Detection and classification of moving objects

Abstract EN

Surveillance systems.is the process of monitoring the behavior of people, objects or processes within systems for conformity to expected "or desired standards in trusted systems for security.

In tracking safety specific regions such as banks, roads, boundaries, forests and so on, video surveillance systems are becoming more" common every day.

As monitoring systems transition from analog to digital devices and numbers rise, there is a need to interpret the captured video automatically".

The objective of this paper is to design a system that can automatically able to detect objects in different" scenes and then classify them.

The proposed system consists of two stages: moving object detection and classification".

The first stage consists of four steps: The first step is input video, the collection database is used in this paper where some videos are captured using fixed camera and other images are acquired from various internet sites.

The second step is the pre-processing of frames using HSV color space, filtering them using Median and Gaussian filters.

The third step is detection object using background subtraction method.

The fourth step is feature extraction of object.

In this paper, 21 features of shape features are extracted.

In the second stage of proposed system is classification the object in to (human and car) using Knearest neighbor.

Two different ratios of training/testing groups which are (50% - 50%) and (70% - 30%) are applied to the classifier.

K-NN gives training accuracy 100 %.

The accuracy of testing is 98.6111% and97.7% when the ratio is (50% - 50%) and (70% - 30%) respectively.

MATLAB R2018 is used to design the proposed system.