Based on our research work in the last five years, this paper highlights our innovative optical sensing system that can identify and separate silkworm gender highly suitable for sericulture industry. The key idea relies on our proposed optical penetration concepts and once combined with simple image processing operations leads to high accuracy in identifying of silkworm gender. Inside the system, there are electronic and mechanical parts that assist in controlling the overall system operation, processing the optical signal, and separating the female from male silkworm pupae. With current system performance, we achieve a very highly accurate more than 95% in identifying gender of silkworm pupae with an average system operational speed of 30 silkworm pupae/minute. Three of our systems are already in operation at Thailand’s Queen Sirikit Sericulture Centers.
Based on our previous work on light penetration-based silkworm gender identification, we find that unwanted optical noises scattering from the surrounding area near the silkworm pupa and the transparent support are sometimes analyzed and misinterpreted leading to incorrect silkworm gender identification. To alleviate this issue, we place a small rectangular hole on a transparent support so that it not only helps the user precisely place the silkworm pupa but also functions as a region of interest (ROI) for blocking unwanted optical noises and for roughly locating the abdomen region in the image for ease of image processing. Apart from the external ROI, we also assign a smaller ROI inside the image in order to remove strong scattering light from all edges of the external ROI and at the same time speed up our image processing operations. With only the external ROI in function, our experiment shows a measured 86% total accuracy in identifying gender of 120 silkworm pupae with a measured average processing time of 38 ms. Combining the external ROI and the image ROI together revamps the total accuracy in identifying the silkworm gender to 95% with a measured faster 18 ms processing time.
We examine the effect of different wavelength spectra in the performance of our optical penetration-based silkworm
pupa sex identification system. With available low-cost light emitting diodes (LEDs), each emitting different wavelength
spectra at 468 nm, 565 nm, 639 nm, 940 nm, and broad white light, we find that the body of the silkworm pupa can block
blue and near infrared light while allowing green and red light pass through. In particular, the red light can clearly
highlight an important organ called “chitin gland” of the female, leading to high accuracy of silkworm gender
identification. In our experiment with 120 silkworm pupae, measured high average 92.8% and lower average 87.5%
accuracies in identifying silkworm gender are obtained under red and white light LEDs, respectively.