Method Research And System Design Of Vegetable Seed Quality Detection Based On Machine Vision Technology

Seed is the most basic production material, so its importance is self-evident. Seed quality directly affects the level of crop yields and the income of farmers. Up to new, foreign companies have occupied more than 50% vegetable seed market share in the last 10 years, almost involving all the vegetable crops. Apparently, China is still at a disadvantage in the fierce competition of world seed market share.To enhance the competitiveness of Chinese seed industry market, the seed quality is the most important. Quality detection of vegetable seed is an essential part of improving seed quality. Conventional methods of seed quality detection could not meet current development demand of intelligent and automated breeding because of subjectivity, high cost and difficult operations. With the rapid development of computer and image processing technology, non-destructive detection technology based on machine vision emerges as the times require. In order to explore the new method of seed quality detection, in this paper, the non-destructive detection technology based on machine vision has been studied using four kinds of vegetable seeds (Cucumber, Chill, Tomato, Aubergine), and we designed the corresponding detection system. The study results could provide a strong theoretical basis and technical support for online and automatic seed breeding detection.The main contents of this paper are as follows:Firstly, this paper not only discussed the advantages and disadvantages of the non-destructive detection technology based on the optical properties, electrical properties and chemical characteristics, but also analyzed the significance, research status and development trends of vegetable seed quality detection based on machine vision technology.Secondly, the image pre-processing program of vegetable seeds, including image transformation, image denoising, threshold processing and morphological processing were researched. On this basis, seed area, seed perimeter, the color characteristics of the external characteristics and seed germination index, average root length, seed vigor index of seed growth characteristics were extracted, breaking the disadvantages of detecting seed quality relying solely on the seed external characteristics. Then, in order to further improve the automation of seed testing, we combined with BP neural network to assess vegetable seed quality.Finally, The Visual Studio 2012 was used to develop the seed quality detection interactive software, and it could be used in image acquisition platform. This system software can automatically extract seed feature and detect seed quality. The experimental results showed that the system has excellent detection accuracy.