To achieve deformation detection of wind turbine blades, an optical detection method is proposed, where an RGBD camera mounted on an unmanned aerial vehicle (UAV) measures track deformations caused by jitter at a close range of 2–4 m, enabling the detection of abnormal deformations. Depth information is captured from the RGBD camera, from which the ROI foreground area is extracted. Optical flow techniques are utilized to counteract overall pixel displacement due to camera shake. The Segment Anything Model (SAM) is used to segment the target object, retrieving the mask of the target object from the first frame of the captured videos. Based on the color image and mask of the first frame, the Personalized Segment Anything Model (PerSAM) conducts a one-shot segmentation of the target object in subsequent video frames, enabling continuous tracking of the object’s edge contours. A GUI interface has been developed to facilitate segmentation and pixel-level tracking of any object within the ROI, and ultimately to calculate and measure the deformation of the track. In the experiments, under conditions of camera hover shake (with shake amplitude not exceeding 20 mm), the method has successfully measured deformations that are approximately 1% of the length of the target object, under complex backgrounds.