Machine vision based flotation froth mobility analysis

In a typical flotation cell circuit, the optimum cell performance depends largely on the ability of the operators to monitor and control the speed at which the froth crosses the lip of each cell. Hence, the froth flowrate is an essential tuning variable for flotation optimal monitoring and control system based on machine vision. However, the froth velocity is difficult to be measured accurately by traditional motion estimation algorithms in digital image processing, since the forth bubbles are stirred by the paddle of the flotation machine, which makes the bubbles lead to inevitable deformations such as rotation, scaling, bursting, coalescence and collapse. SIFT (Scaling Invariant Feature Transform) is a robust local descriptor, invariant to geometric distortion. Taking advantage of the SIFT, the froth velocity field reporting to the cell lip is measured precisely by the displacement estimation of the feature points with SIFT feature registration in the successive frames. The relation between the froth flowrate and flotation production performance indexes is discussed after the froth velocity field has been extracted, in order to accumulate the operation knowledge based on the judgment of froth surface visual information. The results executed in industrial case demonstrate that this method can extract the froth velocity field precisely and it is insensitive to bubble deformations. The method to froth velocity extraction and the operation knowledge accumulated provide guidance for the automatic control in flotation operation and lay a foundation for the establishment of a flotation optimal automatic control system based on machine vision.