The machine-harvested cotton was processed through multistage seed cotton cleaning and lint cleaning, and cotton fiber was damaged inevitably. With the balance of appearance quality and inherent quality, the research method and testing program for process optimization control of machine-harvested cotton processing technology were proposed. According to the latest cotton quality inspection standard, nine parameters optimization targets such as trash area, trash count, reflectance, yellowness, upper half mean length, length uniformity, short fiber index, micronaire and strength were determined, and global optimization goal for the maximum transaction price of lint processing products was established. Seven rotational speed variables of cleaning machines, including inclined seed cotton cleaners I and II, recovery seed cotton cleaner, upper cotton gin, stripper and stick cleaner, saw lint cleaners I and II were selected as optimized control variables, which had significant effect on cotton cleaning. Architecture model based on monitoring layer, control layer and equipment layer was adopted, and upgrading key equipment automation was completed. The data model between control targets and control variables was built by using central composite design of response surface methodology. Taking global optimization control goal as fitness evaluation function, genetic algorithm was proposed to calculate the multivariate data model solution. Seven rotational speeds were 495r/min, 484r/min, 727r/min, 472r/min, 1131r/min, 822r/min, 763r/min, respectively. The test results showed that the change rate of trash area for processed lint products was reduced by 7 percentage points, the change rate of upper half mean length was increased by 2 percentage points, and product quality was more stable. The suggested method guaranteed fiber quality effectively with reduction of impurity content.