CN-122017810-A - Vehicle distance measurement calibration method based on laser radar and binocular vision fusion

CN122017810ACN 122017810 ACN122017810 ACN 122017810ACN-122017810-A

Abstract

The invention discloses a vehicle distance measurement and calibration method based on laser radar and binocular vision fusion, which relates to the technical field of measurement and calibration, is used for solving the problem that the accuracy of a vehicle distance calibration result of a laser radar used in the process of calibrating a highway vehicle distance snapshot system is reduced under the condition of rain, snow and slippery or ponding road surface, the method comprises the steps of constructing a specular reflection state code, adaptively triggering a self-calibration mechanism according to the specular reflection state code, respectively carrying out quantitative analysis on discrete characteristics of a laser radar distance time sequence and degradation trend of binocular vision parallax connectivity, judging a dominant source of distance measurement abnormality through symbol analysis, and selecting an offset calibration or fusion correction strategy according to the dominant source, so that reliability and engineering adaptability of distance measurement under a wet road surface condition can be improved under the condition of not depending on an additional sensor.

Inventors

LUO TENGKE
ZHAO PENG
YE FUYU
WU JIUNIU
Wei Chenkai
JIANG WEIDONG
XU ZHAOXIN

Assignees

甘肃省计量研究院

Dates

Publication Date: 20260512
Application Date: 20260310

Claims (10)

