Abstract: [Objective]In this paper, the unmanned aerial vehicle (UAV) remote sensing technique was used to monitor mousehole density by great gerbil (Rhombomys opimus) in the southern margin of the Gurbantunggut desert forest. The purpose was to explore the best interpretation method of aerial imagery of the rat-holes and to provide a solution for the rapid interpretation of the data of the low altitude remote sensing aerial rodent damage.[Method]Fixed-wing UAV equipped with a digital cameras sensor was used to obtain 2.4 cm resolution aerial images, and TDOM imagery was acquired after pre-processing. The TDOM images were interpreted by artificial visual interpretation, supervised classification and object-oriented classification method, respectively, by which we extracted the spatial distribution information of the rodent cave, and then analyzed and evaluated the applicability and classification accuracy of each method in the rodent extraction in desert forest.[Result]By comparing the three methods of interpretation, it was found that the accuracy of artificial visual interpretation was high, but workload was too heavy for a large scale image; the classification accuracy of the supervised classification method was low and it was not applicable; and the object-oriented classification method not only had high precision, the classification results were consistent with visual interpretation method, but also improved the processing speed of big data.[Conclusion]This paper validated the availability of the UAV technique for dynamically monitoring rodent pests in desert forest, and laid a critical foundation for further study.

Key words: UAV, desert forest, Rhombomys opimus, object-oriented methods