CN103077639B - Curve driving detection system and detection method thereof - Google Patents

Abstract

The invention discloses a curve driving detection system and a detection method thereof. The system includes a sign board, a camera, an image analyzer, a transmission network, a server and a client; The physical size of the coordinate system is calibrated; the image analyzer continuously receives the monitoring images captured by the camera and recognizes the signboard on the training vehicle, and calculates the actual position, attitude, contour range and contour to the marking line of the vehicle through the calibrated coordinate system The shortest distance, and capture pictures when the vehicle outline crosses the marking line; the server receives the measurement data, pictures and videos from the image analyzer through the transmission network, and calculates the training score, and the driver and the coach pass the back-end client to the information Inquire or read video recordings, so as to achieve effective training for drivers and improve the learning efficiency of driving trainers.

Description

Curved driving detection system and its detection method

technical field

The invention relates to a curve driving detection system and a detection method thereof.

Background technique

As a motor vehicle driver, curve driving is one of the driving skills that must be mastered. However, in traditional driver training, trainees generally use fixed reference objects for training, which leads to the need to re-cultivate position sense and turning skills after the actual road. Although it is possible to collect the position of the vehicle by installing a certain number of sensors on the road, such as the permanent magnet used in Chinese utility model 201120277850.8, or the infrared rays, lasers, and ultrasonics commonly used in engineering distance measurement, such sensors can only be used at individual fixed points. On-site perception, complex construction and maintenance, and high overall cost, and the linkage of cameras is still required to achieve evidence collection.

Whether the students are learning or the coaches are teaching, judgments are mainly based on visual information. Therefore, the detection method of the entire dynamic process of curve driving based on machine vision is also the most acceptable.

Although China Utility Model No. 201120489062.5 introduces a video detection device for linear margins used in driver tests, its core is: "The image processing unit maps the obtained images to the HSV and LAB color spaces respectively, and thresholds in each space The pixel comparison unit uses the probability principle to make statistics on the input pixels, and obtains the linear element that best matches the characteristics of the right line, and calculates the right margin of the linear element. As we all know, this method is difficult to be operable in the real environment, because there are many linear pixels in the vehicle itself, such as windows, ceilings, etc., and there are often a large number of linear pixels in the background environment, such as curbs, In this case, it is difficult to extract the "car" from the background environment at the pixel level, which eventually leads to inaccurate detection.

Contents of the invention

The purpose of the present invention is to provide a curve driving detection system and detection method thereof, which can realize quantitative analysis of the training vehicle during curve driving, and is beneficial to real-time monitoring and tracking of training results.

In order to solve these problems in the prior art, the technical solution provided by the invention is:

A curve driving detection system, which includes a front-end system installed near the curve driving position and a back-end system deployed in the management center, the front-end system includes multiple main cameras, image analyzers and signs placed on the training car , the back-end system includes a server and a client. The front-end system is connected to the back-end system through a transmission network. The monitoring screen covers the main camera in the curved driving position area to record the driving conditions of the training vehicle at the warehouse. The image analysis connected to the main camera The instrument continuously receives the monitoring images captured by the main camera and compresses them into monitoring images. At the same time, the image analyzer monitors the running posture of the vehicle. The image analyzer sends the monitoring results and monitoring images to the server of the back-end system. The client can call to view the Monitoring results and monitoring images.

As an optimization, the multiple main cameras are fixed on the pole above the curved driving position, each of which can overlook a section of the curved area of the curved curve.

As an optimization, there are auxiliary cameras on both sides of each main camera to assist in supplementary capture of the running track of the training vehicle in the warehouse.

As an optimization, the signage is arranged on the roof, the engine roof or the trunk lid, and the signage on the signboard is a number of marking points, and the marking points are arranged as a specific line segment with a fixed angle and a unique intersection point.

As an optimization, a supplementary light is provided next to the main camera and the auxiliary camera, and the synchronization signal of the supplementary light is given by the main camera or the auxiliary camera, which is used for strobe supplementary light when the light is too dark.

