单曝光压缩成像开放世界目标重建方法、系统及应用
By using dual-text semantic guidance and physical motion trajectory perception detection, the problems of long global inverse problem solving and feature extraction failure in traditional single-exposure video compression imaging technology are solved, achieving efficient and accurate open-world target reconstruction.
CN122199839BActive Publication Date: 2026-07-17CHENGDU INSTITUTE OF BIOLOGY CHINESE ACADEMY OF SCIENCES
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU INSTITUTE OF BIOLOGY CHINESE ACADEMY OF SCIENCES
- Filing Date
- 2026-05-18
- Publication Date
- 2026-07-17
Smart Images
Figure CN122199839B_ABST
Abstract
本发明具体涉及单曝光压缩成像开放世界目标重建方法、系统及应用，解决了现有技术引发无效内存消耗与计算资源浪费的问题。通过预训练语言模型将文本提示解耦为静态外观与动态运动双先验，并提取降维后的二维混叠测量特征；利用动态先验引导可变形卷积，结合交叉注意力机制进行跨模态特征的交互与增强，对齐目标的时空拖影轨迹；通过掩码物理调制均衡和双文本联合语义相似度计算，在二维特征图上检测与定位目标；根据预测的边界框在测量域中裁剪出目标局部数据，利用轻量级重建网络进行靶向反向重建。极大地缩小了逆问题的求解维度空间，大幅降低了显存与计算冗余，在保证特定目标重建质量的前提下提升了整体视频重建效率。
Need to check novelty before this filing date? Find Prior Art