1. A vehicle distance measurement and calibration method based on laser radar and binocular vision fusion is characterized by comprising the following steps: Step S1, monitoring point cloud data of a laser radar carried by a test vehicle when the test vehicle normally runs, carrying out partition processing on the point cloud data, screening and marking Lei Dadian clouds, counting the number of the point clouds of the marked radar point clouds, and detecting echo intensity data of the marked radar point clouds; S2, calculating a point cloud echo coefficient according to echo intensity data, analyzing the specular reflection state of the radar point cloud by combining the number of the radar point clouds, judging whether to trigger a self-calibration mechanism or not based on the specular reflection state, and setting a ranging period after triggering the self-calibration mechanism; step S3, scanning radar distance of the test vehicle in a distance measurement period, performing traveling evaluation, generating a distance discrete feature according to an evaluation result, storing the distance discrete feature in a self-calibration database, acquiring a distance target area of the test vehicle by binocular vision, identifying parallax pixels in the distance target area and calculating pixel communication length; And S4, generating parallax degradation trend of the test vehicle based on the pixel communication length, executing symbol analysis processing by combining with the vehicle distance discrete features in the self-calibration database, and selecting to perform fusion correction or offset calibration on the radar vehicle distance of the test vehicle according to the analysis result.
2. The vehicle distance measurement calibration method based on laser radar and binocular vision fusion according to claim 1, wherein the method comprises the following steps: in step S1, when a test vehicle runs normally, starting a laser radar to scan the environment space of the test vehicle and collecting point cloud data output by the laser radar in the current scanning period; The point cloud data consists of a plurality of point cloud sampling points, the point cloud sampling points are characterized by three-dimensional space coordinates under a laser radar coordinate system, and the three-dimensional space coordinates consist of three mutually orthogonal coordinate components, including a longitudinal distance component, a transverse offset component and a height component; Defining a point cloud sampling point set formed by three-dimensional space coordinates as radar point clouds; After the radar point cloud is obtained, carrying out partition processing on Lei Dadian clouds according to a preset space partition rule, wherein the space partition rule divides point cloud sampling points with height components lower than a preset height threshold into a low-height analysis area, and otherwise, dividing the point cloud sampling points into a non-analysis area.
3. The vehicle distance measurement calibration method based on laser radar and binocular vision fusion according to claim 2, wherein the method comprises the following steps: in step S1, point cloud sampling points located in the low-altitude analysis area are marked and combined into a marked radar point cloud, and point cloud sampling points which do not fall into the low-altitude analysis area do not participate in subsequent statistics; Counting the point cloud sampling points of the point clouds of the mark radar to obtain the number of the point clouds of the mark radar; and detecting the echo intensity of the cloud sampling point of the mark radar point Yun Nadian, wherein the echo intensity is synchronously output by the laser radar in the sampling process, and the return energy of the laser signal is represented.
4. A method for calibrating vehicle distance measurement based on fusion of laser radar and binocular vision according to claim 3, wherein the method comprises the following steps: In step S2, calculating the sum of echo intensities of all point cloud sampling points in the point cloud of the mark radar to obtain the mark echo intensity; the method comprises the steps of performing descending order sorting on echo intensity data corresponding to each point cloud sampling point in a marked radar point cloud, and extracting high-energy sampling points with echo intensities in a preset high-intensity interval; Counting the sum of echo intensities of the high-energy sampling points to obtain high-energy echo intensity, and calculating the duty ratio of the high-energy echo intensity in the marked echo intensity to obtain a point cloud echo coefficient of the marked radar point cloud; and calculating the ratio of the number of the point clouds of the marked radar point clouds to the total number of the radar point Yun Nadian cloud sampling points acquired by the laser radar in the current scanning period to obtain the low-height number duty ratio.
5. The method for measuring and calibrating the vehicle distance based on the fusion of the laser radar and the binocular vision according to claim 4, wherein the method comprises the following steps: in step S2, comparing the point cloud echo coefficient with a preset multi-level echo coefficient threshold value, dividing a plurality of echo level states and representing the echo level states by digital codes; Comparing the low height quantity ratio with a preset multilevel quantity ratio threshold value, dividing a plurality of quantity level states and representing the states by letter codes; the echo grade states and the quantity grade states are combined and encoded to generate a specular reflection state code, and the specular reflection state code is formed by combining a numerical code and a letter code; Comparing the specular reflection state code with a preset specular reflection trigger threshold, and when the specular reflection state code exceeds the preset specular reflection trigger threshold, judging Lei Dadian the specular reflection state of the cloud to be an abnormal reflection state, and triggering a self-calibration mechanism at the moment; When the specular reflection state code does not exceed a preset specular reflection triggering threshold, determining Lei Dadian that the specular reflection state of the cloud is a non-abnormal reflection state, and not triggering a self-calibration mechanism at the moment; After the self-calibration mechanism is triggered, a ranging period is further set, wherein the ranging period is a time window for limiting the time range of the radar distance to be acquired subsequently.
6. The vehicle distance measurement calibration method based on laser radar and binocular vision fusion according to claim 1, wherein the method comprises the following steps: In step S3, in the ranging period, radar distances of test vehicles are obtained through a vehicle-mounted laser radar ranging interface, and each radar distance in the ranging period is formed into a distance time sequence according to a time sequence; Carrying out fluctuation evaluation on the vehicle distance time sequence, specifically, selecting the median in the vehicle distance time sequence as the vehicle distance median, respectively calculating the absolute value of the difference between the vehicle distance of each radar and the vehicle distance median to obtain vehicle distance deviation, carrying out sorting treatment on each vehicle distance deviation, and selecting the median as the vehicle distance deviation median; Taking the ratio of the median value of the vehicle distance deviation to the median value of the vehicle distance as a vehicle distance discrete index; Taking the product result of the vehicle distance discrete index and the preset adjustment discrete coefficient as the vehicle distance discrete characteristic.
7. The vehicle distance measurement calibration method based on laser radar and binocular vision fusion according to claim 1, wherein the method comprises the following steps: In step S3, in the ranging period, synchronously acquiring a road area in the front driving direction of the test vehicle by using a binocular camera in binocular vision, respectively obtaining a left-eye image and a right-eye image, and performing front vehicle target detection processing on the left-eye image based on an image target recognition unit; the front vehicle target detection process outputs a vehicle distance target area of a target vehicle in the left-eye image; Selecting pixel points in the left-eye image pixel by pixel in a vehicle distance target area by taking the left-eye image as a reference image, and obtaining right-eye pixel points corresponding to the left-eye pixel points based on the calibration relation of the binocular camera; and obtaining the parallax value of the pixel point based on the coordinate difference value of the left-eye pixel point and the corresponding right-eye pixel point in the image coordinates, and mapping the parallax value into a parallax result image.
8. The method for measuring and calibrating the vehicle distance based on the fusion of the laser radar and the binocular vision according to claim 7, wherein the method comprises the following steps: In step S3, in the parallax result image, if the parallax value of the pixel point is greater than or equal to the preset lower limit parallax threshold, marking the pixel point as an effective parallax pixel, otherwise, not marking the pixel point; Combining the marked pixel points into an effective parallax pixel distribution map, carrying out progressive scanning on the effective parallax pixel distribution map, and counting the lengths of adjacent effective parallax pixels in each row of pixels, wherein the maximum value of the lengths is used as a continuous pixel span; And selecting the maximum continuous pixel span from all rows of pixels in the vehicle distance target area as the pixel communication length.
9. The vehicle distance measurement calibration method based on laser radar and binocular vision fusion according to claim 8, wherein the method comprises the following steps: In step S4, a plurality of pixel communication lengths in a ranging period are called, the pixel communication lengths are arranged according to a time sequence, a communication difference value is obtained by differencing adjacent pixel communication lengths, the number of the communication difference values smaller than zero is counted, a degradation count is obtained, and a parallax degradation trend is obtained by utilizing the degradation count; and calling the vehicle distance discrete features in the self-calibration database, and respectively presetting a parallax degradation threshold value and a preset vehicle distance discrete threshold value.
10. The vehicle distance measurement calibration method based on laser radar and binocular vision fusion according to claim 9, wherein the method comprises the following steps: In step S4, symbol analysis processing is performed based on the radar discrete coefficient and the parallax discrete coefficient, specifically: Taking the ratio of the vehicle distance discrete feature to a preset vehicle distance discrete threshold value as a radar discrete coefficient, and taking the ratio of the parallax degradation trend to the preset parallax degradation threshold value as a parallax discrete coefficient; The radar discrete coefficient and the parallax discrete coefficient are subjected to difference to obtain a deviation dominant difference value, and the deviation dominant difference value is compared with a preset dominant difference threshold value; If the absolute value of the deviation dominant difference value is larger than a preset dominant difference threshold value and the deviation dominant difference value is positive, marking the deviation dominant abnormal symbol, and executing offset calibration on the radar distance of the test vehicle; if the absolute value of the deviation dominant difference value is larger than a preset dominant difference threshold value and the deviation dominant difference value is negative, marking as a visual dominant abnormal sign, and executing fusion correction on the radar distance of the test vehicle; otherwise, no marking is performed.

