This survey reviews human-centered trustworthy visual intelligence by summarizing its application landscape， key techniques， and emerging trends. As computer vision advances from perception to highly autonomous decision making and physical execution， risks related to privacy， fairness， robustness， transparency， and safety become increasingly salient. When system outputs may affect human safety and rights， performance optimization alone can no longer satisfy the requirements for trustworthiness. From a computer vision perspective， the paper traces the concept and evolution of trustworthy visual intelligence， emphasizing the multiple roles of humans as data subjects， cognitive participants， and ultimate controllers. A unified framework is then presented along three complementary spaces， information， cognitive， and physical， and a progressive paradigm is formulated that focuses on humans， serves humans， and remains under human control. The survey synthesizes human-oriented visual data analysis methods under fairness and privacy constraints， robust and responsible model design strategies， and human-machine collaborative control mechanisms centered on transparency and safety， with discussions across representative scenarios such as image enhancement， video analysis， robotic manipulation， and 3D visual perception. Finally， open challenges and future directions are outlined， including robustness evaluation， cross-scenario generalization， collaborative governance， and sustainable deployment， providing a roadmap for trustworthy visual intelligence in real-world systems.