Abstract: Testing equipment is still lacking on the fruits post-production, particularly in a major fruit consuming of China. The current online device cannot fully meet to collect the full surface image information of spherical fruits, leading fail to accurately calculate the defect area. This study aims to rapidly and accurately detect the surface defect area of spherical fruits. The full surface image synthesis and defect area correction were proposed using the ideal ball model. A series of online detection and grading device were designed for the external quality of spherical fruits. This device was also different from the traditional full-surface detection device. Specifically, four cameras were used to collect images at the same time to obtain the full surface images of spherical fruits. The collected images were used to synthesize and correct, in order to obtain more accurate values of the surface defect area. Taking the apple as an example, an online detection device was designed to explore the best excitation light source required for the apple image acquisition. The refraction effect was then evaluated to clarify the influence on the fruit cup material. Then, the ball model was established to accurately calculate the surface defect area of apples using the apple full surface image synthesis and the defect area correction. A series of experiments were carried out to verify after segmentation and synthesis of the apple surface image, indicating no missing rate in the overall images. A defect area correction was proposed to calculate the real area of apple defects at any position in the image. 120 samples were selected for verification, including 30 scratch samples, 30 bruise samples, 30 spot samples, and 30 surface corruption samples, respectively. The determination coefficient (R2) was 0.978 7 between the predicted and the real value of the scratch sample defect area, where the Root Mean Squared Error (RMSE) was 3.577 4 mm2, R2 was 0.975 8 in the deflection angle experiment , and the RMSE was 3.466 3 mm2. The R2 was 0.973 0 between the predicted and the real value for the defect area of the impact sample, where the RMSE was 3.981 9 mm2, the R2 was 0.974 2 in the deflection angle experiment, and the RMSE was 4.062 4 mm2. The R2 was 0.970 8 between the predicted and real value for the defect area of the speckled spot sample, where the RMSE was 3.836 6 mm2, and the R2 was 0.977 9 in the deflection angle experiment, the RMSE was 3.895 3 mm2. In the surface corruption sample defect area, the R2 was 0.981 2 between the predicted and real values, the RMSE was 3.178 1 mm2, whereas, the R2 was 0.974 8 in the deflection angle experiment, and the RMSE was 6.304 4 mm2. The detection speed of the device was 2 apples/s, the rating accuracy was 95%, indicating a higher detection and apple rating accuracy than before. The relatively stable running was realized for the detection and grading evaluation of external defects of apples. The finding can provide technical support for the external quality detection of spherical fruits.