Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Accurate peduncle detection in 3D space is a vital step in reliable autonomous harvesting of fruits, as this can lead to precise cutting while avoiding damage to the surrounding plant. This study presents a visual detection method for the challenging task of detecting peduncles of sweet peppers. This method makes use of both color and geometry information acquired from sweet pepper point cloud and utilizes a supervised-learning approach for the peduncle detection task. First, a whole sweet pepper point cloud was subdivided into fruit and peduncle to provide harvesting system with a more comprehensive perception capability. Then, an improved 3D descriptor with the fusion of Hue, Saturation, Value (HSV) features and fast point feature histogram (FPFH) features was extracted separately from the preprocessed point clouds. At last, the HSV-FPFH descriptor was used for developing partial least squares discriminant analysis (PLSDA) classification model and yielded the 88.96% average classification accuracy for challenging sweet pepper point cloud dataset, which is superior to the 83.60% average accuracy based on the HSV descriptor. Compared with the HSV descriptor, the HSV-FPFH descriptor proposed in this study can meet the accuracy requirements of automatic harvesting systems better.