Abstract: The spatial adaptability and motion degrees of freedom of the boom of an aerial work vehicle directly determine its operational efficiency and safety. However, existing booms exhibit limitations in motion coupling and real-time state perception, which restrict their ability to perform specialized operations in narrow working spaces. In this study, a spatial three-degree-of-freedom boom structure is proposed. By coordinating a pair of left-right symmetric hydraulic cylinders with a lifting hydraulic cylinder, coupled motion of the fly boom in luffing, slewing, and telescopic directions is achieved, while machine vision is employed for online pose perception. Kinematic mobility analysis demonstrates that the slewing and luffing mechanisms each possess one independent degree of freedom, ensuring overall motion controllability from the viewpoint of mechanism theory. Based on static equilibrium and moment analysis, a thrust model for the driving hydraulic cylinders is established. Using Matlab, thrust variation surfaces under slewing and luffing working conditions are obtained, from which the optimal installation angle range of the slewing and luffing cylinders is identified as 30°,40°. A machine-vision-based position monitoring method is further developed, in which a two-dimensional-to-three-dimensional inversion approach using pixel offsets and image area ratios is adopted to reconstruct the three-dimensional pose of the fly boom and back-calculate the required hydraulic driving force. Finite element analysis is conducted to verify deformation of the boom under extreme slewing, extreme luffing, and combined extreme working conditions. The results show that the deformation at the lug plates is significantly smaller than that at the boom tip, satisfying the stiffness requirements for engineering applications. This work provides an integrated structural and perception design framework for multi-degree-of-freedom positioning and intelligent operation and maintenance of aerial work vehicle booms in confined spaces, and lays a theoretical and engineering validation foundation for future research on vision-based life prediction and adaptive control.