The invention discloses an infrared image monitoring method for a high-temperature pressure-bearing element based on gain tensor. The method comprises the following steps: (1) installing an infrared thermal imager near the high-temperature pressure-bearing element to be detected and acquiring an image; (2) storing the image and performing feature extraction to form a fault type image database; (3)establishing a fault monitoring model through a neural network learning method, and completing training, evaluation and storage; (4) monitoring the fault of the high-temperature pressure-bearing element; performing image interception on the acquired real-time infrared spectrum data of the element, predicting by using the trained fault monitoring model, and prompting that the element is normal orhas a problem according to a prediction result; and (5) when the prediction category of the collected infrared spectrum data is that there is a problem, calculating and processing the infrared spectrum data as follows. Temperature image data of an element can be monitored on line, timely response and disposal are achieved, and accident potential hazards are eliminated.