CN110633492A - Lane departure early warning method of Android platform of simulation robot - Google Patents
Lane departure early warning method of Android platform of simulation robot Download PDFInfo
- Publication number
- CN110633492A CN110633492A CN201910711496.6A CN201910711496A CN110633492A CN 110633492 A CN110633492 A CN 110633492A CN 201910711496 A CN201910711496 A CN 201910711496A CN 110633492 A CN110633492 A CN 110633492A
- Authority
- CN
- China
- Prior art keywords
- lane
- early warning
- line
- negative edge
- edge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000004088 simulation Methods 0.000 title claims abstract description 35
- 238000001514 detection method Methods 0.000 claims description 33
- 238000004422 calculation algorithm Methods 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 12
- 238000002474 experimental method Methods 0.000 claims description 9
- 235000008733 Citrus aurantifolia Nutrition 0.000 claims description 7
- 235000011941 Tilia x europaea Nutrition 0.000 claims description 7
- 239000004571 lime Substances 0.000 claims description 7
- 239000000843 powder Substances 0.000 claims description 7
- 208000001491 myopia Diseases 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000013459 approach Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000000630 rising effect Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 11
- 238000009434 installation Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- TVZRAEYQIKYCPH-UHFFFAOYSA-N 3-(trimethylsilyl)propane-1-sulfonic acid Chemical compound C[Si](C)(C)CCCS(O)(=O)=O TVZRAEYQIKYCPH-UHFFFAOYSA-N 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 239000005441 aurora Substances 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
Provided is a lane departure early warning method of an Android platform of a simulation robot. The early warning method comprises five steps, namely initializing a parameter list in the first step, detecting the lane line based on the gradient direction in the second step, updating the parameter list in the third step, detecting the lane line based on the line scanning line in the fourth step, and performing deviation early warning in the fifth step. The lane departure early warning method is used for lane departure early warning of the Android platform of the simulation robot.
Description
The technical field is as follows:
the invention relates to a lane departure early warning method of an Android platform of a simulation robot.
Background art:
the Lane Departure Warning System (LDWS) is an important component of a vehicle driving assistance System, and provides timely Warning to a driver, thereby preventing Lane Departure accidents caused by negligence of the driver.
The installation method of the sensor can be divided into a overlooking system and a forward-looking system.
In the overlooking system, the most representative is the AURORA system; the system consists of a color camera with a wide-angle lens, a digitizer, a portable workstation and the like. The system calculates the transverse position of the vehicle after performing lane mark recognition of a secondary standardized template technology on each frame of image, and then adopts a proper warning triggering criterion. The overlooking system has the advantages that the method is simple and easy to implement, the execution efficiency on structural roads such as highways and the like is higher, and higher positioning precision is easy to obtain; but the disadvantage is that the application range is limited, and the method is only suitable for the structured road with clear identification.
The foresight system is typified by an ultra-strong navigator AutoVue system, a bean-washing net AWSTM system and a decision support DSS system. The AutoVue system comprises a camera arranged at the rear part of a windshield in an automobile, road marking line recognition and tracking software, 2 three-dimensional sound boxes, 1 small display device, a control unit and the like. The position of the vehicle in the current lane is monitored in real time through the camera, the distance from the vehicle to the lane marking line is calculated, and then the distance is compared with the set alarm distance to judge whether to carry out early warning or not. The auss AWS system is based on the research results of intel mobiley corporation, monitoring lane markings and measuring and monitoring the distance of the vehicle from the road boundaries using a single camera mounted on the front windshield. The system can detect different lane markings such as solid continuous lines, broken lines, etc. The lane departure warning module calculates the position of the vehicle relative to the lane and the lateral movement of the vehicle by detecting lane boundaries to predict the time when the vehicle will cross the lane markings, and when the time is below a set threshold, visual and audible warnings are triggered to allow the driver to react appropriately to different dangerous conditions to reduce the occurrence of accidents. The DSS system consists of 1 small shear torque CCD camera, a speed sensor, and audible warning devices such as a visual and indicator and warning buzzer. The method comprises the steps of identifying a lane line by utilizing a CCD camera arranged in a rearview mirror in an image processing mode, calculating the deviation between the actual running track and the expected running track of the current vehicle through a controller, determining the relative distance and speed between the current vehicle and the lane line to judge whether the vehicle starts to deviate from a lane, and calculating the lane crossing time so as to judge whether to give an alarm. The forward looking system may utilize more road information and may be used on roads without road markings. The disadvantage is that some image feature points used to locate the lateral position of the vehicle may be disturbed by other vehicles or pedestrians.
The method in the prior art has strict requirements on installation and equipment, and has a good effect on the resource requirement of the existing rear-mounted rearview mirror (Android platform) and the distorted image generated due to angle change.
The invention content is as follows:
the invention aims to provide a lane departure early warning method of an Android platform of a simulation robot by acquiring continuous images transmitted by a camera in a rearview mirror.
