To determine the optimal threshold in image automatic segmentation

a new thresholding method is proposed. In the method

the genetic algorithm carries on the global optimization to get the threshold rapidly

and the computational method of the crossover probability and mutation probability of the Adapted Genetic Algorithm(AGA) is improved. The improved AGA can guarantee that the multifamily of population and the astringency of the algorithm

and overcom the problems of a poor astringency and a premature occurrence in Simple Genetic Algorithm(SGA). Using the AGA to optimize the 2-D Fisher evaluation function that is taken as a threshold criteria

the best threshold is obtained. Then the image is segmented using this best threshold. Taking the TMS320VC5416 as the core processor

an embedded multi-objects measurement platform is built

and the multi-objects image fast segmentation and real-time measurement are realized in the platform. The experimental results show that the AGA has a good convergence rate and stability

also can not only ensure the accuracy of the image segmentation

but also shortened the time of image segmentation.