Dangerous Behavior Detection Of Human Body On The Terrace Based On Machine Vision

With the development of information technology and the improvement of people’s living conditions,people’s awareness of family safety protection has been unprecedentedly enhanced.With the rapid development of network technology,people can still know the abnormal situation inside the house while not in the monitoring range.In everyday family life,the terrace is a dangerous area.On the one hand,in the outdoor part,the invasion of thieves may bring property damage to the family and even threaten personal safety.On the other hand,in the indoor part,parents can’t accompany children all the time,and children climbing the terrace lead to the tragedy of falling down.In response to the above problems,the detection of human dangerous behavior on the terrace was studied.In the process of risk behavior detection,the extraction part of human behavior features improves the LBP feature extraction algorithm and improves the accuracy of target detection.The morphological determination method based on the judgment of the centroid position is used to further reduce the false detection rate of the algorithm.The process of risk behavior detection is divided into three parts: the detection of moving targets,the extraction of human behavior characteristics and the classification of human behavior.The moving target detection part was subjected to the ViBe algorithm after comparison experiments.Because of its ability to use a single frame image to establish the characteristics of the background model,the operation speed and operational stability are improved.The image frame obtained by the operation is then processed to further reduce the influence of image noise.In order to improve the accuracy of the target detection,the process of the traditional LBP algorithm is analyzed.Since it only considers the point directly close to the center point of the pixel,ignoring the useful information of other surrounding points,thereby making the edge information less sensitive.By improving the utilization of the pixel point information around the LBP operator,on the basis of retaining the rotation invariance of the algorithm,the point closer to the position of the target pixel is given a greater weight,and the use of surrounding pixel information is enhanced.The rate is improved compared to the traditional LBP algorithm.The human behavior classification part adopts the SVM algorithm.Since there is no ready-made data set,it is chosen to shoot and create a training set for training.Because of its certain false positive rate,in the final process,by dividing the dangerous area,a morphological judgment method based on the judgment of the position of the center of mass is added,and the correct rate is increased to 94.6% by the combination of the two.The experimental results show that the design process of the human body dangerous behavior in the terrace can achieve the design requirements and obtain good experimental results.