Abstract: Thermal barrier coatings (TBCs) are widely used in blade protection of aero engines， but the harsh service environment inevitably leads to debonding defects inside TBC.The advancement of terahertz imaging technology provides higher quality guarantee for the service safety of TBC.Feature fusion imaging can improve the imaging effect， but it is difficult to characterize the internal relationship between different feature data by linear feature fusion method.Therefore， a quantitative evaluation method for debonding defects of thermal barrier coatings basd on terahertz nonlinear feature fusion is proposed. Firstly， the single characteristic parameter of the terahertz signal of the coating debonding defect is extracted for imaging， and high-quality feature parameters are selected by image quality evaluation index to compose a high-dimensional data set.Then， Laplacian eigenmaps is used to process high dimensional feature data.The intrinsic low-dimensional manifold features hidden between the terahertz features are fused and used for imaging the debonding defects of thermal barrier coatings.Finally， the region growth algorithm is used to segment the image of the feature fusion image to achieve the quantitative evaluation of the defect area.Compared with other methods， the defect imaging quality of the proposed method is better.The standard deviation， average gradient and energy gradient of the imaging results of the proposed method are increased by 57.8%， 37.7% and 33.2% compared with the single feature imaging results with the best index， and the accuracy rate of defect area identification is as high as 87.2%.