Monocular multi-frame video image splicing system, method and device and storage medium

Abstract

The invention provides a monocular multi-frame video image splicing system, method and device and a storage medium. The system comprises a server; a SOC chip used for decoding, wherein the SOC chip comprises a comparison module which is connected with the server, at least one video data stream channel is connected with the SOC chip, and the video data stream is decoded into a picture to be identified frame by frame; a graphic acceleration card formed by a programmable logic device, wherein the graphic acceleration card comprises a neural network module, the neural network module recognizes a to-be-recognized picture frame by frame, when it is recognized that one to-be-recognized picture contains a preset target, a target initial image area occupied by the target is obtained, the to-be-recognized picture serves as a first frame image, and through comparison by a comparison module, the target newly-added image areas after the preset target moves in the picture between the adjacent frames are obtained in sequence and are spliced to one side of the target initial image area, monocular multi-frame splicing can be achieved, the complete vehicle picture is obtained based on the multiple local pictures, and the inspection accuracy is improved; a graphics card is not needed, and the equipment cost is reduced.

Description

technical field [0001] The present invention relates to a graphics deep learning device, in particular to a monocular multi-frame video image splicing system, method, device and storage medium. Background technique [0002] At present, cameras are widely used in scenes such as container terminal gates to assist staff in checking vehicle-related information, and some of the information requires staff to be able to intuitively and comprehensively see the complete situation of vehicles and containers. [0003] The traditional solution will provide the staff with multiple continuous and independent screenshots of the vehicle, and then let the staff manually check these multiple pictures. The traditional method will increase the workload of the staff and increase the possible errors in the inspection. [0004] Moreover, at some gates where vehicles pass through at high speed, since the collected pictures are only part of the car and cannot reflect the view of the entire car body...

AI Technical Summary

Problems solved by technology

The traditional method will increase the workload of the staff and increase the possible errors in the inspection

[0004] Moreover, at some gates where vehicles pass through at high speed, since the collected pictures are only part of the car and cannot reflect the view of the entire car body, it is difficult to obtain a complete car body data number in one picture, and it is also difficult to identify the car body, car body, etc. Image recognition is performed on the top and bottom of the vehicle, and the system cannot automatically detect the safety of the vehicle (for example: scanning suspicious items on the bottom or roof of the vehicle through image recognition, etc.)

[0005] The existing technology mainly uses two methods to realize graphic machine learning. The first one is to use CPU decoding, and then realize the splicing task through CPU resources, which consumes a lot of CPU resources. Once the resources are consumed too much, the running speed of other services will be reduced. And the cost is high; the second is to use the graphics card to realize the splicing task, the cost of the graphics card is relatively high, and the power consumption is relatively high

Images