The above purpose is realized by the following technical scheme:
the early warning method comprises five steps, namely initializing a parameter list in the first step, detecting a lane line based on a gradient direction in the second step, updating the parameter list in the third step, detecting the lane line based on a line scanning line in the fourth step, and performing the early warning method for the lane departure in the fifth step.
According to the lane departure early warning method of the Android platform of the simulation robot, the first step of initialization parameter list is that a lane departure early warning system is installed at the position of a rearview mirror, a camera shoots a road surface through an inclination angle, and road surface information of an obtained image is obtained; setting the width of a pixel of an input image as W and the height as H, and detecting the range [ W/5, 4W/5 ] of the width and the range [ H/3, 7H/8 ] of the height in initialization parameters; the coordinate of the vanishing point is (W/2, H/2), the ordinate of the far vision field line is y1 ═ H/2+ d, the ordinate of the near vision field line is y3 ═ 7 × (H/8-d), wherein d is an adjustable threshold value, and the empirical range is [20,50 ]; the ordinate of the mesoscopic field line is y2 ═ (y3+2 × y 2)/3.
The lane departure early warning method of the Android platform of the simulation robot comprises the following steps of firstly, detecting lane lines based on the gradient direction, namely detecting the lane lines V1,
(a) the feature extraction V1 is carried out,
the lane line edge characteristic may be measured in first or second derivatives, which may be represented by a gradient ∑ G:
Judging the types of the positive and negative edge lines: firstly, eliminating edge pixels with approximate horizontal gradient angles, then taking out a row of pixels on an edge gradient image, sequentially judging pixel gradient values from left to right, and judging pixels with gradient directions biased to the right side of the image as rising edges; and (3) judging the positive and negative edge candidate pairs: for each positive edge, a negative edge exists in an effective range (corresponding to reasonable line width in an experiment) on the right side, and the difference between the gradient direction angle of the positive edge and the gradient direction angle of the negative edge meets an angle threshold value, so that the two edges are considered to form a positive and negative edge candidate pair;
neighborhood discrimination: for each positive and negative edge candidate pair, if the average gray value of the left neighborhood of the positive edge is greater than Ti,laneIf so, the background of the positive edge is considered to be consistent; if the average gray value of the right neighborhood of the negative edge is greater than Ti,laneIf the background of the negative edge is consistent with the background of the negative edge, judging that the background of the negative edge is consistent with the background of the negative edge; if the positive and negative edge candidate pairs meet the background coincidence at the same time, the positive and negative edge candidate pairs are considered to pass the domain discrimination;
(b) the model is fitted to the model V1,
longitudinally sorting the features according to the positions of the images, selecting an unused pixel from the sorted list as a seed point, and using a region growing algorithm to generate a region; recursively, those unused neighborhood pixels of the pixel are used for testing, and pixels with an error between angle and region angle of between 22.5 are added to the region; each time a new pixel is added to the region, the region angle is updated; this continues until no pixels can be added to the rectangular area.
In the lane departure early warning method of the Android platform of the simulation robot, the third step of updating the parameter list is to count the intersection point position V meeting two straight lines after detecting the lane line V1 and when the number of the straight lines is more than 2i(x, y) is in the vicinity of vanishing point P (x, y)Is stored into the g _ vps set, i.e.
According to the lane departure early warning method of the Android platform of the simulation robot, lane line detection based on the line scanning line in the fourth step is that a lane line filtering module is added in lane line detection V2 compared with lane line detection V1, and optimization is carried out;
a rapid extraction method based on line scanning is selected for extraction; positive edge e of lane linepAnd a negative edge evThe following conditions are satisfied,
and the average gray value
The parameter fitting uses a straight line model, and the fitting algorithm adopts a least square method.
The fifth step of departure early warning method is to warn based on the transverse departure of the vehicle at the current position of the lane; giving an early warning when the vehicle approaches or remains near the lane edge; the current position of the vehicle in the lane is obtained by a lane line detection algorithm in real time; x represents the distance between the center of the vehicle and the center of the lane, and assuming that the vehicle is parallel to the lane, the vehicle width is P, and the lane width is W, the current positions of the front wheels of the vehicle relative to the left and right lane edges are as follows:
the simulated lane of the lane departure early warning method of the Android platform of the simulation robot comprises the following steps: simulating a left track, simulating a right track, the simulated left track is provided with a left track groove, the bottom of the left track groove is provided with a group of left fixing holes, the left fixing hole is connected with the left reaming hole, the left fixing rod is inserted into the left fixing hole in an interference fit manner, the bottom of the left fixing rod is a conical tip, one side of the left fixing rod is provided with a left inclined threaded hole, the left inclined threaded hole is connected with the left inclined screw rod, the other side of the left fixed rod is provided with a left inclined threaded hole, the left diagonal threaded hole is connected with a left diagonal screw rod, the included angle between the left fixed rod and the left diagonal screw rod is less than 90 degrees, the left fixing rod, the left inclined screw, the left diagonal screw and the conical tip are all inserted underground; simulation right track open and to have right track groove, the bottom in right track groove open and to have a set of right fixed orifices, right fixed orifices connect right reaming, right fixed orifices in interference fit insert right dead lever, the bottom of right dead lever described be circular cone point, one side of right dead lever open and to have right oblique screw hole, right oblique screw hole connect right oblique screw rod described, the opposite side of right dead lever open and to have right oblique screw hole, right oblique screw hole connect right oblique screw rod, right dead lever with right oblique screw rod between contained angle be less than 90, right dead lever right oblique screw rod circular cone point all insert underground.
