Abstract
To overcome the limitations of the existing non-contact detection methods of sugarcane nodes, a dynamic detection method was developed based on linear array charge coupled device (CCD) photoelectric signal processing, with the goal of reducing the cost of mechanized sugarcane production and enhancing the efficiency of automatic pre-cut seeds production. First, control the mechanical drive unit to drive a photoelectric detection system to acquire signals of the vertical projection of the sugarcane profile. Then, use the Otsu algorithm for binarization to transform the projection information into profile information. Afterward, the profile signal was decomposed by a variational mode decomposition algorithm that has been optimized based on the sparrow search algorithm, and the component representing the node region was seen as the feature signal. Finally, in the normalized feature signals, the position whose wave peaks were higher than the judgment threshold was considered the position of sugarcane node. Also, one-factor and two-factor experiments were conducted to determine the effects of scanning speed and illuminance on identification accuracy. The results showed that when the scan speed of was 75 mm/s and illuminance was 91.91 lx, the identification rate, average response time and average error value reached 98.42%, 0.13 s and 1.36 mm, respectively. Compared to other node identification methods mentioned in this paper, the method proposed in the paper has a high identification rate and accuracy with faster response, which can enhance the efficiency of sugarcane seed production automation.