Blade components are the core power elements of aero engines, and their operational performance and reliability are directly related to flight safety. Borescope inspection—also known as borescope examination—is one of the most frequent and technically demanding routine inspections for aircraft engines, requiring a high level of expertise and skill from maintenance engineers. However, current manual inspection methods are limited by factors such as technical knowledge, work experience, and subjective judgment, making it difficult to meet the accuracy and real-time requirements of blade damage detection. This paper proposes an intelligent blade counting and detection system based on an improved YOLOv11-DeepSORT deep collaborative architecture. By integrating Switchable Atrous Convolution (SAC) with an Online Convolutional Re-parameterization (OREPA) module, the system achieves multi-scale feature optimization and computational lightweighting. While maintaining model efficiency, the precision of blade feature detection is improved to 98.5%. In addition, to enhance the spatiotemporal continuity of object tracking, an occlusion-aware matching strategy (OS-NMS) is introduced, effectively mitigating the issue of dynamic target switching during detection. Validation using borescope inspection video data from multiple engines demonstrates that the absolute error rate of blade counting is below 0.05%. Compared with traditional manual inspection methods, detection efficiency is improved by a factor of 8.9, providing reliable technical support for the intelligent development of aircraft condition monitoring.