To improve the accuracy of ignition-point target location recognition and reduce recognition time, this paper proposes an intelligent fire-extinguishing robot target recognition method based on machine vision. First, a binocular stereo vision system is used to capture images of the ignition point. Second, background noise is reduced through image segmentation, while occlusion is processed using normalised colour-difference segmentation and linear projection methods. Finally, by combining Shannon's entropy mutual information theory with colour moment feature extraction, accurate recognition of the ignition-point target position is achieved through quantification of image information and evaluation of information-sharing degrees between image regions. Experimental results demonstrate that the proposed method maintains over 95% recognition accuracy, with the maximum recognition time not exceeding 3 seconds.