CNN-based Video Time Discrimmination Method

With the development of information technology,electric pictures and videos have become a major way for people to acquire information.Video plays a more and more important role in people's life.Moreover,recently,digital resources have a explosive increase trend.Many videos don't have time information or lost their time information during transmit.The time information in a video can have many useful information for the following analyze.Hence,analyze video and get the time information in a video is meaningful.The paper implemented a video time discrimination system based on CNN.The main contents of this paper are as follows:(1)Studied and implemented an OCR system based on OpenCV and Tesseract.This system can automatically divide and tag pictures by time tag on the picture and based on this system I built a time divide picture database.(2)Studied and implemented a CNN based on AlexNet structure and time divide database.By adjust the net parameters,the net reached a better performance.Also used the test data to verify the performance of the system.(3)Studied and implemented a CNN based on GoogleNet structure and weather divide database.By adjust the net parameters,the net reached a better performance.Also used the test data to verify the performance of the system.(4)Came up with a fuse strategy for time net and weather net.Adjust the output of time net base on the influence of weather in different times.And tested the performance of the system using the test data and verified the result.