Pengcheng Ruan

2023.07—2024.02: The development and optimization of key AI model of mobile monitoring machine for drain outlet inspection.

• it solves the problem of how to locating drain outlets and identifying drainage behavior in sunny days.
• In Shanghai, Ningbo, Jiashan and other places, the actual detection rate was higher than 93%

2024.02—2024.07: Upgrade the software framework of river and lake camera monitoring station.

• Strong support for the rapid adaptation of new products.
• Improve the versatility and performance of the model.
• The visual flow measurement algorithm is upgraded to achieve accurate estimation of the discharge.

2024.07—2024.12:Marine electronic boat nameplate completed

• Completed the reproduction and delivery of the SDK of SM9 encryption algorithm (both C/Python versions).
• Marks the realization of China's most advanced encryption technology in ship-borne equipment communication.

2024.05—2024.09: UAV inspection of four typical river and lake problems

• Object detection of four kinds of riverbank problems.
• Realized and optimized the algorithm of GPS location and area measurement of detected problem.

2024.09—2024.11:Rapid recognition and classification of remote sensing images

• Fast iterative upgrade algorithm of only 2% landform samples.

2024.11—2025.01:Optimize ai model in spectral camera product.

• Optimize the prediction using LightGBM.
• Optimize data flow and performance.

2025.01 present: Research on unmanned ship visual (USV) navigation algorithm

• Expected to solve the urban canyon problem through visual navigation.
• Expected to combine semantic segmentation algorithm with obstacle avoidance algorithm to improve safety.

2025.01 present: Selection and investigation of low-cost visual monitoring scheme

Face Recognition

1. Liveness Detection Topic ( Moire/mobile phone, etc)

• Single recognition accuracy >99.9%, overall accuracy >99%.

2. Pruning and optimization of commercial face fast location algorithm

• The accuracy of the algorithm remained at 92%.

Stock Trading

1. Complete the stock factor mining framework based on genetic algorithm

2. Explore stock picking strategies using deep reinforcement learning (DQN,PPO).

3. Designed a stock selection model based on Informer.

• Live trading with 400K RMB.