The present invention also provides a method for detecting curve driving, the method involves a system including a front-end system installed near the curve driving position and a back-end system deployed in the management center, the front-end system includes multiple main cameras, auxiliary cameras, The image analyzer and the sign board placed on the training vehicle, the back-end system includes the server and the client, the front-end system is connected to the back-end system through the transmission network, and the main camera covering the storage area with the monitoring screen is used to record the training vehicle at the storage area driving conditions, the image analyzer is connected with the main camera and the auxiliary camera, and the specific detection method includes the following steps:

Step S1: By directly manually measuring the length and width of the storage location, and running the camera calibration program, the mutual conversion between the image coordinate system and the world coordinate system is realized;

Step S2: The image analyzer continuously receives the monitoring images captured by the main camera, compresses them into video recordings, and runs the sign recognition program at the same time to obtain the position and direction of the sign;

Step S3: The image analyzer runs the vehicle position and attitude detection program according to the camera calibration results and the sign recognition results to obtain the position and attitude of the vehicle; combined with the known vehicle size, then determine the range of the vehicle outline and the outer edge of the outline to the lane boundary The shortest distance, and trigger the main camera or auxiliary camera to capture pictures when the vehicle presses the line; when the sign recognized by the image analyzer disappears from a specific area of the image, it is determined that the vehicle has left the corresponding main camera coverage area, and the image analyzer will divide The measurement information such as the time spent in the segment, together with the captured pictures and video recordings are sent to the server;

Step S4: After receiving the measurement information sent by each image analyzer, the server runs the scoring program to obtain the comprehensive score of this curve driving;

Step S5: The trainee or coach accesses the server through the client, browses, downloads and prints scores related to curve driving, snapped pictures and video recordings.

For the above detection method, the inventor also has a further optimized implementation.

As an optimization, in step S3, the main camera 1 or the auxiliary camera 2 is triggered to capture pictures and send them to the image analyzer when the vehicle presses the line or the distance is less than the set threshold during the parking process.

As an optimization, in step S1, when performing camera calibration, (the world coordinate system is the coordinate system of the real world where the vehicle is located, consisting of three coordinate axes: X axis, Y axis, and Z axis, and the image coordinate system is the image captured by the camera The planar image is composed of two coordinate axes: U axis and V axis), in step S1, when performing camera calibration, the steps of the camera calibration program are:

The camera captures an image, takes any pixel as the origin of the image coordinate system, and manually marks the image coordinates of the end points of the four lane boundary lines of this section of curved road section;

Mark the physical position point where the origin of the image coordinate system is located, and use this as the origin of the world coordinate system to measure the world coordinates of the four end points of the lane boundary line in the actual site of the curved road section;

Substitute the world coordinates and pixel coordinates of the vertices of the four storage locations into the image-world coordinate transformation linear equations to solve the image-world coordinate transformation matrix;

Substituting the coordinate value of a certain point in any world coordinate system into the image-world coordinate transformation linear equation set known by the image-world coordinate transformation matrix, the corresponding image coordinates of the point can be calculated;

On the premise that the world coordinate height is known, any image coordinate can be substituted into the known image-world coordinate transformation linear equations of the image-world coordinate transformation matrix, and the world coordinate corresponding to the point can be calculated; so far, the camera calibration is completed .

As an optimization, the process of sign recognition in step S2 is as follows:

Perform threshold segmentation on the image to obtain the segmented image;

Extract the connected domain from the image after the threshold segmentation, judge whether the length and width of the connected domain meet the length and width setting values of the signboard marking point graphics, and if so, mark the connected domain;

Determine whether the marked connected domain can be connected into a shape consistent with the actual signboard, including the number of marked points and the angle formed by the line connecting the marked points. If yes, the connected area is regarded as a marked point detection graph, otherwise the connection is discarded domain; repeat this step until all connected domains are traversed;

Use the mark point detection graphic as a reference point to identify and identify other additional letter and number areas, and complete the identification of the entire signboard.

In the detection method, the step flow for vehicle position and posture detection in step S3 is:

Substituting the image coordinates of the mark point detection graphics obtained in the sign recognition program into the image-world coordinate transformation linear equations of the camera calibration program, the world coordinates of the mark point detection graphics can be obtained;

In the world coordinate system, the mark point detection graphics are connected into line segments with intersection points according to the arrangement rules of the original mark point graphics on the signboard;

In the world coordinate system, the coordinates of the intersection point and the included angle of the line segment are calculated. The coordinates of the intersection point are the vehicle position, and the included angle of the line segment is the vehicle attitude.

Compared with the scheme in the prior art, the advantages of the present invention are:

1. The present invention describes a curve detection system and detection method thereof. The system includes a sign board, a camera, an image analyzer, a transmission network, a server and a client; the detection method: first record the camera picture, manually measure the curve, The physical size of the training car, complete the calibration of the coordinate system; the image analyzer continues to receive the monitoring images captured by the camera and recognize the signboard on the training car, and calculate the actual position, attitude, contour range and contour to the vehicle through the calibrated coordinate system. The shortest distance of the marking line, and capture pictures when the vehicle outline crosses the marking line; the server receives the measurement data, pictures and videos from the image analyzer through the transmission network, and calculates the training score, and the driver and trainer pass the back-end client Query information or access video recordings, so as to achieve effective training for drivers and improve the learning efficiency of driver trainers;

2. The present invention is completely based on video technology, does not need to embed other sensors on the site, and does not need to add any on-board electrical equipment, with simple structure and low cost;

3. The present invention is directly based on video technology, and supports picture capture of abnormal situations such as pressure lines, which is convenient for evidence collection.

Description of drawings

The present invention will be further described below in conjunction with accompanying drawing and embodiment:

Fig. 1 is a schematic diagram of the system structure of an embodiment of the present invention;

Fig. 2 is a schematic diagram of a signboard of an embodiment of the present invention;

Fig. 3 is the system overall work flowchart of the embodiment of the present invention;

FIG. 4 is a flowchart of a camera calibration program according to an embodiment of the present invention;

Fig. 5 is a flow chart of a signboard recognition program according to an embodiment of the present invention;

Fig. 6 is a flow chart of vehicle position and attitude detection according to an embodiment of the present invention;

Fig. 7 is a flow chart of the scoring program of the embodiment of the present invention.

Among them: 1. Main camera; 2. Auxiliary camera; 3. Fill light; 4. Image analyzer; 5. Signboard; 6. Transmission network; 7. Server; 8. Client.

Detailed ways

Example:

This embodiment describes a curve driving detection system and its detection method. The system structure is shown in Figure 1. The system consists of a front-end system installed near a curved road and a back-end system deployed in a management center. The front-end system mainly includes: main camera 1, auxiliary camera 2, supplementary light 3, image analyzer 4 and sign board 5 placed on the training vehicle; the back-end system mainly includes: server 7, client 8; The systems are interconnected via a transport network 6 .

The curved road is composed of two 180-degree semicircular curves. In order to obtain a good monitoring effect and save deployment costs, three main cameras 1 are respectively installed in the front, middle and rear sections of the curved road. The main camera 1 in the front section covers The 135-degree area of the first semicircle of the curved curve, the main camera 1 in the middle section covers the remaining 45-degree area of the first semicircle and the 45-degree area of the second semicircle of the curved curve, and the main camera 1 of the rear section covers the remaining 45-degree area of the second semicircle of the curved curve 135 degree area. The main camera 1 is installed on the cross arm of the pole about 4-8 meters above the road, and is on the center line of the road. The installation angle needs to ensure that the lane boundary line of the covered local curved road is continuous in the picture, that is, two There are two intersection points between the lane boundary line and the edge of the image. The main camera 1 adopts industrial-grade wide temperature design, the sensor adopts 5 million CCD, the resolution is greater than 2592x1936, the frame rate is not lower than 8 frames per second, and the built-in Gigabit Ethernet interface is equipped with a variety of fixed-focus lenses from 10mm to 35mm or Zoom wide-angle lens, which can effectively ensure that the monitoring screen covers at least 8m x 6m curve area under the above installation conditions, and ensure that the horizontal pixel width of the signboard is not less than 500 pixels. The auxiliary camera 2 can be installed on the same pole cross arm as the main camera 1. The specific position is 0.2-1 meter outside the boundary line of the left and right lanes. The auxiliary camera adopts an industrial-grade wide temperature design. The sensor uses a 2 million CCD with a resolution greater than 1920x1080. The frame rate is not lower than 8 frames per second, built-in Gigabit Ethernet interface, and equipped with a variety of fixed-focus lenses or zoom lenses from 10mm to 25mm, which can effectively ensure that the wheel pressure line can be effectively monitored under the above installation conditions.

The supplementary light 3 adopts a strobe LED supplementary light, the synchronous signal given by the main camera 1 is called the main supplementary light, and the synchronous signal given by the auxiliary camera is called the auxiliary supplementary light. The power and viewing angle of the fill light 3 depend on the installation location. In order to avoid overexposure of the camera caused by the reflection of the vehicle body, the supplementary light 3 and the camera controlling its synchronization should be installed at a certain distance, generally not less than 0.5 meters. When the installation height of the supplementary light 2 is 6 meters, and the horizontal distance from the center of the parking space is 6 meters, the viewing angle is not less than 40 degrees, the power is not less than 15W, and the installation at a distance of 1 meter from the camera can meet the requirements.

The image analyzer 4 uses an embedded industrial control computer, and its shell is an integrated heat dissipation shell, which does not require a heat dissipation fan, effectively prevents internal dust accumulation, and improves system stability; when connecting a 5 million-pixel CCD camera, it is required to configure the main frequency No less than 2.4GHz CPU, no less than 4GB memory, no less than 32GB hard disk as external storage, Gigabit Ethernet, RS232 and other hardware interfaces, the above configurations ensure that the image analyzer 4 has sufficient computing, storage and communication resources to run various image processing algorithms and applications.

The server 7 can be a tower server, a quad-core processor with a main frequency greater than 2.8GHz and an 8MB cache, and a memory of no less than 4GB, so as to ensure real-time system response when multiple image analyzers 4 are deployed.

The client 8 can be an ordinary PC, which can be equipped with peripheral devices such as a printer, an IC card reader, or a fingerprint collector.

The transmission network 6 uses optical fiber or 3G wireless communication network when connecting the front-end and back-end systems, while the local front-end and back-end systems mostly use Gigabit Ethernet based on twisted pairs.

The main camera 1 , the auxiliary camera 2 , the image analyzer 4 , the server 7 , and the client 8 realize data exchange through the transmission network 6 .

As shown in Figure 2, signboard 5 schematic diagrams of the present invention are made up of three parts: mark point figure, driving school code number character, training vehicle serial number character: mark point figure must be able to clearly indicate sign plate position and direction, as adopting T type mark, And the horizontal line segment is composed of 5 black circle figures on a white background, and the vertical line segment is composed of 3 white circle figures on a black background. The horizontal line segment and the vertical line segment can be easily distinguished and identified through graphic binarization and connected domain extraction. There is a vertical angle between the horizontal line segment and the vertical line segment; the driving school code characters use two English letters, such as "Huafeng Driving School" can be referred to by "HF"; the training car number characters use three Arabic numerals.

As shown in Figure 3 is the overall system work flow chart of the present invention:

Step S1: By directly manually measuring the size of the curved road and running the camera calibration program, the mutual conversion between the image coordinate system and the world coordinate system is realized;

Step S2: The image analyzer 4 continuously receives the monitoring images captured by the main camera 1, compresses them into video recordings, and simultaneously runs the signboard recognition program to obtain the position and direction of the signboard 5;

Step S3: The image analyzer 4 runs the vehicle position and attitude detection program according to the calibration result of the main camera 1 and the recognition result of the sign plate 5 to obtain the position and attitude of the vehicle in the world coordinate system; according to the vehicle position, attitude and known vehicle size, Determine the position of the vehicle outline and the shortest distance from the outer edge of the outline to the lane boundary line, and trigger the main camera 1 or auxiliary camera 2 to capture pictures when the vehicle presses the line or the distance is less than the set threshold.

Step S4: The image analyzer 4 sends the vehicle position, attitude, distance measurement results, snapshot pictures, video recordings and other data obtained in the step S3 to the server 7 through the transmission network 6, and the server 7 runs a data management program to analyze various information and send Output application data such as comprehensive scores and project time, store them in the database as data records, store pictures and videos in specified file paths, and run application service programs to support and manage client access.

Step S5: The trainee operates the client 8, logs in to the server 7 through the transmission network 6, browses training-related data records, pictures, videos and other information, and prints and downloads information as required; the coach operates the client 8 and logs in through the transmission network 6 The server 7 browses the data records, pictures, videos and other information of the students being taught, and prints and downloads the information as required.

