Based on LASSO regression,a kind of triple-channel infrared flame detector together with the corresponding recognition algorithm was put forward,and the hardware circuit and software programs were designed in this paper.Because of the data instability and environment diversity in the process of detection,the features extracted are diverse and complex.In this paper,resulted from the good predictive power,coefficient compression capability and feature selection ability of the LASSO regression,the accuracy and sensitivity of the triple-channel infrared flame detector was improved.LASSO regression also has the characteristics of good efficiency,high prediction accuracy and strong interpretation.At last,by comparing LASSO regression with traditional flame recognition algorithms,an experiment was provided to show the improvement in approximation accuracy,convergence speed and robustness.