Detection Method of Coconut Development Intelligence Based on Improved YOLO V5 Model

Because of the special structure of coconut, it is difficult to observe the development of its internal structure accurately, so the research on the development of coconut organs is blank at present, which leads to the quality and quality of coconut can not be guaranteed. To solve this problem, we use CT to carry out non-destructive scanning, to obtain different kinds, different stages of coconut internal image images. In this paper, a novel YOLO V5 model was designed to detect Coconut haustorium and plumule, which are the key structures of Coconut. In order to improve the detection accuracy of coconut haustoriums and buds, an ASFF mechanism and a GAM module were incorporated into YOLO V5 model. The experimental results show that the model can detect the haustoriums and plumules in different periods, and the average precision is improved to 92.43%, which meets the need of high precision detection, thus, it can provide information for intelligent prediction of coconut development for decision making.

I. Introduction

The annual demand for coconuts in our country is as high as 2.5 billion, but due to the particularity of the growing environment of coconuts, only the Hainan area is more suitable for cultivation, and the output of coconuts in Hainan accounts for more than 90% of the national total, with an annual supply of about 233 million, it is extremely dependent on imports, and coconut itself can not only be eaten as a fresh fruit, mature coconut clothing, fruit, shell, water, meat and other deep-processing products are widely used in chemical raw materials, food, cosmetics, pharmaceuticals and boards and other categories, with the increasing demand for coconut raw materials, many countries begin to brew a policy of banning the export of coconut. Our coconut industry will probably face the severe situation of no fresh fruit to eat and no old fruit to use [1]. Therefore, based on the tropical climate and rich variety resources in Hainan, it is very important to improve the coconut cultivation process to meet the huge demand. And because the coconut has a very special structure, its shell is a very strong material, on the one hand to protect the growth of internal tissue, on the other hand, also hindered the observation of its development, at present, we can only rely on the experience of artificial planting to determine the timing of water and fertilizer [2]. Coconut haustorium is an embryo located near the germinating hole. When the germinating conditions are right, Coconut haustorium will gradually grow into puffy tissue filled with coconut cavity. Then the coconut fruit continues to grow, and the embryo begins to germinate. At this point, the haustoriums provide nutrients for the coconut bud to absorb. As the coconut plumule continues to grow, the whole coconut fruit gradually moves towards the end stage, in addition, some coconuts have stunted development or pathological changes, leading to the loss or disappearance of haustoriums and the breaking of plumules. Therefore, these two structures are important markers in the whole development process of coconuts, it is essential that they be non-destructive tested in a non-destructive manner. In view of this kind of present situation, we propose this kind of solution, first under the premise that does not destroy the coconut itself structure, uses the CT to carry on the non-destructive scanning, obtains the internal physiology image, then on this basis, real-time capture of coconut haustorium and plumule, to provide information support for growers.