研究目的
To improve the accuracy and speed of ROI detection in disaster victim detection systems by reducing false negative error rates through the addition of local entropy features to the GBVS model.
研究成果
The addition of local entropy features to the GBVS model effectively reduces false negative rates in ROI detection, enhancing the accuracy of disaster victim detection systems. Future work will involve testing the model on the GMVRT database to assess its performance with UAV images.
研究不足
The study is limited by the complexity of images taken during disaster events and the potential impact of size variation in UAV images on detection rates.
1：Experimental Design and Method Selection:
The study employs the GBVS model enhanced with local entropy features for ROI detection.
2：Sample Selection and Data Sources:
The model is tested on images from the GMVRT database, which contains images of humans taken by drones.
3：List of Experimental Equipment and Materials:
The study uses visual saliency maps and local entropy features for analysis.
4：Experimental Procedures and Operational Workflow:
The process involves feature extraction, activation map creation, and normalization/combination to form the final master map.
5：Data Analysis Methods:
Performance is measured using correlation coefficients and mutual information between human fixation density and master saliency map.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