The simulated lane of the lane departure early warning method of the Android platform of the simulation robot is characterized in that the top of the left fixing rod is connected with a left cap plate, the left cap plate is inserted into the left expanding hole, the top of the right fixing rod is connected with a right cap plate, the right cap plate is inserted into the right expanding hole, white lime powder is filled in the left track groove, and white lime powder is filled in the right track groove.
Has the advantages that:
1. the lane departure warning method is used for the technical fields of unmanned autonomous driving, safe auxiliary driving and the like, and the lane departure warning standard is realized under the strict requirement of the algorithm on the resource utilization rate; good verification was performed on the vehicle mirror.
2. The invention is suitable for equipment sensitive to the CPU occupancy rate of the central processing unit or uncertain in installation position, and has very good use effect.
3. The invention can provide stable lane line detection information and is suitable for the situation that a lane line is blocked or a part of the lane line is damaged.
4. The invention provides a lane departure early warning algorithm for a rear-view mirror system.
Description of the drawings:
fig. 1 is a flow chart of the implementation of the lane departure warning system of the present invention.
Fig. 2 is a flow chart of the lane departure early warning system implemented based on the Android platform.
FIG. 3 is a frame diagram for realizing an algorithm of a lane departure early warning system based on an Android platform.
Fig. 4 is a diagram of the region of interest and line of sight setting ((a) initialization parameter setting) of the present invention.
Fig. 5 is a view of the region of interest and the view line set ((b) region of interest) of the present invention.
Fig. 6 is a flow chart of updating a parameter list of the present invention.
FIG. 7 is a view showing a case a where the straight line intersects with the present invention.
FIG. 8 is a b-diagram showing the intersection of straight lines according to the present invention.
FIG. 9 is a view of a straight line intersection of the present invention.
FIG. 10 is a schematic diagram of the neighborhood search method of the present invention for determining vanishing points.
Fig. 11 is a diagram of a lane line detection method based on line scanning according to the present invention.
Fig. 12 is a schematic diagram of the lane departure calculation method of the present invention.
FIG. 13 is a diagram of a real-time warning result a on a rearview mirror according to the method of the invention.
FIG. 14 is a diagram of the real-time warning result b of the method on the rearview mirror.
FIG. 15 is a diagram of the real-time warning result c of the method of the present invention on the rearview mirror.
FIG. 16 is a diagram of the real-time warning result d of the method on the rearview mirror.
Fig. 17 is a simulated orbital diagram of the invention.
The specific implementation mode is as follows:
the technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the present invention.
Example 1:
a lane departure early warning method of an Android platform of a simulation robot comprises five steps, wherein a parameter list is initialized in the first step, lane line detection (lane line detection V1) based on a gradient direction is detected in the second step, the parameter list is updated in the third step, the lane line detection based on a line scanning line is performed in the fourth step, and a departure early warning method is performed in the fifth step.
Example 2:
the lane departure early warning method of the Android platform of the simulation robot in embodiment 1 is characterized in that the initialization parameter list of the first step is that a rear view mirror is installed in the lane departure early warning system, a camera shoots a road surface through a certain inclination angle, and most of the obtained images are road surface information; setting the width of a pixel of an input image as W and the height as H, and detecting the range [ W/5, 4W/5 ] of the width and the range [ H/3, 7H/8 ] of the height in initialization parameters;
the coordinate of the vanishing point is (W/2, H/2), the ordinate of the far vision field line is y1 ═ H/2+ d, the ordinate of the near vision field line is y3 ═ 7 × -H/8-d, wherein d is an adjustable threshold value, and the empirical range is [20,50 ]; the ordinate of the mesoscopic field line is y2 ═ (y3+2 × y 2)/3.
Therefore, after the above initialization, we can see that the detected region of interest is as shown in fig. 5.
Example 3:
in the lane departure warning method based on the Android platform of the simulation robot in embodiment 1, the lane line detection based on the gradient direction in the second step is lane line detection V1, which is a lane line detection method based on the gradient direction, and an algorithm flow is shown in fig. 6. This part of the content is aimed at detecting camera mounting tilt or specified detection Region (ROI) errors, and specifically involves the feature extraction V1 and model fitting V1 sub-modules.
(a) Feature extraction V1.
The lane line edge characteristic may be measured in first or second derivatives, which may be represented by a gradient ∑ G:
gxAnd gyAnd (5) adopting edge detection Sobel to obtain. The horizontal and vertical Soble operators can accurately represent the gradient similar to the self direction, but the gradient with large direction difference represents large error, and in order to emphasize the edge along the 45-degree direction, the Soble operator in the 45-degree direction is used, and the gradient vector with the largest amplitude is taken as a final value.
The gradient magnitude threshold used to measure whether it is a lane line edge is also strongly related to image contrast. The invention designs a high threshold value T for local non-maximum inhibition▽H＝βHσiAnd a low threshold T▽L＝βLσi. Wherein beta isH,βLAdjustment coefficients for high and low thresholds, respectively, are set for beta in the experimentH/βLThe better effect is achieved when the ratio is 3; sigmaiIs the gray variance of the ith statistical window image. The statistical window in the invention refers to the area range for calculating the gray characteristic, and the gray variance calculation formula is as follows:
wherein Ii[j]Is the ith gray value of the jth statistical window of statistics.
After the lane line edge is extracted, the lane line feature is discriminated by using the following criteria.
Judging the types of the positive and negative edge lines: firstly, eliminating edge pixels with approximate horizontal gradient angles, then taking out a row of pixels on an edge gradient image, sequentially judging pixel gradient values from left to right, and judging pixels with gradient directions biased to the right side of the image as rising edges; abbreviated as Positive (Positive) edge; similarly, the pixels with gradient direction biased to the left of the image are falling edges, and are marked as Negative (Negative) edges.
And (3) judging the positive and negative edge candidate pairs: for each positive edge, if a negative edge exists in a right effective range (corresponding to a reasonable line width in an experiment), and the difference between the gradient direction angle of the positive edge and the gradient direction angle of the negative edge meets an angle threshold (set to 20 in the experiment), the two edges are considered to form a positive and negative edge candidate pair;
neighborhood discrimination: for each positive and negative edge candidate pair, if the average gray value of the left neighborhood (set to 2 pixels) of the positive edge is greater than Ti,laneIf so, the background of the positive edge is considered to be consistent; if the average gray value of the right neighborhood of the negative edge (set to 2 pixels) is greater than Ti,laneIf the background of the negative edge is consistent with the background of the negative edge, judging that the background of the negative edge is consistent with the background of the negative edge; and if the positive and negative edge candidate pairs meet the background coincidence at the same time, determining that the positive and negative edge candidate pairs pass the domain discrimination. Wherein T isi,laneIs the set background gray scale value.
And after the judgment, taking the central position of the coordinate position of the central edge as the characteristic position of the center of the lane line, namely finishing the characteristic extraction work.
(b) Model fitting V1.
Longitudinally sorting the features according to the positions of the images, selecting an unused pixel from the sorted list as a seed point, and using a region growing algorithm to generate a region; recursively, those unused neighborhood pixels of the pixel are used for testing, and pixels with an error between angle and region angle of between 22.5 are added to the region; each time a new pixel is added to the region, the region angle is updated; this continues until no pixels can be added to the rectangular area. And extracting the point columns in the rectangular area, and performing minimum two-multiplication fitting after the data meet 10 and the point distance is less than 50 pixels to generate a straight line.
Example 4:
an imitation as described in example 1In the lane departure early warning method of the Android platform of the real robot, after the parameter list is updated in the third step and the lane line detection V1 is detected, the parameter list is updated as shown in the attached figure 7. When the number of the straight lines is more than 2, counting the intersection point position V of the two straight linesi(x, y) the straight line around the vanishing point P (x, y) is stored into the g _ vps set, i.e.
In the experiment, Tx is 100 and Ty is 100. When the number of g _ vps sets is larger than AutoNum/3 (the AutoNum is the number of frames for automatic detection), an algorithm is designed to carry out a weight clustering method to screen out credible vanishing points.
Ideally, the inclination angles of the parallel straight lines in the space projected on the image plane are close and both intersect at a point, but in practice, errors always occur in the projection process and the line extraction process of the space straight lines on the image plane, and these errors make the intersection points of the parallel straight lines projected on the image plane in the space not unique, but form a relatively dense point cluster, as shown in fig. 7, ideally, all the straight lines intersect at a point, but most of the intersection points are as shown in fig. 8, and do not intersect at a point any more, but have a plurality of intersection points, and the intersection points are dense, but when there is interference, the intersection points are as shown in fig. 9, and then a plurality of wrong intersection points occur.
Through optimization selection, proper intersection points are selected from each frame and aggregated to form a candidate vanishing point set N, points are classified through a neighborhood search method, and the central point of the point set with the most classified data is selected as the last vanishing point.
