<p>Aircraft play a vital role in aviation safety and security, especially in improving the accuracy and efficiency of aircraft inspections. However, traditional inspection methods face challenges in terms of low efficiency. To address this, drones are increasingly used for inspection tasks. However, traditional drone image feature extraction lacks high precision. This study proposes an optimized approach for stage target detection algorithms using a 50-layer residual network, which results in a new image detection algorithm. This method is applied to drone-based aircraft inspection and integrated into a safety monitoring system. Experimental validation shows that the average precision and recall rates of the proposed method are 91.2% and 93.4%, respectively, with both run time and loss rate lower than those of the compared algorithms. In actual detection scenarios, the system achieves an average Intersection over Union of 0.92 and successfully detects all target defects. The system also reports an average precision rate of 98.4% and an obstacle avoidance rate of 98.1%. The results demonstrate that incorporating ResNet-50 to enhance conventional object detection algorithms improves both the detection accuracy and operational efficiency for identifying minor defects on aircraft, thereby providing more precise and efficient technical support for unmanned aerial vehicle inspections.</p>