The invention discloses an automatic detection method for a texture distortion area of a UAV (unmanned aerial vehicle) image based on GPU parallel computing, and the method comprises the steps that aGPU end loads an original UAV image, image interior orientation elements, exterior orientation elements and a DEM (digital elevation model), and calculates the size and range of an orthoimage according to a central projection imaging equation after ortho-rectification; the GPU end builds a GPU thread grid at a GPU end, opens a corresponding graphic memory, and inputs data into the graphic memory through copying; the GPU end carries out the inverse computation of the pixel coordinates of an original image, corresponding to the corrected image, from the corrected image at the GPU end according to the thread grid; the GPU end carries out the judgment of the texture distortion pixels; the GPU end carries out the image corrosion and image expansion processing of a binary image of a detection result according to the thread grid; the GPU end copies the binary image result after the texture distortion detection to a memory from a GPU global memory, carries out the vectorization of the binary image, and obtains the vector data of the texture distortion region. The obvious effects of the invention are that the method achieves the automatic detection of the texture distortion region in a single orthoimage of the UAV, and greatly improves the detection efficiency of the texture distortion region in the single orthoimage of the UAV.