Two-Dimensional Sketch Recognition with Design Intent Capture

The high misrecognition rate and poor intelligence are the central flaws of current 2D sketch recognition methods.This paper presents a design intent capture approach to identify 2D sketches more accurately by improving the Bayesian network.Through analyzing the sketching behavior,we built the relationship between the sketching regularity such as the velocity and pressure from drawing and design intent.Then the Bayesian network for recognition reasoning is constructed,and a conditional probability table(CPT) is assigned to each node according to the sketch geometric feature.By the design intent probability,the CPTs of constraint nodes are modified to improve recognition precision.Finally,the recognition process is proposed,and the weights are automatically adjusted by the user's feedback.Compared to the work without intent capture,this technology can figure out the design elements and constraints during sketching,decrease the misrecognition by 30%,and distinguish the drawing disturbance.