In this study, an unmanned aerial vehicle (UAV) intelligent detection system based on target recognition was designed and implemented under the background of low efficiency, high risk and poor environmental adaptability faced by traditional manual inspection. The system obtains high-precision 3D models and visual features of the target area through multi-sensor fusion technologies (such as lidar, visible light and infrared cameras), and combines optimized deep learning algorithms (convolutional neural network CNN) to achieve efficient target recognition in complex environments. For data transmission, multi-protocol wireless transmission and encryption technology are utilized to guarantee the reliability and security of data. At the level of UAV inspection execution, a multi-source positioning fusion scheme based on GPS and inertial navigation is proposed, combined with adaptive PID control algorithm to achieve high-precision cruise tracking, and the visual feature extraction ability in complex environments is enhanced by color histogram and edge detection. Experimental results demonstrate that the system significantly improves the inspection efficiency and target recognition accuracy, and reduces the labor cost and safety risk. Future research will focus on the optimization of extreme environment adaptability, the improvement of micro-target recognition algorithms, and the collaborative operation of multiple intelligent devices, so as to further expand application scenarios and system reliability.