Abstract
Complicated image scene of the agricultural greenhouse plant images makes it very difficult to obtain precise manual labeling, leading to the hardship of getting the accurate training set of the conditional random field (CRF). Considering this problem, this paper proposed an unsupervised conditional random field image segmentation algorithm ULCRF (Unsupervised Learning Conditional Random Field), which can perform fast unsupervised segmentation of greenhouse plant images, and further the plant organs in the image, i.e. fruits, leaves and stems, are segmented. The main idea of this algorithm is to calculate the unary potential, namely the initial label of the Dense CRF, by the unsupervised learning model LDA (Latent Dirichlet Allocation). In view of the ever-changing image features at different stages of fruit growth, a multi-resolution ULCRF is proposed to improve the accuracy of image segmentation in the middle stage and late stage of the fruit growth. An image is down-sampled twice to obtain three layers of different resolution images, and the features of each layer are interrelated with each other. Experiment results show that the proposed method can segment greenhouse plant images in an unsupervised method automatically and obtain a high segmentation accuracy together with a high extraction precision of the fruit part.
Highlights
Till various image segmentation algorithms have been proposed in literature, among which the ones that can extract image features through statistical methods are important and practical scientific techniques
Ref.[22] proposed an algorithm called Spatial Latent Dirichlet Allocation (SLDA) to encode the spatial structure of visual words better. It designed the vision documents considering the spatial structure of image and got a better image segmentation result than that obtained by conducting LDA directly
The segmentation result of LDA is used as the initial labels of conditional random field (CRF)
Summary
Various image segmentation algorithms have been proposed in literature, among which the ones that can extract image features through statistical methods are important and practical scientific techniques. Shotton J. et al.[11,12] proposed a new approach to represent the features of image combined with boosting classifier It optimized the unary potential of Dense CRF, and the precision of segmentation can be improved even when the number of categories of objects in the image is large. Since CRF is a supervised learning model, generally, its unary potential is obtained in supervised methods It needs a high-quality training set containing a large amount of labeled images to learn related models of all kinds of objects. LDA can get more reliable label information than manual labeling to obtain the training set in the process of greenhouse plant image segmentation This method takes advantage of CRF to reflect the differences between pixels of different classes. Afterwards, we determine the label assigned to each pixel by computing the probability distribution
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
