Motion Tracking in the Construction and Storage Area with Perspective Approach

Türkçe

Occupational safety in the work area is not only to take measures after accidents, but to try to take precaution by determining the situations that will cause the accident. Image processing techniques are the process of re-visualizing recorded images or snapshots according to the desired requirements. For this reason, by using image processing techniques, an application can be realized to ensure the safety of the work by following the workers and other objects on the recorded or live images in the worksite. The aim is to try to avoid work accidents that may occur to a large extent and also to provide the use of multiple algorithms and features with person and vehicle analysis. In the study, the Matlab software program was deemed appropriate due to the density of mathematical algorithms. In this paper a motion tracking algorithm with the addition of new double-sided and adaptive perspective methods have been implemented. The main purpose of this paper is to propose a motion tracking algorithm to help solving safety concerns at the storage and construction areas. With the proposed algorithm, workers on the work site are detected and they are checked if they comply with the safety requirements at the work site by using computers entirely. Detailed algorithm explanations can be found in the methodology part of the paper. The live image was prepared ready in Matlab environment by using basic image processing techniques such as filtering and noise cleaning processes and then it was subjected to object finding and classification for motion tracking in the image. Proper detection and tracking of the object seems to be more successful with preprocessing. Luminance conditions, weather conditions and moving data density are important factors that affect accuracy. As a first step, identification of workers was done using simple classification techniques. These classification criteria are explained in the methodology section. The classification process provided the human data set for motion tracking in four steps. Field workers who do not wear helmets or not wearing vests are defined as “-” and those who follow the rules are indicated as “+” on the image. The performance of the algorithms on this type of images and the results of the improvements are evaluated in the last section.