基于小波重构与FastICA的巨菌草种茎节点快速检测算法研究OA

To meet the node positioning requirements in the automated production of Pennisetum giganteum seedlings,this paper pro-posed a detection algorithm based on the wavelet multi-scale denoising and fast independent component analysis(FastICA)method.The contour signal of seed stem was obtained by a laser displacement sensor,and the db5 wavelet base was used for eight-layer dis-crete wavelet decomposition.The fifth to seventh layer coefficients were selected for multi-threshold processing and reconstruction to suppress noise interference.By combining this with the time delay embedding technology,a multidimensional observation matrix was constructed,and the FastICA was used to realize blind source separation.Fuzzy constraints were applied to node localization based on the biological characteristics of internodes.The experiment involved 100 Pennisetum giganteum seed stems,totaling 600 nodes with buds,for verification.The results showed that the node positioning accuracy was 100%,the maximum error was 1.10 mm,and the average absolute error was 0.42 mm.When the sensor moved and collected data at a speed of 63 cm/s,the average detection time was 0.19 seconds,and both the accuracy and speed met the requirements of the production line.The algorithm provided an ef-fective technical scheme for automatic seed production of Pennisetum giganteum seedlings.