Due to the complex process of copper flotation and the frequently diversified conditions of ore sources, it is difficult to identify rougher flotation conditions and maintain the stability of production process. By deeply analyzing the characteristics of the copper flotation process, the recognition system for working conditions in copper rougher is established and the cascaded recognition method is presented. At the first stage, the recognition model is built to identify feeding ore types based on fusion information of froth image local colour features and process parameters. At the second stage, the asymmetry binary tree SVM multi-class classification method with working condition priority rating (WCP-BTSVM) is used to recognize copper rougher flotation conditions. As demonstrated in the industrial experiment, the proposed method can relatively accurate identify the working conditions in copper rougher and thus can provide a solid foundation for decision-making in follow-up process control. (C) 2017 Elsevier Ltd. All rights reserved.