Krstinic' proposed a new color model called HS'I specifically for smoke detection. applied the rule to YCbCr color space and proposed a new representing method. statistically extracted the feature of smoke images in RGB color space and proposed a rule of color for smoke classification. Yuan proposed a method of using Local Binary Pattern (LBP) and the variance of Local Binary Pattern (LBPV) to detect smoke. Maruta proposed a method of using Co-occurrence Matrix to detect smoke. Ferrari proposed a method of performing wavelet transformation on images and building a Hidden Markov Tree Model to extract smoke texture. The performance of smoke detection can be improved if more discriminative static features are extracted from images.Ĭurrently, there are many methods proposed to extract static features of smoke image. It is very difficult for users to specify appropriate thresholds, which greatly affect experimental results. However, it is worth noting that dynamic feature extraction requires background modeling or frame differences, which are often based on thresholds. ![]() Smoke not only has some obvious features such as color, texture, and shape, etc., but also has important dynamic features such as flowing, and flicker, etc. Smoke is a very important clue for early fire detection. In addition, early flame may not necessarily fall into the video surveillance area due to the size of flame, occlusion and so on. It has become a more important branch of fire detection.Īccording to objects for detection, vision based fire detection methods can be classified into two categories: flame detection and smoke detection. Among a variety of methods for fire detection, computer vision based methods are considered promising because they are quick in responses, low in cost, and less susceptible to environmental factors, etc. Timely detection of fire is the basic guarantee of avoiding harm caused by fire, which is a serious threat to people's lives and property. Experiments show that our approach not only has better performance than existing methods in smoke detection, but also has good performance in texture classification. Finally, all the sub oriented histograms are concatenated together to form a robust feature vector, which is input into SVM for training and classifying. For each pair of the two discrete orientations, we generate a sub LBP code map from the original LBP code map, and compute sub oriented histograms for all sub LBP code maps. We propose to use two coordinates systems to compute two orientations, which are quantized into discrete bins. Since an LBP code is just a label without any numerical meaning, we use Hamming distance to estimate the gradient of LBP codes instead of Euclidean distance. We first extract LBP codes from an image, compute the gradient of LBP codes, and then calculate sub oriented histograms to capture spatial relations of LBP codes. ![]() In this paper, we propose sub oriented histograms of LBP for smoke detection and image classification. ![]() Local Binary Pattern (LBP) and its variants have powerful discriminative capabilities but most of them just consider each LBP code independently.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |