Object recognition networks in Iran

Amir hossein Shakerin1

IOI Code : AADK-ABACCB
Publication : 3rd International Congress on Science, engineering & New Technologies(secongress.com/3rd)
Abstract :
    Any traffic sign recognition is an important part of advanced driver assistance systems and autonomous driving systems. As the first key step of traffic sign recognition, traffic sign recognition is required for various reasons such as their diversity, size Very small signs, complex driving scenes, large image dimensions, and occlusion by other objects are a challenging problem. In fact, most object detection networks, including single-stage ones like the YOLO family and two-stage ones like the RCNN family, are unable to handle large dynamic objects. They can identify objects such as people, cars, motorbikes, etc. in the scene well, but they still have difficulty in identifying small static objects such as various types of traffic signs and traffic lights. Among traffic signs, traffic light detection is difficult due to the nature of the detection.Its role in moving or stopping vehicles is more essential than other signs. In recent years, a large number of algorithms for traffic sign recognition based on machine vision and pattern recognition have been introduced. But most of them have only limited themselves to sign recognition.And they have not addressed their classification. Therefore, in this research, a two-phase deep model has been introduced for traffic light detection and then color classification for autonomous vehicles to make movement decisions.

Keywords : detection network driving scenes traffic lights