IndexIntroductionBackground SubtractionFrame Segmentation BlockFeature ExtractionFeature VectorSVMS-Based ClassificationExperimental ResultsExperimental EnvironmentPerformance AnalysisConclusionFires are a major cause of death and property destruction that they cost billions in losses every year. Smoke is the initial stage of fire ignition. If you can accurately detect smoke in its early stages, large fires can be prevented. This research proposes a video-based smoke detection approach based on smoke growth analysis. In this proposed model, the preprocessing phase; Image frames are extracted from videos. From the extracted frames, the background subtraction method was applied to subtract the background and obtain the objects using the Gaussian mixture model (GMM). Now, the stolen objects are put through a process to extract their feature. To extract features, we proposed a method that determines the growth region of the object using feature vectors. Finally, these features are fed to a support vector machine (SVM) to discriminate and cluster our data in a more usable way, paving the way to providing accurate smoke detection. Furthermore, a growth analysis histogram was provided to observe the movements of the objects. The proposed approach can work more effectively with a low false alarm rate. Say no to plagiarism. Get a tailor-made essay on "Why Violent Video Games Shouldn't Be Banned"? Get an original essay IntroductionDetecting fire in its early stage can prevent mass destruction and save thousands of lives and valuable property. Smoke components represent the beginning or initial phase of a fire. That's why early smoke detection can provide an effective approach to alerting you to the occurrence of a fire. Fires cause economic degradation and environmental damage2, so early fire detection can avert the fire, minimize damage, and save lives and property. Video-based fire detection system has been developed as a new technique in recent years. Its effectiveness in fire detection quickly made it popular. Higher accuracy and low false alarm rate made it more robust and dynamic. The use of surveillance in smoke detection is the strong point to meet the requirements and essentiality in large rooms and tall buildings or in the woods. There are many methods for smoke detection in the field of computer image processing. And most of them even use combined approaches of different methods to improve performance and reliability. Some methods are similar in most smoke detection approaches; These are the stages of motion detection, particle region analysis, dynamic volume analysis and smoke classification. The growth rate is calculated from the increasing number of pixels of the moving particle area. The differences are observed by the algorithms used in those different phases. All the methods and techniques mainly use three phases, preprocessing phase, feature extraction and classification. Researchers from all over the world have worked on this new technique and have produced numerous works.R. Islam used optical flow characteristics for fire alarm systems. Y. Cappellini has created an intelligent system for the automatic detection of fires in forests. G. Healey have created a system forreal-time fire detection. H. Yamagishi worked on an algorithm that uses a color camera for fire flame detection. But fire detection based on flame detection was very inaccurate. The traditional fire detection system uses electrical sensors that work as a heat transfer through the sensors. Any type of interference in the sensor could delay the alarm time or generate a false alarm. Then new methods are developed for detecting fire and smoke. Movement and color are very important characteristics of fire and smoke.Y. Chunyu worked on a video-based fire and smoke detection system using motion and color features. I. Kolesov worked with optimal mass transport based on optical flow and neural networks for fire and smoke detection in video. Wavelet in smoke, image processing for automatic smoke detection and adoptive background modeling for real-time monitoring have been developed to build a very effective and accurate fire alarm system. In this paper, we proposed an algorithm that uses image processing and works on smoke growth in its region. Smoke has growth in the region that is different from the growth of non-smoke particles over a certain period of time. The features are classified by a Support Vector Machine or briefly called SVM. The proposed smoke growth smoke detection model This paper presents a technique for motion detection that organizes various novel mechanisms. In this paper, the preprocessing stage combines background subtraction and image segmentation. We used the Support Vector Machine (SVM) algorithm to classify moving objects that should be smoke. The characteristics of the proposed focused smoke approach are the smoke growth area and the growth rate over time. A flowchart of the proposed smoke detection method is shown in Fig. 2.1 and are further discussed in more details below. Frame extraction Several videos are taken from surveillance. Some videos are shot with cameras and phones that ignite smoke intentionally for experiments as a dataset. Then we extracted 4 frames from the videos with an existing algorithm. In our case, a 10-second video was divided into several frames at 1 second interval, among which 5 frames at 2, 4, 6, 8 and 10 seconds were selected for further analysis. Background Subtraction Extracted frames are required to be subtracted for foreground objects. There are several existing methods for background subtraction. In this experiment, Gaussian mixture model (GMM) is used for background subtraction of frames. GMM subtracts the background from the foreground. Uses a threshold value that determines the gray level. And normalize the image to a gray image depending on the threshold value as color depth. Frame Segmentation Block Frame segmentation is an important part of frame analysis. In our case, we segmented each frame into blocks of 16 densities which are subsequently used for further analysis of smoke growth. The following table shows the annotations of each frame segmentation called density blocks with their respective percentage growth values. Feature ExtractionIn this section, based on the binary model, local binary models are introduced to classify and characterize smoke. All these determined features are extracted based on a block method. In the studyof machine learning, pattern recognition, and image processing, feature extraction starts from an initial set of measured data and constructs derived values ​​that are effectively the features, intended as informative and necessary, simplifying future stages of learning and generalization, and in some cases, leads to better human interpretations. Feature extraction is related to differences in dimensionality reduction. When the input data for an algorithm is too large to process and is suspected to be unnecessary (e.g. the same measurement in both feet and meters, or repetitive presentation of the image as pixels), then it can be simplified as a reduced set of features (also called a feature vector). Determining a necessary subset of the initial features is called feature selection. The selected features contain the relevant information from the input data, so that the intended action can be performed using this reduced simplified representation instead of all the initial data. The following smoke frames, related characteristic tables and smoke growth diagrams are shown according to the corresponding letter a, b, c. The frames extracted from the video are used to calculate the growth values, and these values ​​are subtracted from the next frame to obtain a percentage growth value in each density block. Feature Vector The feature vector is the average feature of the features of five frames. To achieve greater accuracy and error handling we not only extracted the smoke frame, but also smoke-like objects such as car lights and light bulbs on the wall. Car lights and bulbs give a similar view of smoke and therefore cause confusion. This makes classification more difficult. Therefore the accuracy may be reduced, which results in a high false alarm rate. In this article we have attempted to reduce this type of disturbance by also experimenting with non-smoking objects. The feature vectors are the density of smoke particles contained in a segment. Smoke particles gradually fill the block. Almost all segmented blocks are covered with a certain amount of smoke particles. SVM-based classification In machine learning, support vector machines are called supervised models associated with particular learning algorithms that analyze data used for clustering, classification, and regression analysis. Given a set of training datasets, each of them classified as one or more certain categories. An SVM model is a representation of the dataset as points in space and mapped so that the dataset is divided into different categories. New examples are then inserted into the one with the same gap and predicted into a category, based on which side of the gap they fall on. Not only can it perform a linear classification approach, but SVM can also perform a nonlinear classification task using the kernel trick method. supervised learning is not possible if the data is not labeled. In such a case, an unsupervised learning approach is required, which intends to find a natural grouping of data into various categories. It then maps the new data to these formed categories. The clustering algorithm provides an improvement to support vector machines called support vector clustering. which is often used in industrial applications when data is not labeled or some data is labeled as preprocessing for a classification. We looked at these.