Vehicle track statistic method based on Hadoop and monitoring video streams

Abstract

The present invention discloses a vehicle track statistic method based on Hadoop and monitoring video streams. The method involves a vehicle detection algorithm, a license plate recognition algorithm, distributed video processing and vehicle track statistics. According to the method of the invention, vehicle detection and positioning are performed on video frames on the basis of an improved tiny-yolo model; license plate regions are positioned from a positioned vehicle region on the basis of an HSV color model, an SVM classifier and a text positioning technology; license plate character recognition is performed through an improved LeNet-5 model; the processing algorithms of license plate detection and license plate recognition are compiled into a dynamic link library; the support of MapReduce for a video type format is expanded, and the Map interacts with the dynamic link library through a JNI interface, so that the distributed processing of the monitoring video streams can be realized; analysis results are summarized through three stages of Combiner, Partition and Reduce; and finally, the statistics of vehicle tracks and the storage of vehicle frames are realized. According to the method, the vehicle detection algorithm and license plate recognition algorithm have high adaptability to complex environments, and execution efficiency is higher through the interaction of the JNI interface and the MapReduce.

Description

technical field [0001] The invention relates to the technical fields of cloud computing and computer vision, in particular to a vehicle trajectory statistics method based on Hadoop and monitoring video streams. Background technique [0002] In the environment of rapid development of vehicle network and intelligent transportation, the popularity of traffic monitoring cameras has promoted the exponential growth of video data. If the effective information in video data cannot be fully exploited, resources will inevitably be wasted. In traffic monitoring video, As the main information, vehicle information plays an important role in the development of traffic supervision and intelligent transportation. How to fully mine vehicle information from massive monitoring data has become a research hotspot. However, the traditional centralized video processing method has the problems of insufficient processing capacity and unscalable in the face of massive growth of surveillance video dat...

AI Technical Summary

Problems solved by technology

However, urban roads and highways are dynamic environments that are easily affected by weather, light and other environments. Traditional processing algorithms are often difficult to adapt to changes in weather and light conditions.

[0004] Since the underlying mechanism of the Hadoop platform is implemented by Java, and image processing is often a computationally intensive task, the traditional methods of JavaCV and Hadoop Streaming are not efficient in the distributed processing of images through Hadoop, and MapReduce does not support direct processing from HDFS. Read video data, because FFMPEG does not support the HDFS protocol, the traditional use of FFMPEG to split the video into frame sequences locally and upload to HDFS often results in additional I / O overhead

Images