The paper presents an experimental substantiation of the operating parameters of a device for separating a fruit flow during sorting using a machine vision system (MVS). A laboratory setup consisting of an MVS with a video camera, lens, and structural illumination, as well as a computing module, was created to conduct the research. The Box-Behnken design was chosen to obtain the regression equation. The main objective of the experiment was to obtain a regression equation that would allow one to evaluate the influence of the influencing factors on the accuracy of separating sorted apples. The following were chosen as optimization factors: main conveyor speed, carriage plate inclination angle, and number of solenoid valve holding steps. As a result of the experimental research, a regression equation was obtained to determine the accuracy of fruit separation. Analysis of the response surfaces made it possible to establish the optimal values of main conveyor speed = 0.41 m/s, carriage plate inclination angle = 27 degrees and the number of steps of holding the valve by the solenoid = 93 steps, at which the sorting accuracy is more than 95%.