Other subject about saffron
Mostafa Dehbashi; Amir Rajaei; Hossein KardanMoghadam
Abstract
Saffron is one of the most valuable crops on the planet and the most expensive agricultural and medicinal product globally; this plant has a special place among the industrial and export products of Iran. One of the challenges in the production of this plant is timely harvest from the ground as well ...
Read More
Saffron is one of the most valuable crops on the planet and the most expensive agricultural and medicinal product globally; this plant has a special place among the industrial and export products of Iran. One of the challenges in the production of this plant is timely harvest from the ground as well as red stigmas (branches) separating from other parts of saffron because the flowers are harvested in a brief period. Further, harvesting and separating at a limited time is a critical point. This research has been tried with the help of image processing techniques to discuss and recognize saffron flowers and how to identify them on the ground. In the first step, saffron flowers are recognized by transforming colored spaces. Then, the histogram and the minimum threshold are used to segment and remove extra pixels. For this purpose, to identify flowers, RGB space is converted to YCbCr space, and the combination of HSI and YCbCr color space is used to distinguish other objects in the image; also, a histogram of Cb component for early identification of saffron flowers are used. Then, those pixels which are misidentified are removed by the threshold value. Next step, the saffron flower needle-shaped leaves that are placed on the flowers are restored by morphological operation of the proposed method, and the flowers that overlap are removed. Then, the type of saffron flower (bud, broken, open flower) is determined, and the ability of harvest or suitability of saffron flowers for harvesting has been determined. Finally, the flower center that can be harvested is recognized and used by the saffron harvesting robot. Average results with respect to the accuracy, recall, F-measure, correctness, and correlation coefficient were 99.79, 99.42, 99.60, 99.91, and 99.50 achieved by the proposed method, respectively.