：工业离心机智能监测系统， Provincial Semi-finalist省赛半决赛入围(北京市二等奖)

• Owned the functional development and data analysis for the monitoring subsystem within a complex project targeted at the nuclear sector.负责一个面向核工业领域的复杂项目中，监测子系统的功能开发和数据分析。
• Delivered critical code modules that ensured the stability and integrity of the multi-sensor data stream, foundational for the system's fault diagnosis and vibration suppression capabilities.交付了确保多传感器数据流稳定性和完整性的关键代码模块，为系统的故障诊断和振动抑制能力奠定了基础。

• Designed PCB layout for an Automatic Sewage and Scale Removal Controller using Altium Designer.使用 Altium Designer 设计“自动排污除垢控制器”控制板 PCB。
• Developed initial firmware in C for a 51 MCU to manage timed discharge and descaling based on sensor inputs.基于 51 单片机用 C 开发初版固件，按传感器输入执行定时排污与除垢。
• Contributed to the electronic control solution from schematic to preliminary code for a new energy-saving product.参与从原理图到初版代码的电子控制方案，服务于节能类新品。

• Analyzed a plastic bottle inspection line to identify critical failure points in the assembly and sensor calibration processes.分析了塑料瓶检测生产线，以确定装配和传感器校准过程中的关键故障点。
• Engineered a revised debugging workflow which was adopted by the team, reducing setup time by an estimated 15% and improving overall operational stability.设计了一套修订的调试工作流程并被团队采纳，将设置时间估计减少了15%，并提高了整体运营稳定性。