Damir Krstinić

Intelligent forest fire monitoring system

This paper presents iForestFire, an Environmental Monitoring Information System for forest fire protection. The system is composed of several components, each having a particular function. Automatic fire detection is a crucial component of the system. It is based on various complex image processing algorithms. Complexity of the system also emerges from integration, based on multi agent technology, of different environment information. The presented system contributes to the environment protection and is in use in Croatia for several years.

Visual spatial-context based wildfire smoke sensor

Sensors for early fire detection based on visual analysis have been under constant development and improvement, especially during the last decade. However, there is still a lot of room for advancement to increase the accuracy and reliability of such sensors. In this paper, a novel method for wildfire smoke detection based on spatial context analysis as well as motion detection, chromatic, texture and shape analysis is introduced. Several measures for evaluating quality of smoke detection are used, both on image and pixel scale. Smoke is a semi-transparent and amorphous phenomenon whose boundaries are hard to determine precisely; therefore, fuzzy measures are introduced for assessing the detection error. The proposed method is evaluated using the proposed measures and compared with two existing methods. The results show that the wildfire sensor based on proposed method is capable of detecting fire-smoke accurately and reliably, and in most detection aspects it outperforms the existing methods.

Laser spot tracking based on modified circular Hough transform and motion pattern analysis.

Laser pointers are one of the most widely used interactive and pointing devices in different human-computer interaction systems. Existing approaches to vision-based laser spot tracking are designed for controlled indoor environments with the main assumption that the laser spot is very bright, if not the brightest, spot in images. In this work, we are interested in developing a method for an outdoor, open-space environment, which could be implemented on embedded devices with limited computational resources. Under these circumstances, none of the assumptions of existing methods for laser spot tracking can be applied, yet a novel and fast method with robust performance is required. Throughout the paper, we will propose and evaluate an efficient method based on modified circular Hough transform and Lucas–Kanade motion analysis. Encouraging results on a representative dataset demonstrate the potential of our method in an uncontrolled outdoor environment, while achieving maximal accuracy indoors. Our dataset and ground truth data are made publicly available for further development.

Vision based wildfire and natural risk observers

Wildfires are natural risk phenomena that cause significant economic and environmental damage. In wildfire fighting strategy it is important to detect the wildfire in its initial stage and to apply, as soon as possible, the most appropriate fire fighting action. In both cases wildfire monitoring and surveillance systems are of great importance, so in the last decade the interest for various wildfire monitoring and surveillance systems has increased, both on the research and the implementation level. This paper describes one such system named iForestFire. It is an example of advanced terrestrial vision based wildfire monitoring and surveillance system, today widely used in various Croatian National and Nature Parks and regions, but it is also a system in constant development and improvement, both on theoretical and practical level. This paper describes its last improvements in video detection part that are based on notation of observer, cogent confabulation theory and mechanism of thought. Inclusion of cogent confabulation theory allows us to expend the use of existing wildfire observers to more general natural risk observers.

Cogent Confabulation based Expert System for Segmentation and Classification of Natural Landscape Images

Ever since there has been an increase in the number of automatic wildfire monitoring and surveillance systems in the last few years, natural landscape images have been of great importa ...

Holonic Multi Agent System for Data Fusion in Vehicle Classification

In this paper we describe holonic organization of a multi agent system for automatic vehicle classification in a road toll system. Classification of vehicles in road toll systems is based on physical vehicle features and in this paper we focus on axle counting as the first discriminant feature for class determination. Our system relies on two main sensors—video camera and depth sensor. Video image and depth image processing is performed in several holons. The results from individual holons are fused into the final decision on a number of axles of a passing vehicle. We show that fusion of results from individual holons gives more precise results than individual holons. Holonic organization of the system aids scalability and simplifies inclusion of new sensors and new algorithms.

Grid-based mode seeking procedure

We propose a novel mode seeking and clustering procedure based on the estimation of the discrete probability density function of the data set. The discrete density is estimated in two steps. Initial density estimate is acquired by counting data samples which populate each cell of the discretized domain. Based on the initial density estimate, each cell of the discretized domain is assigned a variable bandwidth kernel, which is afterwards used to compute final discrete density estimate. Modes of the estimated density, corresponding to the patterns in the input data, are obtained by hill climbing procedure. The proposed technique is highly efficient, running in time linear to the number of input data samples, with low constant factors. The proposed technique has no embedded assumptions about structure of the data or the feature space, like number of clusters or their shape, thus arbitrarily structured data sets can be analyzed.

Color cues in Human Motion Analysis

An important part in the analysis of human activity video data is the silhouette segmentation. The results of segmentation are greatly affected by imaging environment, thus posing a problem for the extraction of the region in the image containing a subject of interest. In this paper we propose a human motion analysis method based on nonparametric clustering of monocular color images. Motivated by the need to automatically extract human silhouettes for kinematic gait analysis, without spurious segmentations naturally occurring in other segmentation methods, we have developed and applied adaptive mesh-based color clustering which can be combined with motion segmentation. The advantage of our method is that it is controlled by few intuitive parameters allowing the method to be adjusted to different capturing environments. The usage of color is not sensitive to illumination variations, or to different color distributions of different cameras, as the color distribution is compared between different regions of the same frame. To provide a qualitative evaluation of our method, our results are compared with the Gaussian Mixture model, a standard vision-based human subject extraction method based on background segmentation techniques.