A non-invasive blood glucose detection system and method. The method comprises the steps of: controlling an infrared light source (13) so that same emits near-infrared light to irradiate a human body part to be detected, and acquiring, by means of a photoelectric sensor (14), a pulse wave signal from the human body part to be detected and under the irradiation of the near-infrared light; pre-processing the pulse wave signal by means of a signal processing circuit (21); extracting a characteristic value of the pulse wave signal; using a predictive neural network to detect the characteristic value of the pulse wave signal so as to obtain an initial detection value of a blood glucose concentration; using a classification neural network to detect the characteristic value of the pulse wave signal so as to obtain a detection interval of the blood glucose concentration; when the initial detection value is within the detection interval, displaying the initial detection value on a display screen (24) as a human body blood glucose concentration value of a person to be detected; and sending, by means of a communication unit (25), the human body blood glucose concentration value and a blood glucose measurement period of the person to be detected to a health management platform (02). The system and method can improve the accuracy of blood glucose concentration detection and enable a person to be detected to accurately understand his/her blood glucose condition.L'invention concerne un système et un procédé non invasifs de détection de glucose sanguin. Le procédé comprend les étapes consistant à : commander une source lumineuse infrarouge (13) de telle sorte qu'elle émet une lumière proche infrarouge pour irradier une partie du corps humain à détecter et acquérir, au moyen d'un capteur photoélectrique (14), un signal d'onde pulsatile provenant de la partie de corps humain à détecter et irradiée par la lumière proche infrarouge ; prétraiter le signal d'onde pulsatile au moyen d'un