With increasing necessities for reliable printed circuit board (PCB) products, there has been a considerable demand for a high speed and high precision vision positioning system. To locate a round pin chip with high accuracy and reliability with the obtained image, a positioning method is proposed based on the analysis of the image features, in which a deformable template is used to detect the deflection angle and the offset. The deformable template is constructed according to the arrangement of pins, whose offset, deflection, and zoom are denoted with five parameters. In addition, an energy function is defined by combining the image gradient, gray, and geometry features, which is optimized with the genetic algorithm to find the best matching position between the deformable template and a target image. The last experimental results show that this method has good accuracy, stability, and computing speed, and the detection errors are <0.1 deg and 0.25 pixels, which can meet the positioning accuracy of the placement machine vision system.