Patents
Patents
Embedded Files
Automated Building Information Determinations From Floor Plan Structural Analysis
Automated Building Information Determinations From Floor Plan Structural Analysis
Lichen Wang, Naji Khosravan, Sai Sateesh Gudapati
Granted U.S. Invention Patent No. 2025/0166256A1
We developed an automated system that analyzes building floor plans and captured images using multimodal computer vision and machine learning to determine structural accessibility, room connectivity, and spatial visibility (isovist) attributes. The system enables property search and comparison, accessibility evaluation, and autonomous navigation, while providing data-driven improvement recommendations based on user behavior analysis.
[PDF]
Automated Building Identification Using Floor Plans and Acquired Building Images
Automated Building Identification Using Floor Plans and Acquired Building Images
Naji Khosravan, Lichen Wang, Sing Bing Kang
Granted U.S. Invention Patent No. 11830135B1
We developed a multimodal building matching system that analyzes both floor plans and panoramic images using computer vision and graph neural networks. The system extracts visual features from building images, constructs adjacency graphs to represent spatial relationships, and generates hierarchical vector embeddings for efficient similarity matching. It automatically identifies similar properties and provides explainable comparisons based on structural and visual characteristics.
[PDF] [Google Patent]
Automated Generation and Use of Building Information from Analysis of Floor Plans and Acquired Building Images
Automated Generation and Use of Building Information from Analysis of Floor Plans and Acquired Building Images
Eric M. Penner, Naji Khosravan, Sing Bing Kang, Lichen Wang, Zachary S. Bessinger
Granted U.S. Invention Patent No. 2024/0096097A1
We developed a multimodal AI agent system that automatically extracts building attributes by analyzing floor plans and images with trained neural networks, and leverages large language models to generate comprehensive textual descriptions. The AI agent autonomously validates existing building information, enables intelligent building search and matching, and enhances user experience through automated content generation for real estate applications.
[PDF] [Google Patent]
Reinforced Text Representation Learning
Reinforced Text Representation Learning
Bo Zong, Haifeng Chen, Lichen Wang
Granted U.S. Invention Patent No. 20210248425
We developed a reinforcement learning system that trains AI agents to navigate dependency trees and selectively extract semantic information for text understanding tasks. The system transforms MLP into graph learning by constructing dependency trees, then employs policy networks and temporal networks to intelligently traverse the graph and identify task-relevant features. This approach significantly improves language sentiment analysis and classification tasks by filtering out irrelevant information.
[PDF] [Google Patent] [Research Paper]
Unsupervised Graph Similarity Learning Based on Stochastic Subgraph Sampling
Unsupervised Graph Similarity Learning Based on Stochastic Subgraph Sampling
Bo Zong, Haifeng Chen, Lichen Wang
Granted U.S. Invention Patent No. 20210089652
We developed a self-supervised deep learning framework using stochastic subgraph sampling and auto-encoder representation learning for graph similarity analysis. The system transforms application syscall data into graphs, employs recurrent neural networks to generate vector embeddings, then leverages similarity metrics to identify anomalous behaviors. This approach enables malware detection, network intrusion detection, fraud detection, and complex system anomaly identification, without requiring manual feature engineering or labeled data.
[PDF] [Google Patent] [Research Paper]
Three-Dimensional (3D) Depth Imaging Systems and Methods for Dynamic Container Auto-Configuration
Three-Dimensional (3D) Depth Imaging Systems and Methods for Dynamic Container Auto-Configuration
Lichen Wang, Yan Zhang, Kevin O'Connell
Granted U.S. and International Invention Patent No.11010915
We developed an automated RGB-D imaging system that performs real-time 3D scene understanding and geometric reasoning during loading operations. The system uses structured light depth sensing to capture point clouds, applies semantic plane segmentation with statistical outlier filtering to handle occlusion, and employs spatially-adaptive 3D bounding box regression to accurately determine container boundaries and poses. This technology enables volumetric analysis, spatial configuration learning, and automated load optimization across diverse container geometries.
[PDF-US] [[PDF-CN]] [Google Patent]
Systems and methods for automatic camera installation guidance (CIG)
Systems and methods for automatic camera installation guidance (CIG)
Yan Zhang, Kevin O'Connell, Jay Williams, Lichen Wang
Granted U.S. and International Invention Patent No.11010915
We developed an automated RGB-D camera installation guidance system that performs real-time pose estimation and orientation assessment during deployment. The system employs 2.5D template matching with penalized least squares depth completion, segments container planes via RANSAC fitting, then evaluates exterior completeness and interior occlusion through 6-DOF pose estimation. It autonomously generates orientation refinement instructions via spatial reasoning, eliminating manual camera calibration and enabling rapid zero-touch deployment across diverse container loading environments.
[PDF-US] [PDF-CN] [Google Patent]
Novel Methods and System for Evaporator Frosting Inspection
Novel Methods and System for Evaporator Frosting Inspection
Lichen Wang, Min Wu, Qinglin Liu
Granted China Invention Patent No. CN201511025257.3
We developed an automated multi-view frost monitoring system for evaporator surfaces using temporal image analysis. The system combines grayscale analysis for frost coverage estimation, morphological segmentation for thickness measurement, and temporal motion analysis using optical flow and frame differencing to characterize frost growth dynamics. It autonomously determines optimal defrosting timing by evaluating frost coverage, thickness, growth rate, and propagation direction, enabling energy-efficient thermal management for refrigeration systems.
[PDF-CN]
Novel Methods and System of Foreign Matter Inspection in Infusion Bottle
Novel Methods and System of Foreign Matter Inspection in Infusion Bottle
Zhenshen Qu, Lichen Wang, Wenhua Jiao, et al.
Granted China Invention Patent No. CN2013102084539
We developed an integrated mechatronic-vision system for automated foreign particle detection in pharmaceutical infusion bottles with controlled rotation imaging. The system implements graduated acceleration with non-uniform velocity profiling, performs synchronized multi-view capture using adaptive dual-source illumination, applies temporal frame differencing and morphological blob segmentation, and employs multi-object tracking to discriminate particulate contaminants from bubbles and scattering artifacts. This technology enables high-throughput automated quality inspection for pharmaceutical manufacturing.
Page updated
Google Sites
Report abuse