Unsupervised Greenhouse Tomato Plant Segmentation Based on Self-Adaptive Iterative Latent Dirichlet Allocation from Surveillance Camera

Qifan Cao,Lihong Xu

doi:10.3390/agronomy9020091

Abstract

It has long been a great concern in deep learning that we lack massive data for high-precision training sets, especially in the agriculture field. Plants in images captured in greenhouses, from a distance or up close, not only have various morphological structures but also can have a busy background, leading to huge challenges in labeling and segmentation. This article proposes an unsupervised statistical algorithm SAI-LDA (self-adaptive iterative latent Dirichlet allocation) to segment greenhouse tomato images from a field surveillance camera automatically, borrowing the language model LDA. Hierarchical wavelet features with an overlapping grid word document design and a modified density-based method quick-shift are adopted, respectively, according to different kinds of images, which are classified by specific proportions between fruits, leaves, and the background. We also utilize the feature correlation between several layers of the image to make further optimization through three rounds of iteration of LDA, with updated documents to achieve finer segmentation. Experiment results show that our method can automatically label the organs of the greenhouse plant under complex circumstances, fast and precisely, overcoming the difficulty of inferior real-time image quality caused by a surveillance camera, and thus obtain large amounts of valuable training sets.

Highlights

The vision information of greenhouse crops is of great significance for plant phenotype analysis, which has been applied to provide guidance to improve crop cultivation in many studies [1,2,3,4,5]
Our evaluation technique is based on the comparison of the segmentation results and groundby human labeling judging from two aspects
One is the pixel of theresults overalland image (IA), Our evaluation technique is based on the comparison of theaccuracy segmentation groundtruth by human labeling judging from two aspects

Summary

Introduction

The vision information of greenhouse crops is of great significance for plant phenotype analysis, which has been applied to provide guidance to improve crop cultivation in many studies [1,2,3,4,5]. It is obvious that an automatic, accurate, and high-throughput imaging processing technique—with high sensitivity for plant phenotypic research— lays a visual foundation for analyzing the effect on the environment in production, but is conducive to comprehensive research of internal and external factors on the physical and biochemical characteristics of plants. It serves as a non-destructive analysis method to make contributions to directive breeding, crop identification, and yield estimation. There is little open source data available in relation

Results

Discussion

Conclusion