This paper presents a three-dimensional chaotic map with five equilibrium points, which, unlike the classic Lorenz system, possesses a symmetric four-wing structure. Through detailed dynamical analysis and verification, the excellent chaotic properties of this map have been proven. Based on this system, an improved lightweight unmanned surface vehicle (USV) image information encryption system is designed. Given the constraints of USV equipment, the system first compresses images to enhance their storage capacity and transmission efficiency. Subsequently, an enhanced lightweight Feistel network encryption method is developed, utilizing a diagonal interpolation permutation model to ensure the security of the encryption process and reduce its complexity through a block structure. Simulation experiments and data analysis demonstrate that this algorithm effectively ensures the security of image information within the USV system. The use of compressed sensing and a block encryption structure reduces the complexity of the encryption process, thereby enhancing both the encryption efficiency and the transmission efficiency.