Abstract: Classification on cultivated land is an important basis for the management of cultivated land resources, the improvement of quality, and the guarantee of regional cultivated land food security. However, the existing classifying factor system of cultivated land still focuses on the influencing factors of productivity, and lacks the consideration of factors such as cultivated land health and ecological function. Taking Guangdong Province as the research area, on the basis of on-the-spot investigation and survey, this paper uses Delphi method and geographical detector synthetically to build a cultivated land classifying factor system for improving comprehensive efficiency. The cultivated land classifying factor system constructed in this paper consists of three basic layers, eight factor layers and 18 factors: natural factors, social economic factors and engineering factors. According to the system, the cultivated land classifying index in the study area is 481-4183, and its class is 3-15, with 7-11 as the main range. The correlation coefficient r between the classifying index and the yield of standard grain is 0.544, and the correlation coefficient r between the classifying index and the physical grading index and the utilization grading index are 0.738 and 0.616, respectively. Moreover, the quality of cultivated land can be improved by more than one class on the premise that the remediation engineering measures are in place. The empirical results show that the cultivated land classifying factor system constructed in this paper can better reflect the advantages and disadvantages of cultivated land productivity, and has a good connection with the original grading results. At the same time, the factors of environmental pollution and ecological land consolidation engineering are increased, which has a strong leading role in farmland protection and sustainable utilization projects in the new period.