Research On Active Contour Model For Plant Image Segmentation

For plant research,plant image obtained solely from equipment such as cameras,drones and satellites can not help humans understand the current state of the plant's growth.In order to obtain the hidden information,it is necessary to extract the target object by image segmentation and study it deeply.In this way,the information of plant leaves or petals can be obtained effectively,and the interference of other objects can be eliminated.At present,one of the main methods to complete image segmentation is the active contour model.Due to the complexity of the plant image background and the inherent shortcomings of the active contour model,on this basis,it is necessary to improve the active contour model to realize the segmentation of plant image.Therefore,the target object can be segmented in the end.The main work is as follows:(1)The parametric active contour model based on vector field is influenced by the vector field when the target object is extracted.It often appears the problem of equilibrium point,such as saddle point,static point,etc.The premature convergence of the contour curve can not enter the depression area,especially the deep and narrow concave region.In this paper,a method of fusion concave point detection and affine transformation is proposed to improve the ability of the contour curve to enter the concave region.(2)For parametric active contour models,the initial contour can not be automatically split and merge,resulting in the problem of the parametric active contour can not use an initial contour to extract multi-target objects.In this paper,by integrating the theory of sparse representation and decomposition,the parametric active contour model based on vector field is improved.It not only successfully solved the shortcomings of the parametric active contour model,but also the accuracy of the segmentation results is improved.(3)Although the geometric active contour model can make the contour curve split and merge automatically,it is much more affected by the noise,and it is easy to extract some unnecessary objects.Therefore,it is more difficult to realize the segmentation of plant images with complex background.In order to solve these problems,so the advanced bi-convex fuzzy variational model is improved in this paper.The robustness of the model to noise is increased and the target object of interest can be extracted accurately.(4)Finally,the improved active contour model is applied to the segmentation of plant images.The improved model is also extended to other images to verify the universality and research value.