Accurately locating the root collar of seedling is important for the measurement of seedling morphological characteristics in the computer vision system of seedling grading. A RC ASL ( Root Collar Automatic Searching and Locating ) method based on fuzzy neural network (FNN) has been studied to resolve the key problem in the field of conifer seedling feature extraction. In the computer vision system

the features of the object may be extracted based on the hue variation of the image. Three line features of the seedling image are selected based on the hue variation in every line of the image as the input vector of the FNN

And the corresponding membership function has been chosen to accomplish the fuzzy processing of these features. BP algorithm is selected to optimize the weights of the membership function and the stimulation function of the neural networks. With the training of the FNN

a FNN structure used to search and locate the root collar of seedling is achieved. The foliage

stem

root and the root collar of the conifer seedlings is recognized from the seedling image and then the root collar is located by the FNN. The experimental results show that this method has the capability to meet the requirement of seedling grading operation.