Convolutional Neural Network Cascaded Surveillance Image Vehicle Detection Method and System

Abstract

The present application relates to the field of vehicle detection, and specifically discloses a convolutional neural network cascaded monitoring image vehicle detection method and system thereof. It adopts the architecture of cascaded convolutional neural network, that is, firstly train the first-level convolutional neural network as a noise canceller, and then train the second-level convolutional neural network as a detector, so that the obtained The monitoring images are processed in order to detect the abnormal behavior of the working vehicles in the grain depot. In this way, the accumulation of errors in traditional detection can be avoided, so that the accuracy of abnormal behavior detection of the working vehicles in the grain depot is higher.

Description

Technical field [0001] The present application relates to the field of vehicle detection, and more particularly, involving a surveillance image vehicle detection method and a system thereof. Background technique [0002] My governments and people have always attached great importance to the safety of food. In recent years, in recent years, in order to fully protect my country's food security, under the leadership of the National Grain Bureau, all provinces and cities are actively promoting smart grain reservoirs, designed to be from the grain security system, grain Several aspects such as library information system, grain smart ventilation system ensure grain storage security. [0003] At present, the security system built in the grain library can monitor the grain library for 24 hours to ensure the safety of grain reservoirs, but the existing security system does not recognize the function, not considering effective analysis of image information of the collected grain vehicle. A...

AI Technical Summary

Problems solved by technology

[0004] However, the identification of grain transport vehicles in grain depots is more difficult than that of ordinary vehicles. The main reason is that the internal environment of grain depots is very complex. When grain transport vehicles work in grain depots, they will generate a lot of dust, resulting in the collected images Not very sharp, with a lot of noise

However, the traditional license plate recognition method requires a series of complex processes such as license plate positioning, correction, segmentation, and recognition for grain storage vehicles. There is a strong dependence between each step, and there will be error accumulation. lead to large errors in the final recognition results

Images