Research On The Segmentation Algorithm And Recognition System Of Trunk Image In Complex Background

This type of machine vision navigation technology, is now popular in computer vision research.A basic part of Forestry Harvesting Robot’s auto navigation work in this paper.Trunk image processing and segmentation the feature extraction by the camera captured image.These pictures with the target trunk separated from complex background.A technique to obtain the image feature point data of the trunk.Based on the research and analysis on conventional image segmentation algorithm, this paper has given a algorithm which can be applied in the trunk image segmentation,according to the characteristics of the trunk image and a priori knowledge the automatically recognition on trunk image has been initially realized. The main contents and conclusions of this study are as follows:1.The trunk image be segmented by the commonly used algorithms and the results shown that the commonly used algorithms can not separate the trunk under the complex background.2.Based on full knowledge of all common image segmentation technology premise,summed up the way to meet the complex context of the trunk split image. Firstly, the tree trunk image is detected by using semi-threshold segmentation; then, filtering method of mean and median is employed to smooth image, as after using edge detection operator made ample preparations are used for image edge detection.3.The traditional edge detection algorithm makes use of some traditional classical operator for image edge detection, and its main idea is in horizontal and vertical directions for gradient calculation, but such detection under complicated background image edge is not fully connected, being disconnection of a certain degree. The BP neural network can be used conveniently to trunk image edge detection under complicated background. The neural network model which consist of three layers BP network can make the edge detection of the trunk.A small pieces of image is regard as value space.Traditional edge detection can detect the different gray level of pixels in the image.It can be used to train image between original image and target image. Then the network model of edge detection has been established, and it can achieve good recognition effect through experimental verification.4.Using VC++ and Matlab to develop trunk image recognition system.In short, the article conceived trunk image recognition algorithm,it is a basis for research in the field of intelligent harvesting robot, the results of this study can push a lot of other intelligent robot research process, it has provided a new method of robot resreach.