Abstract: In order to predict pig carcasses lean-percentage as soon as possible on the slaughtering line, finally the price policy that much content lean with a little fat carcass will be pay much money and little content lean with much fat will be pay a little money is coming true. This study was to assess carcass lean-percentage(y) calibration equation on crossbreed of ab extra pigs，and indexes which measured were chosen for pig carcases grading. Number of 114 samples about half yearly age was selected in this study. The usual predictors have been used：hot carcass weight，backfat of different parts and other characteristics measured on each carcass with ruler. By commercial cutting，multiple regression analysis of dissectible lean meat(%) on linear form based on the variables with different indexes was performed by SAS（SAS version 8.2）. The results suggested that the seventh equation: y=58.496-0.1046 x1-0.513 x7+0.273 x10 is the best multiple linear regression prediction model，in the model R2 is 0.8290 and RMSE is 2.07336. Backfat between sacrum & lumbar vertebra(x7) and M（M is the distance between the end of musculus gluteus medius and the edge of spinal cord canal ,x10）will be measured by ruler for pig carcasses grading in the future.