Because of the linear CCD is used in non-contact dimensional measurement, especially at the edge of the detects object, the edge of the object can not be well coincided with the pixel edge of the sensor, there will be some errors. Currently, in order to reduce this errors, there are some edge subdivision algorithm based on dynamic probability subdivision method, interpolation method, fitting method and modem method and so on. Since the detection platform of this design is based on the color linear CCD, as well as taking into account the actual situation. On the basis of analysis of interpolation and fitting, an improved fitting is proposed, which is based on the most advantage points. At the same time, the design idea of this algorithm is introduced in detail, by measuring the same object, then using the above analysis of several algorithms, the experiment results show that the accuracy of the least-squares fitting algorithm based on the most advantage points is better. This result not only has important significance to improve the accuracy for CCD to detect object edge, but also provides a good basis for the application of the color linear CCD.