The performance of ultra-high frequency (UHF) antennas is crucial to the detection effect of UHF sensors. UHF composite antennas integrate different fractal structures, striving to retain the advantages of multiple structures. However, there are many variables to be optimized for composite antennas, and using traditional parameter scanning methods is time-consuming and laborious. In this paper, HFSS software and MATLAB software are used for joint simulation, thereby greatly improving the optimization efficiency of composite antennas. The composite antenna described in this paper combines the bow-tie and Sierpinski triangle structures, and draws on the design ideas of pixel antennas. The sub triangles in the Sierpinski triangle are treated as pixel like units, and the operations of filling or removing the copper layer are performed, enriching the antenna radiation layer design. This paper also proposes an improved bald eagle search algorithm, which combines SPM chaotic mapping, Brownian motion, and dynamic search step size. This method not only retains the good convergence of the original algorithm, but also partially improves the global search ability. After joint simulation and optimization, the UHF composite antenna is tested in the laboratory. The results show that the voltage standing wave ratio of the composite antenna is less than 3 in the frequency band range of 0.983 9 GHz to 3 GHz, close to 2 in the frequency band range of 1.643 8 GHz to 2.375 8 GHz, and less than 2 in most frequency bands of 1.127 9 GHz to 1.487 9 GHz and 2.375 8 GHz to 3 GHz, which meets the design expectations.