ConvLSTM 학습 기반의 깊이 지도 및 위치 예측 알고리즘 설계 및 성능분석

다국어 초록 (Multilingual Abstract)

In this paper, we propose the design and performance analysis of a depth mapping and location prediction algorithm optimized for road driving environments with complex geometric features. The proposed algorithm uses a convolutional recurrent cell (ConvLSTM), which combines Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), to predict the depth map and location of the next frame from consecutive image frames. By integrating an Autoencoder for encoding spatial information and an RNN for handling temporal continuity, the algorithm can utilize both spatial and temporal information from visual data for more accurate predictions. The performance of the algorithm was validated using the KITTI vision Benchmark dataset, and the results confirmed that our approach is more efficient than location estimation methods based on complex sensors. These findings indicate that the proposed algorithm can significantly improve performance in understanding the complex spatial structure of road driving environments and estimating location.

목차 (Table of Contents)

• Abstract
• I. 서론
• II. 순환 신경망 학습 기반 깊이 지도 및 위치 예측 알고리즘
• III. KITTI 벤치 마크 데이터를 이용한성능 분석
• IV. 결론 및 향후 연구방향
• 참고문헌