Detection of moving objects using thermal imaging sensors for occupancy estimation

The article considers the method of image processing proposed by the author in relation to the problem of automatic detection of moving objects in optoelectronic thermal imaging systems. Moving objects on the observed scene are subject to investigation, so it is advisable to use algorithms based on background subtraction methods to solve the detection problem. However, the observed objects may include objects of interest (a person, a vehicle), as well as other objects and background elements that increase the noise component of the observed situation. Also, the increase in the noise component is greatly influenced by false segmentation in the foreground of the areas of processed images when transferring the field of view of the sensor of the optical-electronic surveillance system. The purpose of this article is to prove the reduction of the probability of false alarm of an automatic detector due to the author's proposed approaches to image processing. The research uses the mathematical apparatus of probability theory and simulation with subsequent statistical processing of data. The article shows that the probability of a false alarm of an automatic detector based on the background subtraction method increases significantly after the transfer of the field of view of the sensor of the optical-electronic surveillance system and decreases after the movement stops as the areas of the processed image that are falsely highlighted in the foreground are automatically segmented. The simulation showed that the approaches proposed by the author can increase the peak signal-to-noise ratio of processed images and reduce the probability of a false alarm of the automatic detector of objects of interest. The results obtained show the feasibility of adapting detection algorithms based on background subtraction methods to work in scanning optoelectronic surveillance systems.

Download Full-text

THERMAL 3D MODELS ENHANCEMENT BASED ON INTEGRATION WITH VISIBLE IMAGERY

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ◽

10.5194/isprs-archives-xlii-4-w18-263-2019 ◽

2019 ◽

Vol XLII-4/W18 ◽

pp. 263-269

Author(s):

F. Dadras Javan ◽

M. Savadkouhi

Keyword(s):

High Resolution ◽

Thermal Imaging ◽

Point Cloud ◽

Point Clouds ◽

3D Models ◽

Thermal Modelling ◽

Surface Models ◽

Geometric Quality ◽

Imaging Sensors ◽

Visible Imagery

Abstract. In the last few years, Unmanned Aerial Vehicles (UAVs) are being frequently used to acquire high resolution photogrammetric images and consequently producing Digital Surface Models (DSMs) and orthophotos in a photogrammetric procedure for topography and surface processing applications. Thermal imaging sensors are mostly used for interpretation and monitoring purposes because of lower geometric resolution. But yet, thermal mapping is getting more important in civil applications, as thermal sensors can be used in condition that visible sensors cannot, such as foggy weather and night times which is not possible for visible cameras. But, low geometric quality and resolution of thermal images is a main drawback that 3D thermal modelling are encountered with. This study aims to offer a solution for to fixing mentioned problem and generating a thermal 3D model with higher spatial resolution based on thermal and visible point clouds integration. This integration leads to generate a more accurate thermal point cloud and DEM with more density and resolution which is appropriate for 3D thermal modelling. The main steps of this study are: generating thermal and RGB point clouds separately, registration of them in two course and fine level and finally adding thermal information to RGB high resolution point cloud by interpolation concept. Experimental results are presented in a mesh that has more faces (With a factor of 23) which leads to a higher resolution textured mesh with thermal information.

Download Full-text