摘要：The paper studied weeds classification system using BP neural network and 6 shape parameters (the ratio of the width and the length, complete degrees, the roundness, the rectangle, ratio of the framework proportion and frame perimeter) based on such characteristic parameters as weed leaf area, perimeter, minimum bounding rectangle, circumcircle, equivalent oval as input feature vectors; and meanwhile trained and improved the system. The experiment results showed when the plant coverage was low, the classification system could identify different weeds better; otherwise, the correct recognition rate was lower.
