Research On Application Of Action Recognition In Traffic Command

Accompanied by the continuous investment of scientific and technological strength and information technology in social development,people’s living and social economy have achieved rapid development.At the same time,the urban traffic congestion is becoming more and more serious.Meanwhile,the advent of unmanned vehicles and the development of intelligent driving assistance system have become a new challenge to the traditional traffic command.In the field of artificial intelligence and machine vision,action recognition technology is gradually entering the mature stage.However,the research of traffic command action recognition is still a new and urgent research hotspot.Based on this background,this paper focuses on the study of action recognition in traffic command.The main research contents are as follows:1.The traffic command action which is achieved by using video capture device,often recorded the sustained lens content within a certain period of time.It contains both the superposition of different command movements and the repetition of the same command action in different cycles.This requires the use of movement characteristics of the rate of change for motion segmentation.For the binary image that has been extracted from the moving target area,it is necessary to obtain the edge of the human body,and then according to its edge changes to find the rate of change in the characteristics of the movement,and finally according to the rate of quantification curve definition action segmentation.2.In addition,the same command action contains the key frame posture with the expression of the decisive motion information,and the non-critical frame posture that does not contribute to the motion information expression,this requires the use of the apparent feature contour vector to extract the key frame posture.Firstly,the Candy operator is used to extract the contour and centroid of the target region firstly.Secondly,the distance from the centroid to the edge of the contour is used as the element of the contour vector,and compute the contour vector of each image and by optimizing the derivation.Through the quantitative vector curve to define the key frame posture and non-key frame posture,and complete the key frame posture extraction.3.In view of the feature extraction of traffic command action,this paper adopts the skeleton parameters as the selection feature,and the traditional algorithm of skeleton extraction based on distance transformation is proposed to optimize and improve.The traditional skeleton extraction algorithm based on distance transform is the result of the existence of error dispute in the determination of the maximum inscribed circle in the discrete domain,and the lack of adjacent points in the skeleton point as a reference,which makes it difficult to guarantee the connectivity and single pixel.In this paper,by defining the seed skeleton point and searching the whole region skeleton point with the growth model,it not only maintains the topological structure consistent with the original image,but also has good connectivity and single pixel,and the noise caused by the edge disturbance has a good shielding effect.4.In this paper,the conventional template matching method is improved to highlight the information expression of the arm.The method is based on the skeleton model extracted from the current standard traffic police command signal.When the distance between the templates is obtained by using the Hausdorff distance,the corresponding weights are set for different regions by analyzing the difference of the motion information contributed by the various parts of the body in different traffic command action skeleton models,and finally to enhance the traffic command action recognition system real-time,to improve the accuracy of recognition purposes.