Daniel Yoshinobu Takada Chino

A New Time Series Mining Approach Applied to Multitemporal Remote Sensing Imagery

In this paper, we present a novel unsupervised algorithm, called CLimate and rEmote sensing Association patteRns Miner, for mining association patterns on heterogeneous time series from climate and remote sensing data integrated in a remote sensing information system developed to improve the monitoring of sugar cane fields. The system, called RemoteAgri, consists of a large database of climate data and low-resolution remote sensing images, an image preprocessing module, a time series extraction module, and time series mining methods. The preprocessing module was projected to perform accurate geometric correction, what is a requirement particularly for land and agriculture applications of satellite images. The time series extraction is accomplished through a graphical interface that allows easy interaction and high flexibility to users. The time series mining method transforms series to symbolic representation in order to identify patterns in a multitemporal satellite images and associate them with patterns in other series within a temporal sliding window. The validation process was achieved with agroclimatic data and NOAA-AVHRR images of sugar cane fields. Results show a correlation between agroclimatic time series and vegetation index images. Rules generated by our new algorithm show the association patterns in different periods of time in each time series, pointing to a time delay between the occurrences of patterns in the series analyzed, corroborating what specialists usually forecast without having the burden of dealing with many data charts.

BoWFire: Detection of Fire in Still Images by Integrating Pixel Color and Texture Analysis

Emergency events involving fire are potentially harmful, demanding a fast and precise decision making. The use of crowd sourcing image and videos on crisis management systems can aid in these situations by providing more information than verbal/textual descriptions. Due to the usual high volume of data, automatic solutions need to discard non-relevant content without losing relevant information. There are several methods for fire detection on video using color-based models. However, they are not adequate for still image processing, because they can suffer on high false-positive results. These methods also suffer from parameters with little physical meaning, which makes fine tuning a difficult task. In this context, we propose a novel fire detection method for still images that uses classification based on color features combined with texture classification on super pixel regions. Our method uses a reduced number of parameters if compared to previous works, easing the process of fine tuning the method. Results show the effectiveness of our method of reducing false-positives while its precision remains compatible with the state-of-the-art methods.

BoSS: image retrieval using bag-of-superpixels signatures

Techniques of bags-of-visual-words based on signature have been employed in image retrieval and analysis, with the benefit of dismissing expensive clustering processes. However, the limitations of such techniques are the requirement of multiple parameters, which may be unintuitive and in most cases depends on the application domain. In this paper, we overcome these limitations by proposing Bag-of-Superpixel Signatures (BoSS), which extracts visual signatures using local features from superpixels. Moreover, our proposal also employs a fractal analysis to extract intrinsic information about the domain application and also to diminish the amount of parameters needed. The results demonstrated that BoSS achieved an improvement up to 31.2% in image retrieval precision during experimental evaluations over five distinct datasets. We conclude that BoSS introduces an intuitive, self-contained, scalable and effective approach for image retrieval using bags-of-visual words.

Discovering Frequent Patterns on Agrometeorological Data with TrieMotif

The “food safety” issue has concerned governments from several countries. The accurate monitoring of agriculture have become important specially due to climate change impacts. In this context, the development of new technologies for monitoring are crucial. Finding previously unknown patterns that frequently occur on time series, known as motifs, is a core task to mine the collected data. In this work we present a method that allows a fast and accurate time series motif discovery. From the experiments we can see that our approach is able to efficiently find motifs even when the size of the time series goes longer. We also evaluated our method using real data time series extracted from remote sensing images regarding sugarcane crops. Our proposed method was able to find relevant patterns, as sugarcane cycles and other land covers inside the same area, which are really useful for data analysis.

The SITSMining Framework

The amount of data generated and stored in many domains has increased in the last years. In remote sensing, this scenario of bursting data is not different. As the volume of satellite images stored in databases grows, the demand for computational algorithms that can handle and analyze this volume of data and extract useful patterns has increased. In this context, the computational support for satellite images data analysis becomes essential. In this work, we present the SITSMining framework, which applies a methodology based on data clustering and classification to extract patterns and information from time series obtained from satellite images. In Brazil, as the agricultural production provides great part of the national resources, the analysis of satellite images is a valuable way to help crops monitoring over seasons, which is an important task to the economy of the country. Thus, we apply the framework to analyze multitemporal satellite images, aiming to help crop monitoring and forecasting of Brazilian agriculture.