The purpose of camera calibration is to calculate the correspondence between the world coordinate system and the image coordinate system. The world coordinate system is the coordinate system of the real world where the vehicle is located. It consists of three coordinate axes: X axis, Y axis, and Z axis. The image coordinate system That is, the plane image captured by the camera is composed of two coordinate axes: U axis and V axis. Assuming that the coordinates of a point in the three coordinate axes of the world coordinate system X, Y, and Z are (x _w , y _w , z _w ), and the coordinates of the corresponding point in the image coordinate system are (u, v), then this The correspondence between the coordinates of two points can be expressed as a system of linear equations:

$\mathrm{<span\; class="notranslate">\; \&alpha;</span>}\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; u</span>}\\ \mathrm{<span\; class="notranslate">\; v</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; =</span>\left[\begin{array}{cccc}{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& {\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}& {\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right]\left[\begin{array}{cc}\mathrm{<span\; class="notranslate">\; R</span>}& \mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}\\ {\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; MX</span>}$

This equation system is called "image-world coordinate conversion linear equation system", where α is an intermediate parameter in the calculation process, (u, v) is the image pixel coordinate, (x _w , y _w , z _w ) is the world coordinate, ( f _x , f _y ) are the focal lengths in the X-axis and Y-axis directions in the image coordinate system, respectively, and (u ₀ , v ₀ ) are the positions of the intersection of the camera optical axis and the image plane in the image coordinates. T=[t _x , t _y , t _z , 1] is the mapping parameter of the origin of the world coordinates in the image coordinates, R is an orthogonal matrix, defined as:

$\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 11</span>}}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 12</span>}}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 13</span>}}\\ {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; three</span>}}\\ {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 31</span>}}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 31</span>}}& {\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 33</span>}}\end{array}\right]$

The M matrix is the image-world coordinate system transformation matrix, which contains all the camera calibration parameters to be determined. Camera calibration is the process of solving the M matrix. We obtain the world coordinate values {(x _wi , y _wi , z _wi )|i=1,...,4} of the four vertices of the training site by measuring, and obtain the four corner points in the image by manual marking. Coordinate values {(u _i , v _i )|i=1,...,4}, the M matrix can be obtained by solving the above linear equations.

After the M matrix is calculated, any coordinate value of a point in a world coordinate system can be used to calculate the corresponding image coordinate value of the point; any coordinate of a point in an image coordinate system is given and the point is known to be in the world coordinates The height z _w of the corresponding point in the middle can calculate the world coordinates of the corresponding point, so the present invention adopts the camera calibration program flowchart as shown in Figure 4:

Step S101: The camera captures an image, takes any pixel as the origin, and manually marks the image coordinates {(u _i , v _i )|i=1,…,4} of the end points of the four lane boundary lines of the curve curve ;

Step S102: Mark the physical position point where the origin of the image coordinate system is located, and use it as the origin of the world coordinate system, measure the world coordinates {(x _wi , y _wi , z _wi )|i=1,...,4};

Step S103: Substituting the world coordinates and image coordinates of the four endpoints into the image-world coordinate transformation linear equation system to solve the image-world coordinate transformation matrix M;

Step S104: Substituting the coordinate value of a certain point in any world coordinate system into the known image-world coordinate conversion linear equations of M, the image coordinate corresponding to the point can be calculated; on the premise of knowing the height of the world coordinate, the Substituting any image coordinate into the known image-world coordinate transformation linear equations of the image-world coordinate transformation matrix, the corresponding world coordinate of the point can be calculated; so far, the camera calibration is completed.

When adopting the T-shaped signboard shown in Figure 2, black circles and white circles are the main basis for identifying signs, and the present invention can adopt the signboard recognition program flow chart as shown in Figure 5:

Step S201: Set the pixel value at the image coordinates (x, y) as p _{x, y} , define two threshold segmentation images LB, LD for detecting black circles and white circles respectively, and their pixel values are respectively:

${\mathrm{<span\; class="notranslate">\; LB</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; if</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; if</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}\end{array}\right.$

${\mathrm{<span\; class="notranslate">\; LD</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; if</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; if</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}}\end{array}\right.$

Where T _B , T _D are preset values. All pixel values of LB and LD are calculated according to the above formula.

Step S202: For the detection of white circles, find out the connected area with 1 in LB, convert the length and width values of this area into the length and width values in the world coordinate system, and judge whether the length and width values conform to the values in the signboard 5 The size of the white circle, if it matches, it will be recorded as the detected white circle. In the same way, black circles can be detected by doing the same processing for LD.

