The invention discloses a carrot surface defect detecting method which comprises the following steps: S1, pretreating: obtaining an RGB image of a carrot surface, carrying out image format conversionand color space conversion on the RGB image to output a target image file; S2, detecting: carrying out contour detection, green head detection, black spot detection, crack detection, variegation detection, length detection, diameter detection, draw ratio detection, tortuosity detection and taper detection on the carrot according to the target image file; and S3, quantizing: judging whether the detecting result information is qualified, if yes, quitting the detecting program, otherwise, outputting the unqualified detecting item. According to the carrot surface defect detecting method, machine vision is applied to automatic grading and sorting of agricultural products, so that strict unification of the quality standard of agricultural products, large scale of the agricultural product, and automatic production and processing of the agricultural products are facilitated.