Detection Of Fertility Of Hatching Eggs Based On Computer Vision And Acoustic Impulse Technique

Automatically detect hatching eggs during early incubation would allow timely removal of infertile eggs and dead embryos from incubation thereby contributing to hatchery profits and application value. Computer vision and acoustic impulse response technique were used to study the fertility of hatching eggs, Bayesian discriminant models which proven to have high detection accuracy were built to distinguish infertile eggs and dead embryos.1. Computer vision system was built to acquire images of hatching eggs. RGB and HSI color models were selected to analysis the images, the image preprocessing method included image de-noise and background segmentation were established, and the average and standard deviation of R, G, B, H, S, I were computed. Total of12color characteristic parameters were used to describe the hatching eggs quality.2. The method of image acquisition by computer vision were tested, and the error rate was lower with the vertically acquisition way, thus adopting the vertically acquisition way was better than adopting the horizontal way. Some useful parameters were selected from the12color characteristics parameters to build Bayesian discriminant models which distinguished fertile from infertile eggs in early incubation days and could distinguish living and dead embryos in middle incubation days. The experimental results showed that, the correct classifications of early incubation detection were92.73%,99.92%,100%,100%respectively on days4,5,6and7for white shell eggs, while90.63%,91.75%,97.1%,97.1%,97.1%respectively on day5,6,7and8for brown shell eggs; the accuracy of the discriminant models reached100%for white shell eggs and98.89%for brown shell eggs during the middle incubation detection.3. The acoustic impulse response technique was used to monitor the fertility of hatching eggs during early incubation days. The standard deviation of acoustic frequency at the equator of the egg was smaller than standard deviation of acoustic frequency at the small end of the egg, thus the suitable place of acquisition of acoustic frequency was at the equator of the egg. According to the variation of the acoustic frequency during early incubation days, three character variables:f54, f64, f74were calculated to establish discriminant functions respectively. The results showed that, the function with the variable f64reached94.87%accuracy rate for white shell eggs, while the correct rate of the function with the variable f54reached94.87%for brown shell eggs.4. A comparative study of computer vision and acoustic impulse response technique to detect fertility of hatching eggs was discussed, the algorithm of computer vision was much complex, but in the applicability, accuracy, stability and convenience, computer vision performed better than the acoustic impulse response technique.