Abstract:A machine vision image acquisition system was constructed to obtain 186 eggs hatching the 4 d image under LED light source to identify the early embryonic sex of chicken eggs. The preprocesses including component extraction,debackgrounding and binarization of egg image were carried out,followed by the using adaptive histogram equalization,top-hat and bottom-hat to enhance the image. Threshold segmentation of iterative add “and” operation was used to highlight the blood line texture. The 11-dimensional feature parameters of the image were extracted with the method of difference box,gray level co-occurrence matrix,gray histogram and geometric. When a BP (back propagation neural network,BPNN) model of chicken egg embryo sex identification was built,genetic algorithm (genetic algorithm,GA) was used to optimize BP neural network initial weights and thresholds initial. The results showed that the comprehensive accuracy of the training set of the GA-BP model was 99.73%,and the comprehensive accuracy of prediction set was 82.80%.