A vehicle lower boundary detection method based on multi-sensor fusion

Abstract

The invention provides a vehicle lower boundary detection method based on multi-sensor fusion. Through measuring information of a millimeter-wave radar and a camera, space alignment points are obtained, an area of interest containing a target vehicle is selected in a camera image according to space alignment point information, by determining k peak values of the area of interest, k-1 threshold values of the area of interest are determined, a minimum threshold value is obtained, and accordingly a vehicle lower boundary shadow area corresponding to a gray level area marked out by the minimum threshold value, finally, a vehicle lower boundary shadow line is obtained in the shadow area, and vehicle lower boundary detection is achieved. By using a certain searching strategy, the target area of interest is determined, the target area of interest can contain the target vehicle, the size of the area can be proper, following computing is convenient, the k-1 threshold values are determined through a particle swarm optimization algorithm, operation steps are simplified, operation speed is improved, and meanwhile area dividing accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of unmanned vehicle environment perception, in particular to a vehicle detection method based on multi-sensor fusion. Background technique [0002] Unmanned vehicles, also known as outdoor intelligent mobile robots, are highly intelligent devices that integrate environmental perception, dynamic decision-making and planning, behavior control and execution. It is inseparable from multi-sensor information fusion technology. Environmental perception technology is the prerequisite for autonomous driving of unmanned vehicles. It is the most basic, critical and challenging subject. Its main function is to determine the target position of vehicles and the safety of unmanned vehicles in dynamic scenarios. driving area. Therefore, vehicle detection is an important research content of unmanned vehicle environment perception, and it is of great significance to the autonomous navigation of unmanned vehicles. According...

AI Technical Summary

Problems solved by technology

According to the types of sensors, vehicle detection methods are mainly divided into vision-based detection methods and radar-based detection methods. Among them, the vision-based detection method is currently the most widely studied, which can obtain the size and outline information of the target, but cannot obtain the position of the target. and relative velocity information

Radar-based detection methods can obtain accurate target position and relative velocity information, but cannot identify the shape and size of the target

Among them, the most critical part of the vision-based detection method is the detection of the lower boundary of the vehicle, which directly affects the accuracy of vehicle detection. If the detection of the lower boundary of the vehicle is inaccurate, other non-vehicle targets will affect the effectiveness of the detection algorithm in the subsequent vehicle detection process. sex

Since the grayscale of the shadow area at the bottom of the vehicle is darker than that of the asphalt pavement, the current lower boundary detection of the vehicle mainly obtains the shadow of the bottom of the vehicle through image segmentation methods (such as the maximum between-class variance method and the area statistics method), and then extracts the shadow However, due to the complexity of the natural environment, the existing segmentation methods cannot adapt to changes in the scene, which in turn affects the detection of the lower boundary of the vehicle.

Images