Abstract: The number of Bactrocera Dorsalis in occurring period is the important parameter which threats the growth situation for fruit trees and is the basis of implementing variable rate technology. In order to realize detecting the occurrence of Bactrocera Dorsalis real-time and fasting in large-scale orchard, machine vision technologies based on moving object trace tracking were employed to trace Bactrocera Dorsalis behavies around traps real-time, so as to achieve statistics of their number into the hole precisely. The fore 50 000 video image were selected as evaluation samples which collected in feed room of Resource and Environment College in South China Agricultural University using vision monitoring platform for Bactrocera Dorsalis. Through comparison of results with methods of artificial and machine vision detecting, the experiment indicated that the number of Bactrocera Dorsalis detected by artificial and machine vision were 85 heads and 78 heads, respectively. The loss rate of detecting using machine vision was 9.4%, which can meet the demands of pests’ monitoring.