The invention discloses an oil casing shaft casing damage image feature pattern recognition method, comprising the following steps: (1) inputting a casing damage image; (2) converting the casing damage image from the two-dimensional space to the three-dimensional space, and dividing data points in the three-dimensional space into data points with damage and corrosion and data points without damageand corrosion according to the data threshold value of data points in the three-dimensional space; and (3) judging the type of damage corrosion according to the set of data points with damage corrosion. In the invention, a two-dimensional image corresponding to the casing damage image is converted into an image in a three-dimensional virtual space, the image is formed into corresponding data setsin the three-dimensional space; and by constraint of the distance threshold between points and the numerical value threshold of points in the image, in combination with the limitation of the maximumperpendicular distance allowed by image search, the data sets are divided into sub-type data sets representing different corrosion pattern features, and different corrosion pattern features correspondto different corrosion shapes, so that the corrosion types of the sub-type data sets can be accurately evaluated and recognized.