Aiming at the problems of omission or duplication of surface image acquisition in the detection of spherical fruits and vegetables, resulting in low accuracy of area identification, a method was proposed to ensure the true area ratio value of spherical fruits and vegetables. Firstly, multiple sets of images were acquired during the rolling process of samples, and the narrow column area facing the camera was cropped for splicing and unfolding. Then the geometric relationship was established by images, and the theoretical length of each row in the unfolded image was calculated. The projected image of the sample was obtained by scaling size with the area ratio of each area conforming to the real value. The standard sphere was used to verify the accuracy. Taguchi OA was used to design the L16(34) orthogonal test. The speed (A), cutting width (B) and height of camera (C) were the main influencing factors. The experimental results showed that the optimal parameters were as follows: A was 0.40m/s, B was 9pixels,C was 255mm, and the projection accuracy was 95.47%. Furthermore, projection experiments were carried out on three spherical fruit and vegetable samples of apple, orange and tomato, and the influence of different sampling angles on projection accuracy was discussed. The detection accuracy of apple, orange and tomato was 84.0%, 92.2% and 87.6%, respectively, which verified the feasibility of the twodimensional projection algorithm of fruit and vegetable surface.