This paper introduces the structure and working principle of a tobacco online foreign body eliminating system

and discusses the key techniques in its design. The testing experiments of sample machine are designed

and the results are analyzed. The HIS color model is used and the chroma is taken as the most characteristic parameter. The threshold value is set to distinguish whether the pixel belongs to tobacco or foreign body. Based on these

some pixels are set to constitute a judgement unit

then the proportion of pixels is counted

which belongs to foreign body

then it is distinguished whether the unit belongs to tobacco or foreign body based on this proportion. Finally

the detected result is provided. High-performance components and devices are employed to make the system. Direct operation on EMS memory is achieved by using image developing library. All above make the system operate realtimely on product line. The results of experiment indicate the sample machine can operate realimely on simulation online

and the distinguishing rate and eliminating rate to foreign body are preferable.