Fusion of time series representations for plant recognition in phenology studies

Abstract

Nowadays, global warming and its resulting environmental changes is a hot topic in different biology research area. Phenology is one effective way of tracking such environmental changes through the study of plant's periodic events and their relationship to climate. One promising research direction in this area relies on the use of vegetation images to track phenology changes over time. In this scenario, the creation of effective image-based plant identification systems is of paramount importance. In this paper, we propose the use of a new representation of time series to improve plants recognition rates. This representation, called recurrence plot (RP), is a technique for nonlinear data analysis, which represents repeated events on time series into a two-dimensional representation (an image). Therefore, image descriptors can be used to characterize visual properties from this RP images so that these features can be used as input of a classifier. To the best of our knowledge, this is the first work that uses recurrence plot for plant recognition task. Performed experiments show that RP can be a good solution to describe time series. In addition, in a comparison with visual rhythms (VR), another technique used for time series representation, RP shows a better performance to describe texture properties than VR. On the other hand, a correlation analysis and the adoption of a well successful classifier fusion framework show that both representations provide complementary information that is useful for improving classification accuracies.

Spatio-Temporal Vegetation Pixel Classification by Using Convolutional Networks

“…Results are presented in the second part of Table IV. Once again, the worst result was yielded by the Recurrence Plots technique [13] (around 27%). Baselines based on ConvNets outperformed such technique achieving very similar performance (around 35%).…”

“…For the Itirapina dataset, obtained results are presented in Table III. The worst results were yielded by the methods based on RP-based technique [13]. Such approaches take only a few hours to be trained but are outperformed by all other approaches.…”

A Multirepresentational Fusion of Time Series for Pixelwise Classification

“…Two-dimensional representations were investigated in time series classification and retrieval problems in [4], [7]. In [7], the phenological visual rhythm was introduced to encode vegetation phenology changes into images.…”

“…Traditional color descriptors are employed in the characterization of such images. In [4], both recurrence plot and visual rhythm representations were explored in the context of a pixelwise classification fusion problem. Different from our approach, however, no data-driven features were used.…”

“…In particular, a promising family of methods relies on the use of two-dimensional representations [7], [12], [13], which can be transformed into images. The goal of such approaches is to benefit from successful computer vision methods proposed for image classification, to more effectively classify time series [4], [7], [14]- [16].…”