Feature Extraction and Identification of Sugarcane Bud Based on S Component in HSV Model

Abstract

Background: Accurate identification and positioning of sugarcane buds is not only the prerequisite for the automatic cutting of sugarcane bud, but also the premise of intelligent directional planting of sugarcane. Objective: In this study, a feature extraction and recognition method for sugarcane bud was proposed based on the S component in HSV model to identify the position of sugarcane bud center in the sugarcane seed image. Method: According to the observation and analysis of 3D image of the S component in a single sugarcane bud segment after median filtering, the S component value of the stem node area containing the sugarcane bud was concentrated, and there were continuous depressions on the edge of the sugarcane bud. A rectangular template with width of 350 pixels and step length of 35 pixels was created to search for the position with the maximum average component value, so as to indirectly determine the position of the stem node. The 3D geometric space of the S-component was analyzed, and the characteristics of the edge of sugarcane bud in the longitudinal section curve were extracted as the troughs. Finally, the characteristic parameters of the sugarcane bud were used to determine the size and position of the ellipse that approximated the shape of the sugarcane bud, indirectly locate the sugarcane bud, and effectively retain the characteristic information of the sugarcane bud. Results: The test results showed that the success rate of sugarcane bud identification was 90%, and it took 1.58 s to identify an image on average. Conclusion: The method can effectively reduce the deviation of the location of sugarcane bud center, prevent the sugarcane bud damage due to the wrong location of the sugarcane bud and the low germination rate caused by the sugarcane bud facing downward during the planting process, and provide a research foundation for the intelligent and precise planting of sugarcane.