Improved SSD based pedestrian detection algorithm for forklift active warning system

Abstract

Forklifts are the common transport vehicles used in the industrial production widely. The accidents related to forklifts such as collisions between people and vehicles, speeding, and overturning often occur. With the rise of deep learning technology, target detection algorithms based on deep learning have also been applied to the detection of vehicles and pedestrians in traffic scenarios. This paper takes pedestrian detection as a research angle, and studies how to make the algorithm not only have a high detection speed, but also ensure the detection accuracy in practical scene applications. In addition, due to the fact that the algorithm is used in practical application scenarios, the road environment is often complicated and the vehicles are dense, and it may also encounter some extreme weather factors such as night, rain, cloudy, etc. These complex traffic conditions will increase Pedestrian detection difficulty. In order to solve the problem of low detection accuracy of pedestrian target detection in complex road conditions, this paper improves the existing target detection algorithm and designs a pedestrian target detection algorithm based on deep learning, so as to accurately detect pedestrians in more complex road scenes The goal is to reduce traffic accidents and ensure driving safety. The validity of the clustering-based anchor parameter setting method is verified by experiments.