CN-116879571-B - Vehicle speed detection method based on epipolar image

Abstract

The invention relates to a vehicle speed detection method based on an outer polar image, and belongs to the field of automatic driving. The method comprises the steps of S1, periodically capturing images of roads in front of a vehicle, acquiring space-time distance images, S2, dividing a curve cluster surface of the acquired space-time distance images into a plurality of parts, S3, acquiring curves corresponding to the parts, S4, obtaining speed curves of the parts through analysis and differentiation of the curves, S5, generating a full-interval speed curve, and calculating the speed of the vehicle according to the displacement of characteristic points. Based on the principle of visual kinematics, the speed of the vehicle is calculated by utilizing images acquired by the vehicle-mounted camera and the distance sensor and observing the motion condition of the characteristic points in the images. Compared with the existing speed detection method based on the physical sensor, the method has lower cost, but has simple principle and higher precision. The method can be used independently, and can also be combined with data of other sensors, so that the detection precision and stability are improved.

Inventors

• DENG TIANMIN
• CHEN YUETIAN
• DENG JIE
• JIANG ZUOBO
• YANG LING
• XIE PENGFEI
• YU YANG
• PENG LIDAN
• LAN YIFAN

Assignees

• 重庆交通大学
• 重庆市佰强科技有限公司

Dates

Publication Date

20260508

Application Date

20230711

Claims (1)

1. 1. A vehicle speed detection method based on an epipolar image is characterized by comprising the following steps: s1, loading a distance sensor for repeated route scanning on a vehicle, and periodically capturing images of roads in front of the vehicle to obtain space-time distance images; S2, dividing the curve cluster surface of the obtained space-time interval image into a plurality of parts; S3, obtaining curves corresponding to all parts, wherein the curves represent the condition that all target objects are changed along with time in the visual range of the sensor, and represent the position and speed change of the vehicle of the device sensor; S4, expressing the corresponding curves of the parts in the S3 by using analytical expressions to show the position change of each local short section; S5, generating a full-interval speed curve, calculating the vehicle speed according to the displacement of the characteristic points, carrying out regression processing on the speed curve obtained from each part, which is the product of the part with the local interval, generating a speed curve which is connected with each speed curve and spans the full interval, wherein the reliability is higher when the speed curve is closer to the middle part; in S1, the vehicle with the distance sensor is driven horizontally to obtain a space image, a side composed of consecutive corresponding points in the space image is drawn, depth information is included, and the constituent points of the outer pole face distance image EPDI are subjected to side detection processing using the depth information: (1) in the formula, Is the slope of the edge; Representing a horizontal movement distance of the image distance sensor; Representing the distance that the feature moves in the image plane, i.e. the amount of change in the image per scan frame; is the interval of arranging scanning lines, standard unit ; Representing scan time consuming; is the frame frequency of the line scanning sensor, the standard unit is Or (b) ; Is the speed of movement of the sensor; depth D and parallax The correlation is performed by the following method by using the moving distance: (2) (3) in the formula, Representing parallax variation amounts at different positions; And Representing the parallax of the image distance sensor at position 1 and position 2, respectively; Representing a horizontal movement distance of the image distance sensor; representing the horizontal distance of the reference point P from the image sensor at position 1; representing a perpendicular distance of the camera plane from the image plane; The speed obtained by combining the formula (1) and the formula (3) is: (4) In the step S5, since a plurality of cluster planes are included in the acquired spatio-temporal distance image data, the cluster planes are used Representation, if From part of the first Frame start to If present in the frame-up range, the first Frame speed estimation value The confidence level of (2) is defined in the following way: (5) (6) (7) in the formula, Represent the first A frame preceding the frame; Represent the first The frame following the frame; Represent the first Average value of frame before and after frame; Represent the first A scaling factor of a difference between a subsequent frame and a previous frame of the frame; Will follow The resulting velocity profile is seen as In other words, the estimated speed profile across the whole interval Expressed in the following manner: (8) (9) in the formula, Is the first A set of frames in a cluster plane; Is the first The sum of the confidence levels of the velocity estimation values of each frame in the cluster planes; The speed profiles of each part corresponding to the respective reliability are connected.

