Abstract: Abstract In the field monitoring of cement pole diseases, many pole diseases characteristics are complex and difficult to distinguish accurately, which reduces the intelligent effect of cement pole disease monitoring. In order to improve the monitoring accuracy of cement pole diseases, a geometric analysis method of binocular stereo vision features of pole diseases was proposed. Using binocular stereo vision 3D imaging algorithm and according to the position relationship between imagers and the corresponding relationship of 3D coordinates between different imager scanning images, the site rod disease position coordinates were converted, the rod disease positioning error was corrected, the difference of state characteristics was analyzed, the geometric feature optimization model of rod disease was constructed, the geometric feature of rod disease was optimized step by step and the random gradient descent method was introduced to solve the problem of distinguishing erosion disease, corrosion disease, external force, stress extrusion disease, etc., and identify the disease type and severity. The experimental results showed that with the help of binocular vision 3D imaging technology, all the crack characteristics of the pole body were completely extracted. Considering the influence of parallax, the geometric characteristics of the pole body disease were optimized, the relative displacement measurement results had a high accuracy and the corrosion strength of the cement pole was consistent with that of the actual pole after optimization. The providing a new idea for on-site monitoring of cement pole diseases.