This paper presents an augmented reality platform for eye laser treatment with region guidance for intelligent surgery. The proposed method aims to assist doctors to plan the surgery, semi-automatically detect and protect sensitive regions within the retina and guide the laser during photocoagulation treatment for Diabetic Retinopathy. In the pre-surgical phase, we mosaic multiple retina fundus images to form a reference image. We then perform automatic optic disc and macula detection to mark out these sensitive regions. During the laser operation, we detect the aiming spot of the laser and track it on the reference image. Alerts are given out when the laser moves near the sensitive areas, and the laser function is disabled when the laser is within the sensitive regions. The proposed method can help doctors to perform laser surgery confidently without harming the patient.