Description

Vehicle speed detection method based on epipolar image Technical Field The invention belongs to the field of automatic driving, and relates to a vehicle speed detection method based on an outer polar surface image. Background In real world measurement or reconstruction with sensors, speed determination of the sensor itself is an important issue. The speed of the macro location may be determined using external sensing means such as GPS and inertial (gyroscopic) sensors, and the speed of the micro location may be determined using a speed or acceleration sensor, for example, a vehicle may use a vehicle speed pulse sensor. However, the GPS has very limited use conditions, such as high-rise shielding and reflection influence, tunnel satellite loss and the like, and has fewer means for connecting an intermediate sensor of macro and micro data and improving the macro and micro data. The vehicle speed detection method based on the outer polar image is relatively simple in algorithm realization, can effectively avoid leakage of sensitive information such as user position information and the like by the GPS, and can be not influenced by signals in remote areas. Disclosure of Invention In view of the above, an object of the present invention is to provide a vehicle speed detection method based on an epipolar image. In order to achieve the above purpose, the present invention provides the following technical solutions: a vehicle speed detection method based on an epipolar image, the method comprising: s1, loading a distance sensor for repeated route scanning on a vehicle, and periodically capturing images of roads in front of the vehicle to obtain space-time distance images; S2, dividing the curve cluster surface of the obtained space-time interval image into a plurality of parts; S3, obtaining curves corresponding to all parts, wherein the curves represent the condition that all target objects are changed along with time in the visual range of the sensor, and represent the position and speed change of the vehicle of the device sensor; S4, expressing the corresponding curves of the parts in the S3 by using analytical expressions to show the position change of each local short section; s5, generating a full-interval speed curve, calculating the speed of the vehicle according to the displacement of the characteristic points, carrying out regression processing on the speed curves obtained from all the parts, which are products with the part having the interval part, generating speed curves which are connected with all the speed curves and span the full interval, wherein the reliability is higher when the speed curves are closer to the middle part, and the reliability is higher when a plurality of speed curves are obtained in a certain specific interval. Optionally, in the step S1, the vehicle with the distance sensor is driven horizontally to obtain a spatial image, a side formed by consecutive corresponding points in the right image is drawn, the side includes depth information, and the forming point of the outer pole face distance image EPDI uses the depth information to perform the detection processing of the side: Wherein m is the slope of the edge, deltax is the horizontal movement distance of the image from the sensor, deltay is the movement distance of the feature in the image plane, namely the variation of each scanned frame image, k is the interval of the arranged scanning lines, the standard unit m -1, deltat is the scanning time consuming time, F 0 is the frame frequency of the line scanning sensor, the standard unit is Hz or s -1, and V is the movement speed of the sensor; The depth D and the parallax u are related by the following method by moving distance: Wherein DeltaU represents parallax variation at different positions, U 1 and U 2 represent parallax at position 1 and position 2 of the image distance sensor, respectively, deltaX represents horizontal movement distance of the image distance sensor, X represents horizontal distance between the reference point P and the image sensor at position 1, and h represents vertical distance between the camera plane and the image plane; The speed obtained by combining the formula (1) and the formula (3) is: Optionally, in the step S5, since a plurality of cluster planes exist in the acquired spatio-temporal distance image data, the cluster planes are denoted by S 1,S2,...,Sn, and if the portion S n exists in a range from the kth n frame to the k' n frame, the credibility of the speed estimate R n (k) of the kth frame is defined in the following manner: 6σ=k'n-kn (7) wherein k n represents a frame preceding the kth frame, k' n represents a frame following the kth frame, μ represents an average value of frames preceding and following the kth frame, σ represents a scaling factor of a difference between the frame following the kth frame and the frame preceding the kth frame; regarding the velocity profile obtained from S n as V n (k), the estimated velocity profile V (k)