Traditionally, classification uses a traversal method, the distance between each point and a certain point is calculated, and then the point smaller than a given threshold value is a neighboring point of the point. If there are n points, the number of times is n (n-1)/2, and the complexity is O (n)2). Analyzing the characteristics of point distribution, and knowing that the positions of the points corresponding to the straight line cluster and the intersection point of the central straight line are concentrated except for the interference points, a field searching method can be adopted.
The domain search method takes each point as a center, searches in a neighborhood range of the point and judges whether other points exist. If so, the points are considered as neighbors of the point. When the size of the neighborhood is M, the search times are M2 n, the complexity of the algorithm is O (n), and the search times are one order of magnitude faster than the traversal method.
The algorithm comprises the following specific steps:
if c (i) is 0, N is the number N for each point i in Nc＝nc+1，C(i)＝ncThen sequentially executing; otherwise, i is equal to i +1, and the step (1) is continued.
Consider each point j (j ═ 1,2, L, m (i)) in the set u (i) of neighboring points of i.
If m (j) > 2 and c (j) ═ 0, c (j) ═ c (i), the set of adjacent points u (j) of j is considered.
If m (j) ≦ 2 and c (j) ≦ 0, c (j) ═ c (i), the set of neighboring points U (j +1) of j +1 is considered again.
If m (j) ≦ 2 and C (j) ≠ 0, we look directly at the neighboring point set U (j +1) of j + 1.
After each point is classified, each weight value of each point is used as the weight value of the point after clustering, and the average value of the coordinates of each point is used as the final coordinate value.
The weight clustering method not only has high searching speed, but also endows different weight values to different intersection points, strengthens the contribution of points at dense positions to final vanishing points, can eliminate the influence of interference points, enhances the robustness of vanishing point detection, and obtains good detection effect.
When updated vanishing point Pi(x, y) within a reasonable range of vanishing point P (x, y), we update the lane line parameters as follows. Vanishing point, update to P ═ Pi。
ROI area and line of sight. Let W be the pixel width of the input image and H be the height of the input image, and in the dynamic parameters, the range [0, W ] of the detection width and the range [ Vh, H ] of the height are detected. The vanishing point coordinates are (Vw, Vh), the far vision field line ordinate is y1 ═ Vh + d, the near vision field line ordinate is y3 ═ H-d, where d is the adjustable threshold, the experimental range is [20,50 ]. The ordinate of the mesoscopic field line is y2 ═ (y3+2 × y 2)/3.
The minimum lane width minWidth. The minimum lane width is the minimum pixel that the pixel distance between two lane lines at least satisfies, and is initialized to the manual designation (140 pix in the experiment). The calculation method is that the vanishing point in the g _ vps set meets the requirement and the distance between two straight lines meets the average value of all distances [ minWidth/2,1.5 minWidth ].
Example 5:
in the method for early warning of lane departure of an Android platform of a simulation robot in embodiment 1, the fourth step of lane line detection based on a line scan line is lane line detection V2, which is a lane line detection method based on line scan, and an algorithm flow is shown in fig. 10. Compared with lane line detection V1, the lane line filtering module is added, and optimization is performed to adapt to the resource requirement of the android platform;
the feature extraction selects a rapid extraction method based on line scanning; positive edge e of lane linepAnd a negative edge evThe following conditions are satisfied.
Wherein g (x, y) is a gray value of the coordinate point (x, y), and the average gray value
The parameter fitting uses a straight line model, and the fitting algorithm adopts a least square method.
Example 6:
the lane departure early warning method of the Android platform of the simulation robot is characterized in that the fifth step of departure early warning method is that a lane departure early warning algorithm warns based on the transverse departure of a vehicle at the current position of a lane; giving an early warning when the vehicle approaches or remains near the lane edge; the current position of the vehicle in the lane is obtained in real time by a lane line detection algorithm; as shown in fig. 11, x represents the distance between the center of the vehicle and the center of the lane, and assuming that the vehicle is parallel to the lane, the vehicle width is P, and the lane width is W, the current positions of the front wheels of the vehicle relative to the left and right lane edges are as follows:
in the formula, the lane width W and the vehicle width P are both known quantities, and x is a calculated value of the lane line detection algorithm. Δ L and Δ R represent the positions of the left and right wheels with respect to the left and right lane boundaries, respectively.
When Δ L > 0 and Δ R > 0, indicating that the vehicle is in the lane, then no warning is needed; when Δ L < 0 and Δ R < 0, the vehicle is considered to be off lane, k > t1When, it is considered as left bias, k < t2Considered as a right deviation. t is t1And t2Can be the same, can also according to the experiment carries on the sensitivity adjustment, in the experiment, t1＝t2＝1。
Example 7:
in the lane departure warning method of the Android platform of the simulation robot in the embodiment, the lane departure warning is implemented by acquiring road surface information of a current driving lane through a camera, processing and analyzing road image information to obtain useful information, and further judging the current state of a vehicle, if the lane departure of the vehicle is judged, a warning signal is sent to a driver to remind the driver to perform relevant operations to correct the current state of the vehicle, and if the lane departure of the vehicle is not judged, the warning signal is not sent.