Description

Vehicle distance measurement calibration method based on laser radar and binocular vision fusion Technical Field The invention relates to the technical field of measurement and calibration, in particular to a vehicle distance measurement and calibration method based on fusion of a laser radar and binocular vision. Background With the continuous increase of the traffic flow of the expressway, the illegal snapshot system based on the vehicle distance judgment is widely applied to automatic evidence obtaining scenes of traffic illegal behaviors such as keeping the safe vehicle distance and approaching the vehicle too close according to regulations. The existing highway vehicle distance snapshot system generally adopts a traffic violation monitoring system mode based on a speed measuring radar, an advanced imaging technology and computer vision, continuously measures the space distances between front vehicles and rear vehicles in the same lane, and judges whether the vehicles have a behavior continuously lower than a safe vehicle distance threshold or not by combining a time dimension. Because the system directly serves for traffic law enforcement evidence collection, the accuracy and the reliability of the vehicle distance measurement result are directly related to the fairness and the legality of law enforcement, the system vehicle distance measurement result is required to be calibrated with higher accuracy, and the stability and the consistency of the calibration value are required to be maintained under complex weather and road surface conditions. Under the road surface condition that rain and snow slide or water exists, road surface reflection characteristic changes from diffuse reflection to specular reflection, and the test vehicle-mounted laser radar for calibration is easy to produce abnormal enhanced echo signals and low-height false point clouds when measuring and scanning, so that the calibration device misjudges road surface reflection or vehicle chassis reflection as an effective target, and therefore systematic offset or fluctuation increase occurs in a vehicle distance calibration result, and the stability of vehicle distance calibration is affected. The prior art has the following defects: at present, under the conditions of rain, snow and wet skid, ponding or high reflectivity road surface, when a laser radar used for calibrating a highway vehicle distance snapshot system scans a target in front of a lane where a test vehicle is located, abnormal echoes and low-height false point clouds are easily generated due to the influence of specular reflection of the road surface, so that systematic deviation or fluctuation of a vehicle distance calibration result occurs, meanwhile, binocular vision is easy to generate parallax degradation in a ponding or low-texture water film area, the stability of vehicle distance calculation is reduced, a self-calibration mechanism aiming at the working conditions is lacked, the reliability of the vehicle distance calibration result is difficult to ensure, and the accuracy of the calibration result is reduced under the conditions of rain, snow and wet skid or ponding road surface. Disclosure of Invention In order to overcome the above-mentioned drawbacks of the prior art, an embodiment of the present invention provides a vehicle distance measurement calibration method based on fusion of a laser radar and binocular vision, which solves the problems set forth in the above-mentioned background art by applying a joint analysis mechanism of adaptive recognition of a road surface reflection state and a multi-source ranging stability difference. In order to achieve the above purpose, the invention provides a vehicle distance measurement and calibration method based on fusion of a laser radar and binocular vision, which comprises the following steps: step S1, monitoring point cloud data of a vehicle-mounted laser radar of a test vehicle when the test vehicle normally runs, carrying out partition processing on the point cloud data, screening and marking Lei Dadian clouds, counting the number of the point clouds of the marked radar point clouds, and detecting echo intensity data of the marked radar point clouds; S2, calculating a point cloud echo coefficient according to echo intensity data, analyzing the specular reflection state of the radar point cloud by combining the number of the radar point clouds, judging whether to trigger a self-calibration mechanism or not based on the specular reflection state, and setting a ranging period after triggering the self-calibration mechanism; step S3, scanning radar distance of the test vehicle in a distance measurement period, performing traveling evaluation, generating a distance discrete feature according to an evaluation result, storing the distance discrete feature in a self-calibration database, acquiring a distance target area of the test vehicle by binocular vision, identifying parallax pixels in the distance target area and c