To address the issues of path redundancy and susceptibility to local optima in traditional path planning methods, this paper proposes an improved dung beetle optimization (IDBO) algorithm for efficient flight path planning of unmanned aerial vehicle (UAV) inspection of construction hoisting machinery. The core improvements include three mechanisms: A population initialization strategy based on a good point set to enhance spatial uniformity and coverage of initial solutions; an exponential decay formula to dynamically adjust the perturbation factor for adaptive balance between global exploration and local exploitation; and a hybrid Cauchy-Gaussian mutation mechanism to mutate stagnant populations, thereby inhibiting premature convergence and enhancing global search performance. Experimental results on benchmark test sets demonstrate that the proposed IDBO algorithm outperforms comparable algorithms in both convergence speed and solution accuracy, securing the top comprehensive ranking. For the UAV inspection application, a comprehensive evaluation model was formulated, integrating critical factors including path length, energy consumption, and threat cost. Simulations conducted within a realistic 3D construction site model populated with multiple hoisting machinery confirm that the paths planned by IDBO not only effectively avoid obstacles but also yield significant improvements in the objective function value. Specifically, in three scenarios of varying complexity, the performance improved by 11.47%, 7.23%, and 9.17%, respectively, when compared to the baseline method. Consequently, the proposed IDBO algorithm provides an effective and robust solution for autonomous UAV path planning in complex construction environments characterized by multiple obstacles, numerous inspection targets, and multi-dimensional costs, demonstrating considerable application potential.