Step S203: Use the detected white circles and black circles to judge one by one whether they can be connected to a specific line segment that is consistent with the signboard. If yes, then regard the white circles and black circles as marker detection graphics, otherwise discard the white circles or Black circle; repeat this step until all white and black circles are traversed;

Step S204: If the mark detection pattern is found in step S203, then use the mark point detection pattern as a reference point to locate other additional letter and number areas. As for the identification of letters and numbers, methods such as template matching can be used to identify, because character and number recognition It is already a common technology for image processing, and it is not a core technical point in this patent, so it will not be repeated here. So far, the identification of the whole signboard 5 is completed.

Since the signboard 5 is pasted on the fixed position and direction of the vehicle, the position and posture of the vehicle can be deduced according to the position and direction of the signboard 5 . It is assumed here that the size parameters of the vehicle, such as length, width, height, and the position and direction of the signboard on the vehicle body, have been obtained through measurement.

When using the T-shaped signboard 5 shown in FIG. 2 and the signboard detection method shown in FIG. 3 , the angle Theta between the line segment composed of multiple black circles and the line segment composed of multiple white circles is the direction of the signboard 5 . Assuming that the image coordinates of the sign are (u _i , v _i ), the angle is Theta, and the sign plate 5 is pasted on the roof of the vehicle, then z _w = the height of the vehicle, which is known, so the position and attitude of the positioning vehicle can be used as shown in Figure 6 The vehicle position and attitude detection process shown:

Step S301: Substituting the image coordinates (u _i , v _i ) of the mark detection figure acquired in the sign recognition program into the image-world coordinate conversion linear equations of the camera calibration program, the world coordinates of the mark point detection figure can be obtained (x _w , y _w , z _w );

Step S302: In the world coordinate system, connect the marker point detection graphics into line segments with intersections according to the arrangement rules of the original marker graphics on the signboard;

Step S303: In the world coordinate system, calculate the coordinates of the intersection point and the included angle Theta _w of the line segment. The coordinates of the intersection point are the vehicle position, and the included angle of the line segment is the vehicle posture.

As shown in FIG. 7 is a flow chart of the scoring program of the present invention.

Step S401: Count whether there are obvious operational errors such as the vehicle pressing the line, the time spent exceeding the maximum time limit, and the parking position is not in the specified area. If there is, it is directly determined that the current curve driving is unqualified, and the scoring ends;

Step S402: Determine whether the time is within the standard range. If it does not exceed the standard range, no points will be deducted; 1 point will be deducted for every 5 seconds overtime, until the timeout exceeds 200 seconds, that is, the time spent in step S401 exceeds the maximum time limit;

Step S403: Calculate the final comprehensive score. If it is lower than the set value, it will be judged as unqualified; otherwise, it will be judged as qualified and a judgment score will be given.

The above examples are only to illustrate the technical conception and characteristics of the present invention, and its purpose is to allow people familiar with this technology to understand the content of the present invention and implement it accordingly, and cannot limit the protection scope of the present invention. All equivalent changes or modifications made according to the spirit of the present invention shall fall within the protection scope of the present invention.

Claims (6)

1. a curve driving detection system, it is characterized in that, it comprises the front end system be arranged near curve driving position and the back-end system being deployed in administrative center, described front end system comprises multiple stage main camera, image analyzer and the sign board be placed on training cart, back-end system comprises server and client, front end system connects back-end system by transmission network, monitored picture covers the main camera in warehouse compartment region for shooting with video-corder the travel conditions of warehouse compartment place training cart, the monitoring image that the image analyzer continuous reception main camera be connected with main camera is caught also is compressed into monitoring image, the operation attitude of image analyzer monitor vehicle simultaneously, monitoring result and monitoring image are sent to the server of back-end system by image analyzer, client can be called and check monitoring result in server and monitoring image,

Described sign board is arranged on roof, engine top cover or case cover, and the marker graphic on described sign board is some gauge points, and gauge point is arranged in the specific line segment with fixed angle and unique intersection point;

Also be provided with auxiliary camera in the both sides of every platform main camera, supplement for auxiliary the running orbit catching training cart in warehouse compartment.

2. curve driving detection system according to claim 1, it is characterized in that, the seat in the plane of main camera, auxiliary camera is other is equipped with light compensating lamp, and the synchronizing signal of described light compensating lamp is provided by main camera or auxiliary camera, for carrying out strobe type light filling when illumination is crossed dark.

