The measurement principle of lateral deviation According to the imaging relationship of the camera, this paper presents a visual measurement method of lateral deviation when the intelligent vehicle is running on a relatively flat road. The solid black line in the figure indicates the road center mark line for guiding the vehicle painted on the road, that is, the vehicle guide line. The world coordinate system OwXwYwZw is fixedly connected to the vehicle, that is, it is relatively stationary with the vehicle, and its XwYw plane is located on the underside of the vehicle and coincides with the road surface. The Q point is a point on the road surface that is fixedly connected to the vehicle and is located at the position L in front of the vehicle. It is a reference point for measuring the lateral deviation l of the vehicle with respect to the vehicle guide line. According to the needs of the study, the value of L is not large, and within a few meters, that is, within the near field of vision. In order to adapt to the near-field measurement requirements, the camera's shooting position and angle are adjusted so that the camera mainly captures the road surface in the near-field of vision, that is, the intersection point between the Zc axis of the camera coordinate system and the ground is located near the Q point.

Parametric calibration of on-board cameras needs to be performed in an outdoor road environment. In this regard, the calibration method for on-board cameras should be fast, flexible, simple, highly accurate, and robust. The calibration methods for on-board cameras are studied in some intelligent vehicle visual navigation documents, such as calibration by drawing a regular grid on the road surface, calibration using known road width and other road information, and linear least squares calibration. The first two methods are constrained by the road surface and have great limitations in practical applications, and the accuracy is not high. The latter method has slow convergence, poor stability, and limited accuracy.

The calibration method of the camera can be mainly divided into linear method, direct nonlinear solution method and nonlinear step method. In order to adapt to the outdoor road environment, this paper uses the proposed calibration method. This method is a non-linear step-by-step method. Compared with the traditional calibration method, which requires a high-precision target composed of two or three mutually orthogonal planes, it only requires a common planar target. The target map does not require special equipment to make, but uses an ordinary printer or plotter to draw a square image consisting of a number of equally spaced black squares, which can then be placed on a flat surface. In the calibration process, only a few camera images with different orientations are taken with the camera, and the camera or the target can be placed arbitrarily according to needs, and it is not necessary to know the displacement or the angle of the change, which is very convenient and practical. This is also unmatched by other calibration methods.

Before the target image is calibrated, it is first necessary to determine the world coordinates and image coordinates of the 64 corner points. For the sake of simplicity, the world coordinate system is directly established on the target, the origin is located at the vertex of the square in the bottom left corner of the target map, and the Xw and Yw axes are parallel to the two sides of the square respectively. The Zw axis is determined according to the right-hand rule. The coordinates of the Zw axes of the 64 corner points are all 0, and the Xw and Yw axis coordinates can also be easily determined. In order to determine the image coordinates of the 64 corners of the target image, the Harris corner extraction algorithm with high stability and reliability is used in this paper. The Harris extraction operator can accurately extract corner points in the case of image rotation, grayscale changes, and noise interference. Using the world coordinates and image coordinates of all the corner points that have been determined, the camera can be calibrated.

Lateral deviation measurement test experiments show that for a smart vehicle on a moderately sloped road, the lateral deviation measurement method proposed in this paper can achieve centimeter-level measurement accuracy in the near field, and the measurement frequency depends on the white mark. The speed of the line was tested and the test speed reached 25 Hz. The measurement results are given under different deviations l and the vehicle reaches a stable tracking black line. The deviation distances l are set to -30, 0, and 30 cm respectively. The statistical results are shown in Table 1. Of course, the above measurement results are indispensable There is a subjective error of the driver. Conclusion Based on the imaging relationship of the camera, this paper proposes a visual measurement method for the lateral deviation of the intelligent vehicle for near-field conditions, and studies a vehicle-mounted camera calibration method that is easy to use and suitable for outdoor road environments. Experiments show that the lateral deviation measurement method proposed in this paper can achieve centimeter-level real-time measurement accuracy for near-field visibility on a moderately graded road.

Disperse Stirring Reaction Equipment

High Shear Mixer,Mobile Lifting Type Mixer,Automatic Lifting Bin Mixer,Automatic Lifting Mixer

Wuxi Mingyan Equipment Co., Ltd , https://www.wxmygroup.com