Abstract: Deep sky surveys provide fundamental data for studies of formation and evolution of galaxies, galaxy clusters, large-scale structures of the universe, and supermassive black holes. The SExtractor software package is an open-source software system. It can build a catalogue of objects from a sky-survey image and output magnitudes, surface coordinates, sizes, shape descriptions, and other information of objects. The SExtractor has been widely used for detection and measurement of objects in deep sky surveys. For object detection and measurement, abilities to detect objects (for completeness) and to separate objects of superimposed profiles (for resolving confusion) often have a larger impact on results than measurement errors due to photon noise. This requires certain balance between the abilities to detect objects and to separate objects of superimposed profiles. When using the SExtractor, the parameter values should then vary according to different types of objects so as to optimize detection and measurement. To detect apparently small faint objects, low detection thresholds need to be adopted, but low detection thresholds can lead to overdeblending of high-redshift galaxies of complex inner structures, or confusion of outer features of bright galaxies/scattered-light features of bright stars to be independent objects. Particularly, overdeblending can cause severe contamination in statistics of high-redshift close pairs identified from deep sky surveys. Confusion from parts of bright galaxies and scattered-light features of stars can be avoided by flexible settings of SExtractor parameter values. Overdeblending of faint galaxies of complex structures remains a difficult problem though. In this paper, we systematically describe the principles and parameters of the SExtractor. By incorporating our own analysis we also discuss accuracies and other measurement problems in using the SExtractor. At last, we give some suggestions to those problems.