CN202454078U - Road vehicle detecting device - Google Patents

Abstract

The utility model discloses a road vehicle detecting device which comprises a field programmable gate array (FPGA) processing module, a charge coupled device (CCD) binocular camera, a flash memory module, a liquid crystal display (LCD) monitor and a voice output module, wherein the FPGA processing module is connected with the CCD binocular camera, the flash memory module, the LCD monitor and the voice output module and is used for controlling actions of the CCD binocular camera, the flash memory module, the LCD monitor and the voice output module. The road vehicle detecting device can mark vehicles on the monitors for drivers, drivers can know driving conditions of vehicles, then the road vehicle detecting device can judge intelligently whether vehicles collide with front vehicles or not and prompts drivers on the dangerous conditions, the safe driving coefficient is increased, and traffic accidents are avoided.

Description

A kind of road vehicle pick-up unit

Technical field

The utility model relates to a kind of pick-up unit, relates in particular to the embedded intelligence road vehicle pick-up unit of a kind of high real-time, high precision, low false drop rate.

Background technology

Along with the progress and development of society and the raising of vehicle owning rate; Traffic hazard becomes the key factor that threatens the human life day by day; The average per minute in the whole world just at least 1 people is died from vehicle traffic accident; Have every year 3 million peoples need carry out long-term treatment, the direct economic loss that causes is up to 23,000,000,000 dollars.Intelligent Vehicle System (ITS:intelligent transportation systems) reduces the generation of traffic hazard and the loss of social property for the security that improves vehicle ', and saving the human life has very important meaning.

In the past 20 years, in the world wide, the research institute of government has carried out the pioneering research of intelligent vehicle control loop through implementing a series of projects, and has produced some prototypes based on difference in functionality.A new generation ccd video camera and computer vision hardware make the vision sensor parts of high-efficiency and low-cost of large multi-sensor device system become possibility.Vehicle detection is a pith in the Intelligent Vehicle System, has obtained widely paying close attention to, and the vehicle detection that has robustness and reliability concurrently is the first step that can carry out the vehicle-mounted automatic driver assistance system of early warning to driving environment and possible collision.

The utility model content

The utility model technical matters to be solved is: a kind of road vehicle pick-up unit is provided, can satisfies low cost, low-power consumption, high real-time and high-precision demand.

For solving the problems of the technologies described above, the utility model adopts following technical scheme:

A kind of road vehicle pick-up unit, said device comprises: FPGA processing module, CCD binocular camera, FLASH memory module, LCD LCD and voice output module;

Said FPGA processing module is connected with CCD binocular camera, FLASH memory module, LCD LCD, voice output module, in order to the action of control CCD binocular camera, FLASH memory module, LCD LCD, voice output module.

As a kind of preferred version of the utility model, said CCD binocular camera comprise possess automatic gain control, ccd video camera that AWB, high light suppress.

As a kind of preferred version of the utility model, said voice output module comprises big volume hummer.

The beneficial effect of the utility model is: the road vehicle pick-up unit that the utility model proposes; Not only can on display, identify vehicle for the driver; Let the driver that the travel situations of vehicle is accomplished to know what's what, can also whether can collide by the intelligent decision vehicle, voice reminder driver when danger occurring with front vehicles; Improve the safe driving coefficient, the generation that avoids traffic accident.

Description of drawings

Fig. 1 is the system module block diagram of the utility model;

Fig. 2 is the system flowchart of the utility model;

Fig. 3 is the Flame Image Process process flow diagram of the utility model;

Fig. 4 is the desirable video camera pin-hole model of the utility model

Fig. 5 is the fitting of a polynomial figure of the utility model;

Fig. 6 is the barrier world coordinates and the camera position graph of a relation of the utility model.

Embodiment

Specify the preferred embodiment of the utility model below in conjunction with accompanying drawing.

Embodiment one

See also Fig. 1; The utility model has disclosed a kind of road vehicle pick-up unit; Comprise FPGA processing module, CCD binocular camera, FLASH memory module, LCD LCD and voice output module, the FPGA processing module is connected with CCD binocular camera, FLASH memory module, LCD LCD and voice output module; Said FPGA processing module has the software of functions such as reading and writing data, control said CCD binocular camera, control voice output module, road vehicle detection.

