Abstract: In order to seperate reeling?of male and female cocoon, a novel method for sexual discrimination of silkworm cocoons based on X-ray imaging technique was proposed. Images of silkworm chrysalis were obtained by X-ray imaging device, and then seven shape features were extracted after pre-processing of silkworm cocoons images. Discrimination models were established by linear discriminant analysis (LDA) and back propagation artificial neural network (BPANN), respectively. The results showed that the performance of LDA was better than that of BPANN between the same species in the same period, the same species at different period of time, and the different species of the same period; the accuracy the recognition of LDA models were larger than 91.33%, 86.73%, 89.53%, respectively; the accuracy of BPANN models were larger than 88.47%, 81.97%, 83.26%, respectively, while the recognition time of LDA model was (68.5 ms) much lower than BPANN model (503.6 ms). The results provide theoretical bases in the real-time testing of cocoon gender.