Fig. 1 shows a flow chart of the lane departure warning system. The invention relates to a lane departure early warning method based on an Android platform, which is characterized in that a camera carried by the Android is used for image acquisition, an algorithm of the system is called after a real-time video stream is obtained, an early warning signal is output, and then the early warning signal is displayed on the Android platform in real time. Fig. 2 shows a flow of implementing the present invention.
The most important two parts in the invention are lane line detection and deviation judgment. In order to obtain a stable real-time lane line detection result, image preprocessing, lane line tracking and other association constraints are required. According to the requirements of installation equipment, the system runs on an Android system platform, the installation position, the environment and the like are uncertain, so that camera calibration cannot be suitable for the scene, and an algorithm implementation framework is shown in attached figure 3.
Example 8:
the simulation lane of the lane departure early warning method of the Android platform of the simulation robot in embodiment 1 comprises a simulation left rail 1 and a simulation right rail 2, wherein the simulation left rail is provided with a left rail groove 3, the bottom of the left rail groove is provided with a group of left fixing holes 4, the left fixing holes are connected with a left reaming hole 5, a left fixing rod 6 is inserted into the left fixing holes in an interference fit manner, the bottom of the left fixing rod is a conical tip 7, one side of the left fixing rod is provided with a left inclined threaded hole 8, the left inclined threaded hole is connected with a left inclined screw rod 9, the other side of the left fixing rod is provided with a left inclined threaded hole 10, the left inclined threaded hole is connected with a left inclined screw rod 11, the included angle between the left fixing rod and the left inclined screw rod is smaller than 90 degrees, and the left fixing rod, the left inclined screw rod and the left inclined screw rod, The left diagonal screw and the conical tip are inserted underground; simulation right track open and to have right track groove 12, the bottom in right track groove open and to have a set of right fixed orifices 13, right fixed orifices connect right reaming 14, right fixed orifices in interference fit insert right dead lever 15, the bottom of right dead lever for coniform point, one side of right dead lever open and to have right oblique screw hole 16, right oblique screw hole connect right oblique screw rod 17, the opposite side of right dead lever open and to have right oblique screw hole 18, right oblique screw hole connect right oblique screw rod 19, right dead lever with right oblique screw rod between contained angle be less than 90, right dead lever said right oblique screw rod circular cone point all insert the underground.
The tops of the simulated left rail and the simulated right rail are on the same plane with the ground, and when the simulated robot passes through the simulated left rail and the simulated right rail, if the simulated robot deviates from the simulated left rail and the simulated right rail, the simulated robot can clearly see and has obvious traces through the displacement of the white lime powder.
Example 9:
the simulated lane of the lane departure warning method of the Android platform of the simulation robot in embodiment 8 is characterized in that the top of the left fixing rod is connected with a left cap plate 20, the left cap plate is inserted into the left enlarged hole, the top of the right fixing rod is connected with a right cap plate 21, the right cap plate is inserted into the right enlarged hole, white lime powder is filled in the left track groove, and white lime powder is filled in the right track groove.
Claims (8)
1. The early warning method for lane departure of the Android platform of the simulation robot is characterized by comprising five steps of initializing a parameter list in the first step, detecting lane lines based on the gradient direction in the second step, updating the parameter list in the third step, detecting the lane lines based on a line scanning line in the fourth step and early warning the lane departure in the fifth step.
2. The lane departure early warning method of the Android platform of the simulation robot according to claim 1, wherein the initialization parameter list of the first step is the position of a rearview mirror installed in the lane departure early warning system, a camera shoots a road surface through an inclination angle, and road surface information of an obtained image is obtained; setting the width of a pixel of an input image as W and the height as H, and detecting the range [ W/5, 4W/5 ] of the width and the range [ H/3, 7H/8 ] of the height in initialization parameters; the coordinate of the vanishing point is (W/2, H/2), the ordinate of the far vision field line is y1 ═ H/2+ d, the ordinate of the near vision field line is y3 ═ 7 × (H/8-d), wherein d is an adjustable threshold value, and the empirical range is [20,50 ]; the ordinate of the mesoscopic field line is y2 ═ (y3+2 × y 2)/3.
