A visual extraction method of multi-dimensional motion features of pedestrians in traffic environment

Abstract

The invention discloses a method for visually extracting pedestrian multi-dimensional motion features in a traffic environment, comprising: step 1: building a pedestrian motion database; step 2: extracting pedestrian detection frame images of the same pedestrian in consecutive image frames; step 3: extracting the same pedestrian motion HOG features of the energy map; Step 4: Construct a pedestrian movement and posture recognition model based on Elman neural network; Step 5: Use the pedestrian movement and posture recognition model based on Elman neural network to judge the pedestrian posture in the current video; Step 6: Calculate and obtain pedestrians The real-time speed of the pedestrian is obtained from the instantaneous speed sequence in the direction of the X-axis and the Y-axis; Step 7: According to the three-dimensional scene in the intersection environment, the position information of the pedestrian in the image is obtained in real time, combined with the posture and real-time speed of the pedestrian, and the real-time speed of the pedestrian is obtained. sporty features. This scheme has the characteristics of high identification accuracy and good robustness, and is easy to apply, so it has a good application and promotion space.

Description

technical field [0001] The invention belongs to the field of traffic monitoring, in particular to a method for visually extracting multi-dimensional motion features of pedestrians in a traffic environment. Background technique [0002] In recent years, with the rapid development of science and technology, more and more intelligent methods have been applied to transportation, especially in the field of intelligent driving. Traffic safety is an eternal topic. In collision accidents, collisions between vehicles and pedestrians also account for a large proportion. The timely detection and posture recognition of pedestrians is the key to the current intelligent traffic active protection system. To achieve accurate identification, the most important thing is pedestrian motion feature extraction. [0003] Pedestrian pose recognition includes global feature method and local feature method. The global feature mostly adopts the method of moving historical images, that is, the frame...

AI Technical Summary

Problems solved by technology

The patent has the following problems: 1. When extracting the motion vector of each pixel, the pixels are not effectively screened, the amount of data is large, and the calculation is complicated; 2. The clustering algorithm used in the patent is prone to local convergence

Images