3. a curve driving detection method, it is characterized in that, described method relates to system and comprises the front end system be arranged near curve driving position and the back-end system being deployed in administrative center, described front end system comprises multiple stage main camera, auxiliary camera, image analyzer and the sign board be placed on training cart, back-end system comprises server and client, front end system connects back-end system by transmission network, monitored picture covers the main camera in warehouse compartment region for shooting with video-corder the travel conditions of warehouse compartment place training cart, image analyzer and main camera, auxiliary camera is connected, concrete detection method comprises the following steps:

Step S1: measure warehouse compartment length and width size by direct labor, runs camera calibration program, realizes the mutual conversion of image coordinate system and world coordinate system;

Step S2: the monitoring image that image analyzer continuous reception main camera is caught, is compressed into video record, simultaneously running mark board recognizer, obtains position and the direction of sign board;

Step S3: image analyzer is according to camera calibration result and sign board recognition result, and operational vehicle position and attitude trace routine, obtains position and the attitude of vehicle; In conjunction with known vehicle dimension, and then determine vehicle ' s contour scope and the profile outer edge bee-line to lane line, and trigger main camera when vehicle line ball or auxiliary camera captures picture; When the sign board of image analyzer identification disappears from the specific region of image, judge that vehicle sails out of corresponding main camera overlay area, image analyzer, by the metrical information of segmentation used time, is all sent to server together with candid photograph picture, video record;

Step S4: after server receives the metrical information that each image analyzer sends, runs scoring procedures, obtains this curve driving comprehensive grading;

Step S5: student or trainer, by client-access server, browse, download and print curve driving relevant scoring, capture picture and video record;

In step S1, when carrying out camera calibration, the step of described camera calibration program is:

Video capture piece image, with any pixel for image coordinate system initial point, hand labeled goes out the image coordinate of four the lane boundary line endpoints in this section of curve bend section;

The physical location point at marking image coordinate origin place, and as world coordinate system initial point, the world coordinates of lane line four end points in actual place, experiment curv bend section;

The world coordinates on four warehouse compartment summits and pixel coordinate are substituted into image-world coordinates transform in system of linear equations, solve image-world coordinates transformed matrix;

Certain point coordinate value under any one world coordinate system is substituted into image-world coordinates transformed matrix known image-world coordinates and transform system of linear equations, the image coordinate of this some correspondence can be calculated; Under the prerequisite of the known world coordinate height, any one image coordinate is substituted into image-world coordinates transformed matrix known image-world coordinates and transform system of linear equations, the world coordinates of this some correspondence can be calculated; So far, camera calibration is completed.

4. curve driving detection method according to claim 3, is characterized in that, in step S3, triggers main camera or auxiliary camera and capture picture and be sent to image analyzer when vehicle line ball or distance are less than setting threshold value in docking process.

5. the curve driving detection method according to claim 3 or 4, is characterized in that, the flow process of carrying out sign board identification in step S2 is as follows:

Threshold segmentation is carried out to image, obtains the image after segmentation;

Connected domain extraction is carried out to the image after Threshold segmentation, judges whether connected domain length and width meets the length and width setting value of sign board gauge point figure, if met, then mark this connected domain;

Can the connected domain of judge mark connect into the shape consistent with real marking board one by one, comprises angle formed by gauge point quantity and gauge point line, if passable, be then reference points detection figure depending on connected region, otherwise give up this connected domain; Repeat this step, until travel through all connected domains;

With reference points detection figure for reference point, identify other additional letter and numeric area and identify, so far completing the identification of whole sign board.

6. curve driving detection method according to claim 5, is characterized in that, step in this detection method

The steps flow chart carrying out vehicle location attitude detection in S3 is:

Image-world coordinates that the image coordinate of the reference points detection figure obtained in sign board recognizer substitutes into camera calibration program is transformed in system of linear equations, the world coordinates of reference points detection figure can be obtained;

In world coordinate system, reference points detection figure is linked to be according to the queueing discipline of sign board mark original tally dot pattern the line segment that there is intersection point;

In world coordinate system, calculate the coordinate of intersection point and the angle of line segment, intersecting point coordinate is vehicle location, and line segment angle is vehicle attitude.