The corona onset characteristic in high altitude areas is one of the important contents of the research on corona characteristics of bundle conductors. Ultraviolet imager is a non-contact discharge detection instrument, in this paper, study on corona characteristics of LGJ-720 bundle conductors in the UHV corona cage has been executed based on UV(ultraviolet) imaging. The experiment of corona characteristics on 6*LGJ-720 bundle conductors under the conditions of dry, wet and rain separately were carried out in the UHV test site in Xining City Ping’an. The UV images of corona discharge were obtained. In this paper UV image processing algorithm has been improved based on the previous one. The small spot is classified using clustering and graph cuts, the small spot near the main spot and with high gray value is considered to the effective spot while the ones away from the main spot to be interference. The optimized algorithm of filtering according to each spot position and gray value is established. The relationship between voltage and equivalent UV spot area under different meteorological conditions has obtained. Corona onset voltage has obtained through "tangent line method", which could provide a reference for UHV conductors’ selection only considering corona onset voltage.