3. The method for early warning of lane departure of the Android platform of the simulation robot of claim 1, wherein the lane marking detection based on the gradient direction in the second step is lane marking detection V1,
(a) the feature extraction V1 is carried out,
the lane line edge characteristics may be measured in first or second derivatives, which may be represented by a gradient ∑ G:
Judging the types of the positive and negative edge lines: firstly, eliminating edge pixels with approximate horizontal gradient angles, then taking out a row of pixels on an edge gradient image, sequentially judging pixel gradient values from left to right, and judging pixels with gradient directions deviated to the right side of the image as rising edges; and (3) judging the positive and negative edge candidate pairs: for each positive edge, a negative edge exists in an effective range (corresponding to reasonable line width in an experiment) on the right side, and the difference between the gradient direction angle of the positive edge and the gradient direction angle of the negative edge meets an angle threshold value, so that the two edges are considered to form a positive and negative edge candidate pair;
neighborhood discrimination: for each positive and negative edge candidate pair, if the average gray value of the left neighborhood of the positive edge is greater than Ti,laneIf so, the background of the positive edge is considered to be consistent; if the average gray value of the right neighborhood of the negative edge is greater than Ti,laneIf the background of the negative edge is consistent with the background of the negative edge, judging that the background of the negative edge is consistent with the background of the negative edge; if the positive and negative edge candidate pairs meet the background coincidence at the same time, the positive and negative edge candidate pairs are considered to pass the domain discrimination;
(b) the model is fitted to the model V1,
longitudinally sorting the features according to the positions of the images, selecting an unused pixel from the sorted list as a seed point, and using a region growing algorithm to generate a region; recursively, those unused neighborhood pixels of the pixel are used for testing, and pixels with an error between angle and region angle of between 22.5 are added to the region; each time a new pixel is added to the region, the region angle is updated; this continues until no pixels can be added to the rectangular area.
4. The method for early warning of lane departure of the Android platform of the simulation robot as claimed in claim 1, wherein the parameter list updated in the third step is after lane line detection V1, and when the number of the straight lines is more than 2, the intersection position V of the two straight lines is countedi(x, y) the straight line around the vanishing point P (x, y) is stored into the g _ vps set, i.e.
5. The lane departure early warning method of the Android platform of the simulation robot according to claim 1, wherein the lane line detection based on the row scanning line in the fourth step is that a lane line filtering module is added to the lane line detection V2 compared with the lane line detection V1, and optimization is performed;
a rapid extraction method based on line scanning is selected for extraction; positive edge e of lane linepAnd a negative edge evThe following conditions are satisfied,
and the average gray value
The parameter fitting uses a straight line model, and the fitting algorithm adopts a least square method.
6. The lane departure early warning method of the Android platform of the simulation robot as claimed in claim 1, wherein the fifth step of departure early warning method is to warn based on the lateral departure of the vehicle at the current position of the lane; giving an early warning when the vehicle approaches or remains near the lane edge; the current position of the vehicle in the lane is obtained in real time by a lane line detection algorithm; x represents the distance between the center of the vehicle and the center of the lane, and assuming that the vehicle is parallel to the lane, the vehicle width is P, and the lane width is W, the current positions of the front wheels of the vehicle relative to the left and right lane edges are as follows:
7. the simulated lane of the lane departure early warning method of the Android platform of the simulation robot according to claim 1, which comprises a simulated left rail and a simulated right rail, and is characterized in that the simulated left rail is provided with a left rail groove, the bottom of the left rail groove is provided with a group of left fixing holes, the left fixing holes are connected with left counterbores, left fixing rods are inserted into the left fixing holes in an interference fit manner, the bottom of each left fixing rod is a conical tip, one side of each left fixing rod is provided with a left inclined threaded hole, the left inclined threaded hole is connected with a left inclined screw rod, the other side of each left fixing rod is provided with a left diagonal threaded hole, the left diagonal threaded hole is connected with a left diagonal screw rod, the included angle between the left fixing rod and the left diagonal screw rod is smaller than 90 degrees, and the left fixing rod, the left inclined screw rod and the left diagonal screw rod are connected with a left diagonal threaded hole, The conical tips are all inserted underground; simulation right track open and to have right track groove, the bottom in right track groove open and to have a set of right fixed orifices, right fixed orifices connect right reaming, right fixed orifices in interference fit insert right dead lever, the bottom of right dead lever be circular cone point, one side of right dead lever open and to have right oblique screw hole, right oblique screw hole connect right oblique screw rod, the opposite side of right dead lever open and to have right oblique screw hole, right oblique screw hole connect right oblique screw rod, right dead lever with right oblique screw rod between contained angle be less than 90, right dead lever right oblique screw rod circular cone point all insert the underground.
