Morel (Morchella spp.) target recognition and size measurement based on machine vision

Morel harvesting mostly rely on manual picking and grading, which are carried out separately. So the morel harvesting was with high labor intensity and much grading errors, which cause the decline of planting efficiency and economic benefits. Using morel harvesting machinery for morel picking and grading can help improve harvesting efficiency and benefits, which requires automatic morel target recognition and size measurement. A machine vision system was built and the morel target recognition and size measurement algorithm was proposed in this study. Firstly, based on the target recognition and measurement requirements of morel, the diameter and length of morel cap and the length of morel stem were determined as the key measurement indicators. And a monocular vision system was constructed and calibrated for morel image acquiring. Then, the morel stem was extracted firstly by liner enhancement and automatic global threshold binary processing, and the adhered morel stems were segmented with cyclic judgment algorithm. Subsequently, the morel cap region was predicted based on the position relationship between morel stem and cap, and morel cap recognition algorithm was designed based on local deformable template matching. Finally, the morel image was affine transformed to get the true size of pixels to calculate and correct these three key indicators based on the imaging principle of the built machine vision system. The experimental results show that the target recognition accuracy of designed morel target machine vision system is 90.73 %, the average error of the morel size measurement is 1.07 mm with average relative error value of 2.72 %. And the grading accuracy is 95 %, which is much higher than the grading accuracy with manual recognition method. That means the morel target recognition and size measurement machine vision system and algorithm designed in this study was with high accuracy, can fulfill the automatic morel picking and grading demands for morel harvesting machine design. The research results in this study will provide technical support for mushroom target recognition and size measurement, and also provide theoretical basis for the design of automatic equipment for morel picking and grading.