In the present embodiment, the ccd video camera that said CCD binocular camera adopts automatic gain control, AWB, high light to suppress.Said FPGA processing module adopts the device that frequency of operation is high, logical block is many, the IO pin is many, the polyp storer is big slightly.Said FLASH memory module adopts read or write speed fast, the device that storage space is big.Said LCD LCD adopts high-clear display.Said voice output module adopts big volume hummer.

As shown in Figure 2, the utility model comprises the following steps: based on the method for the vehicle detection of road vehicle pick-up unit

The left order of said CCD binocular camera and right lens camera are used for collection vehicle the place ahead binocular view data; And the view data that collects passed to said FPGA processing module with the speed of 12 frames/S; Said FPGA processing module is used to control the performance parameter of said CCD binocular camera, and the image data storage that will receive from described CCD binocular camera is in said FLASH memory module.

Said FPGA processing module is read the binocular view data from said FLASH memory module, at first, use the software of binocular measurement, measures the three-dimensional coordinate of object point in the viewfinder range; Then, the software that uses road vehicle to detect is handled the binocular view data of reading; From the view data of reading, detect vehicle, obtain the distance of front vehicles and square vehicle, in image, identify vehicle; Judge whether safety of vehicle ', if safety then continues above-mentioned software processes; If dangerous, then control the voice output module and send audio alert signal prompting driver.

Said FPGA processing module will identify the image data storage of vehicle in said FLASH memory module, and said FPGA processing module read-out mark from said FLASH memory module goes out the view data of vehicle, and passes to said LCD liquid crystal display displays.

Elaborate below:

As shown in Figure 2, the CCD binocular camera is with the binocular image of the speed acquisition vehicle front of 12 frames/S, and the FPGA processing module is carried out the image pre-service to the image that collects, and is used for follow-up road vehicle through pretreated image and detects.

As shown in Figure 3, the flow process that road vehicle detects is:

Convert the coloured image that collects into gray level image, reduce the complexity of subsequent treatment algorithm; The binocular gray level image is carried out gaussian filtering, and filtering noise wherein is to reduce interference.

Three-dimensional coupling; With in the left order image certain a bit is seed points; Use polar curve constraint and parallax constraint; From right order image, find corresponding candidate's point set, with each matching degree of matching degree algorithm computation to seed points and candidate point, the match point of seed points of having got the corresponding candidate point of wherein maximum matching degree; Use the fitting of a polynomial algorithm; With the match point is the center; The pixel in the window of field and the matching degree of corresponding seed points are carried out match, obtain the matching degree function, through differentiate; Pixel abscissa value when calculating the matching degree function and getting maximal value, this value is the horizontal ordinate of the optimal match point of this seed points; The repeating step above-mentioned steps finishes up to the entire image coupling, obtains disparity map.Use is optimized disparity map based on the optimized Algorithm in field.Three-dimensionalreconstruction, the three-dimensional coordinate of use binocular stereo vision Model Calculation object point.

Rim detection uses the Sobel operator to detect the marginal point in the left order image; Cast out the wherein less marginal point of parallax, obtain the marginal point within the certain distance scope; Detection of obstacles uses the relation between object point three-dimensional coordinate and the camera position from left order edge of image point, to detect the marginal point on the barrier.

The marginal point that obtains is projected in XOZ and the XOY plane successively, utilize space length information, carry out area-of-interest and cut apart; Utilize the symmetry feature detection of vehicle to cut apart whether area-of-interest afterwards is vehicle.

Desirable camera model specifies:

Video camera is hinting obliquely at the image device that geometry is principle, can use pin-hole model to come it is carried out modeling, and pin-hole model is a kind of desirable linear camera model.As shown in Figure 4, the plane of delineation of video camera is arranged in camera coordinate system Oc-Xc Yc Zc z=f position, any point thing P (Xw in the space; Yw; Zw) picture point on the plane of delineation is that (u, v), picture point Q is the line OP of photocentre Oc and object point P and the intersection point of plane of delineation O-UV to Q.The world coordinates and the relation between the image coordinate of object point are following:

$z_c\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=\left[\begin{array}{ccc}\frac{1}{d_x}& 0& u_o\\ 0& \frac{1}{d_y}& v_o\\ 0& 0& 1\end{array}\right]\left[\begin{array}{cccc}f& 0& 0& 0\\ 0& f& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{cc}R& t\end{array}\right]\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]$

Wherein, dx, dy are respectively the size factors of video camera level and vertical direction; Uo, vo are the coordinates of image center; F is the focal length of camera lens; R is the rotation matrix of 3*3, and 3 column vectors all are vector of unit length, and mutually orthogonal; T is the translation matrix of 3*1; Rotation matrix R and translation matrix T have quantized the spatial relation of world coordinate system and camera coordinate system.

The detailed description of the important step that road vehicle detects:

Three-dimensional coupling: comprise three-dimensional coupling of thick level and the three-dimensional coupling of smart level

For a specific seed points, candidate point be the set of a pixel, from this a series of candidate point, make a strategic decision out and the highest point of seed points matching degree is crucial.The three-dimensional coupling of thick level is used constraint of level line and parallax constraint, candidate's point set of location seed points.Employing in conjunction with the thought in field, is measured the matching degree of seed points and candidate point based on statistical linearly dependent coefficient.In left order image, as central point, setting up a size is m*n1 (m is a positive integer, and n1 is an odd number) field window with seed points; In right order image, be the center with current candidate point, setting up a size is m*n1 field window.The matching degree computing formula is following:

$\mathrm{coef}=\frac{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\left[(x_i-\stackrel{\‾}{x})(y_i-\stackrel{\‾}{y})\right]}{\sqrt{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(x_i-\stackrel{\‾}{x})}^2}\sqrt{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2}}$

Wherein, xi (i=1,2,3 ..., N) be to be the pixel point value in the field window at center with the seed points;

is the average of all xi; Yi (i=1,2,3 ..., N) be to be the pixel point value in the field window at center with the candidate point;

is the average of all yi; N is a window interior pixel sum, is m*n1.

Choosing candidate point concentrates maximum matching degree to be worth the optimal candidate point that pairing candidate point is a seed points, i.e. Pixel-level optimal match point.

In the three-dimensional coupling of smart level, be the center with the optimal candidate point, in right order image, getting a size is the field window of 1*n2 (n2 is an odd number); Use 3 order polynomials that the matching degree of pixel in the window of field and seed points is carried out match, obtain the function y=f (x) between matching degree and the pixel number, through function is carried out differentiate; When obtaining function f (x) for maximal value, the value of x, i.e. seed points sub-pixel optimal match point; As shown in Figure 5, n2 gets 3 here.

Parallax is optimized: based on the thought in field

This paper has adopted the parallax smoothing algorithm based on the field, and the parallax that the stereo coupling obtains is optimized.In disparity map, choosing any as current point, is the center with the current point, extracts the window of one 5 row 5 row; Successively each point in the window is subtracted with the parallax of central point, takes absolute value again, among the statistical computation result greater than 1 number; If number is within threshold value; Then the central point parallax remains unchanged, otherwise then the central point parallax changes to the average that a parallax is arranged in the window.

Three-dimensionalreconstruction: based on desirable binocular tri-dimensional vision system

Comprise parallax factor in the three-dimensional coordinate of object point,, after parallax optimization, carry out three-dimensionalreconstruction again in order to obtain object point world coordinates preferably.The one object point P in space (Xw, Yw, Zw) picture point in left order image and right order image be respectively Q1 (u1, v1), Q2 (u2 v2), in conjunction with the pin-hole model of video camera, obtains three-dimensionalreconstruction formula in the binocular tri-dimensional vision system:

$\left\{\begin{array}{c}{x}_w=\frac{b(u_1-u_0)}{u_1-u_2}\\ y_w=\frac{b(v_1-v_0)}{u_1-u_2}\\ z_w=\frac{b}{u_1-u_2}f\end{array}\right.$

Wherein, b is left order and the distance of right order photographed images center machine on the Xc direction.

The cognitive disorders thing:

As shown in Figure 6, object point P is positioned at ground, and PR (except that the P point) is an object point collection above the ground, and video camera becomes the θ angle with vertical direction, and the vertical height of video camera is H.The coordinate Z of ordering through P can obtain coordinate Y, and computation process is following:

Y _m＝(Z*cos(θ)-H)/sin(θ)

From Fig. 6, can know, for arbitrary object point, if its Y coordinate, thinks so that this object point is positioned on the barrier less than Ym.

Project to the XOZ plane:

Making the Yw coordinate of marginal point is 0, and the coordinate resolution of choosing Xw and Zw axle is Dx and Dz, and the discrete also projection of the object point space three-dimensional scene of left lens camera viewfinder range is become the two dimensional image on the XOZ plane.For object point P (Zw), the coordinate Calculation formula of projection is following for Xw, Yw:

$\left\{\begin{array}{c}x=[X_w/D_x]\\ y=0\\ x=[Z_w/D_z]\end{array}\right.$

Area-of-interest is cut apart:

Two dimensional image behind the scanning projection detects all eight UNICOMs zones, the number of adding up eight UNICOMs zone interior pixel.For the zone of number less than certain threshold value, think the view field of false barrier, cast out pixels all in this zone.

Choose segmentation threshold TNx and TNz on X and the Z-direction; Detect per pixel distance Nx and the Nz of two eight UNICOM zones on X and Z-direction, if it is two threshold values that pixel distance all makes progress greater than the counterparty are then thought these two projections that the zone is different barriers, in former left order image that it is separated.

Project to XOY plane and area-of-interest is cut apart:

Through after the above-mentioned processing, the barrier in the left order image is divided into several zones to a certain extent, but the barrier that has is not yet by separated.Adopt identical thinking, elementary zone after cutting apart is projected to XOY plane successively, and carry out area-of-interest and cut apart.So, just detected barriers all in the image.

Vehicle detection:

Edge image pixel to current region scans, and utilizes statistical method, finds out axis of symmetry, calculates its symmetry:

$S_{\mathrm{level}}=\frac{1}{K}\underset{(t,k)\∈P}{\mathrm{\Σ}}\frac{|f(t-s,k)-f(t+s,k)|}{255}$

Wherein, P is the point set of current region; (x y) is point (x, the pixel value of y) locating to f; K is the number of symmetric points; S belongs to [0,1], the expression measure of symmetry.

In sum; The road vehicle pick-up unit that the utility model proposes not only can let the driver that the travel situations of vehicle is accomplished to know what's what for the driver identifies vehicle on display; Can also whether can collide by the intelligent decision vehicle with front vehicles; Voice reminder driver when danger occurring improves the safe driving coefficient, the generation that avoids traffic accident.

Here description of the utility model and application is illustrative, is not to want the scope of the utility model is limited in the above-described embodiments.Here the distortion of the embodiment that is disclosed and change are possible, and the replacement of embodiment is known with the various parts of equivalence for those those of ordinary skill in the art.Those skilled in the art are noted that under the situation of spirit that does not break away from the utility model or essential characteristic, and the utility model can be with other form, structure, layout, ratio, and realize with other assembly, material and parts.Under the situation that does not break away from the utility model scope and spirit, can carry out other distortion and change here to the embodiment that is disclosed.

Claims (3)

1. a road vehicle pick-up unit is characterized in that, said device comprises: FPGA processing module, CCD binocular camera, FLASH memory module, LCD LCD and voice output module;

Said FPGA processing module is connected with CCD binocular camera, FLASH memory module, LCD LCD, voice output module, in order to the action of control CCD binocular camera, FLASH memory module, LCD LCD, voice output module.

2. road vehicle pick-up unit according to claim 1 is characterized in that:

Said CCD binocular camera comprise possess automatic gain control, ccd video camera that AWB, high light suppress.

3. road vehicle pick-up unit according to claim 1 is characterized in that:

Said voice output module comprises big volume hummer.

Images