8. The simulated lane of the lane departure early warning method of the Android platform of the simulation robot according to claim 7, wherein the top of the left fixing rod is connected with a left cap plate, the left cap plate is inserted into the left enlarged hole, the top of the right fixing rod is connected with a right cap plate, the right cap plate is inserted into the right enlarged hole, white lime powder is filled in the left track groove, and white lime powder is filled in the right track groove.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910711496.6A CN110633492A (en) | 2019-08-02 | 2019-08-02 | Lane departure early warning method of Android platform of simulation robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910711496.6A CN110633492A (en) | 2019-08-02 | 2019-08-02 | Lane departure early warning method of Android platform of simulation robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110633492A true CN110633492A (en) | 2019-12-31 |
Family
ID=68969241
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910711496.6A Pending CN110633492A (en) | 2019-08-02 | 2019-08-02 | Lane departure early warning method of Android platform of simulation robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110633492A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111368659A (en) * | 2020-02-24 | 2020-07-03 | 同济大学 | Intelligent port AGV unmanned vehicle driving track correction method |
| CN115713758A (en) * | 2022-11-10 | 2023-02-24 | 国能黄骅港务有限责任公司 | Carriage identification method, system and device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140204212A1 (en) * | 2002-05-03 | 2014-07-24 | Magna Electronics Inc. | Vision system for vehicle |
| CN103996053A (en) * | 2014-06-05 | 2014-08-20 | 中交第一公路勘察设计研究院有限公司 | Lane departure alarm method based on machine vision |
| CN105261020A (en) * | 2015-10-16 | 2016-01-20 | 桂林电子科技大学 | Method for detecting fast lane line |
| CN108875657A (en) * | 2018-06-26 | 2018-11-23 | 北京茵沃汽车科技有限公司 | A kind of method for detecting lane lines |
| CN109948552A (en) * | 2019-03-20 | 2019-06-28 | 四川大学 | It is a kind of complexity traffic environment in lane detection method |
-
2019
- 2019-08-02 CN CN201910711496.6A patent/CN110633492A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140204212A1 (en) * | 2002-05-03 | 2014-07-24 | Magna Electronics Inc. | Vision system for vehicle |
| CN103996053A (en) * | 2014-06-05 | 2014-08-20 | 中交第一公路勘察设计研究院有限公司 | Lane departure alarm method based on machine vision |
| CN105261020A (en) * | 2015-10-16 | 2016-01-20 | 桂林电子科技大学 | Method for detecting fast lane line |
| CN108875657A (en) * | 2018-06-26 | 2018-11-23 | 北京茵沃汽车科技有限公司 | A kind of method for detecting lane lines |
| CN109948552A (en) * | 2019-03-20 | 2019-06-28 | 四川大学 | It is a kind of complexity traffic environment in lane detection method |
Non-Patent Citations (1)
| Title |
|---|
| 高海龙，等: "基于VBOX-ADAS模块车道保持系统测试研究", 《汽车科技》, 31 December 2018 (2018-12-31), pages 95 - 99 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111368659A (en) * | 2020-02-24 | 2020-07-03 | 同济大学 | Intelligent port AGV unmanned vehicle driving track correction method |
| CN115713758A (en) * | 2022-11-10 | 2023-02-24 | 国能黄骅港务有限责任公司 | Carriage identification method, system and device and storage medium |
| CN115713758B (en) * | 2022-11-10 | 2024-03-19 | 国能黄骅港务有限责任公司 | Carriage identification method, system, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106919915B (en) | Map road marking and road quality acquisition device and method based on ADAS system | |
| EP3296923B1 (en) | A method of detecting an overtaking vehicle, related processing system, overtaking vehicle detection system and vehicle | |
| CN109583267B (en) | Vehicle target detection method, vehicle target detection device, and vehicle | |
| CN103996053B (en) | Lane departure alarm method based on machine vision | |
| CN112700470B (en) | Target detection and track extraction method based on traffic video stream | |
| CN110992693B (en) | Deep learning-based traffic congestion degree multi-dimensional analysis method | |
| US11727799B2 (en) | Automatically perceiving travel signals | |
| CN110379168B (en) | Traffic vehicle information acquisition method based on Mask R-CNN | |
| US10650256B2 (en) | Automatically perceiving travel signals | |
| CN110298307B (en) | Abnormal parking real-time detection method based on deep learning | |
| US20180299893A1 (en) | Automatically perceiving travel signals | |
| JP2003016430A (en) | Travel path detecting device | |
| RU2636121C2 (en) | Three-dimensional object detecting device | |
| CN110929676A (en) | Deep learning-based real-time detection method for illegal turning around | |
| KR101772438B1 (en) | Apparatus and method for detecting bar-type traffic sign in traffic sign recognition system | |
| CN115240471B (en) | Intelligent factory collision avoidance early warning method and system based on image acquisition | |
| CN112798811A (en) | Speed measurement method, device and equipment | |
| WO2018195150A1 (en) | Automatically perceiving travel signals | |
| CN113658427A (en) | Road condition monitoring method, system and equipment based on vision and radar | |
| US20180300566A1 (en) | Automatically perceiving travel signals | |
| CN110633492A (en) | Lane departure early warning method of Android platform of simulation robot | |
| CN113380038A (en) | Vehicle dangerous behavior detection method, device and system | |
| Seo et al. | Use of a monocular camera to analyze a ground vehicle’s lateral movements for reliable autonomous city driving | |
| CN211293945U (en) | Simulated lane of Android platform lane departure early warning method of simulation robot | |
| CN106898023B (en) | Method and system for measuring vehicle head distance based on video image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |