A method and device for monitoring a lifting vehicle

Abstract

The invention relates to the field of crane monitoring technology, in particular to a crane monitoring method, the technical scheme of which includes ADAS acquisition, image preprocessing, lane recognition, lane departure prediction, vehicle distance prediction and collision prediction; the HSV space is used to recognize the lane and the road surface, so as to avoid the influence of illumination distribution and foggy weather on the recognition accuracy; when the straight lane is not recognized, the shape of the curved lane is judged, the parabolic lane is set as a Bezier curve, the clothoid lane is set as a B-spline curve, and the mathematical model of the corresponding curved lane is constructed according to the algorithm properties of the Bezier curve and the B-spline curve, so as to improve the accuracy of lane recognition and lane departure prediction; the accuracy of vehicle distance prediction and collision prediction is improved according to the curvature of the curved lane in combination with the corresponding mathematical model; the accurate prediction of lane departure, vehicle distance and collision by the crane enhances the safety of the crane during driving, ensures the safety of construction work and reduces the possibility of accidents.

Description

Technical Field

[0001] The invention relates to the technical field of crane vehicle monitoring, and in particular to a crane vehicle monitoring method and device. Background Art

[0002] Advanced driver assistance systems use a variety of sensors installed on the vehicle to sense the surrounding environment at any time during the driving process, allowing the driver to be aware of possible dangers in advance, effectively increasing the comfort and safety of car driving. Today, ADAS cameras have the functions of lane departure prediction, vehicle distance prediction and collision prediction. However, the ADAS camera monitoring technology for cranes needs to be improved. When cranes are driving on parabolic lanes or spiral lanes, ADAS cameras cannot accurately identify curved lanes, resulting in the inability to accurately predict lane departure, vehicle distance and collision, which will inevitably cause the driver to misjudge and cause traffic accidents. In addition, uneven road lighting distribution and foggy weather will also affect the accuracy of lane recognition.

[0003] In view of this, we propose a crane vehicle monitoring method and device to solve the existing problems. Summary of the invention

[0004] The object of the present invention is to provide a method and device for monitoring a lifting vehicle to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a crane vehicle monitoring method, characterized in that it includes:

[0006] S1, ADAS camera collects images of the road conditions ahead and pre-processes the collected images;

[0007] S2, using Hough line detection to extract straight lanes, and screening the straight lines by the slope of the straight line and the intersection of the straight line and the horizontal coordinate to identify the straight lanes;

[0008] S3. If S2 does not identify a straight lane, the HSV space is used to identify the curved lane and its shape is judged; if the judgment result is a parabolic lane, the inner side of the parabolic lane is set as a Bezier curve, and S4 is performed; if the judgment result is a clothoid lane, the inner side of the clothoid lane is set as a B-spline curve, and S5 is performed; if the judgment result is a straight lane or S2 identifies a straight lane, the lane departure is predicted according to the slope of the straight line and the intersection of the straight line and the horizontal coordinate, and S6 is performed;

[0009] S4, obtaining the control points of the Bezier curve, using the control points as evaluation indicators of the arc of the parabolic lane and predicting lane departure, and then proceeding to S6;

[0010] S5, obtaining the control points of the B-spline curve, using the control points as evaluation indicators of the arc of the clothoid lane and performing lane departure prediction, and then proceeding to S6;

[0011] S6. Predict vehicle distance and collision.

[0012] Furthermore, in step S1, the preprocessing includes scale transformation, perspective transformation, grayscale processing, image filtering, threshold segmentation, and edge detection, wherein the scale transformation includes image scaling and ROI selection.

[0013] Furthermore, the ROI is selected to extract the road surface area in the image according to the HSV space recognition.

[0014] Further, in step S4, the evaluation index of the curvature of the parabolic lane is the turning amplitude of the parabolic lane, and a corresponding color tone is marked on the parabolic lane of the image according to the turning amplitude of the parabolic lane.

[0015] Further, in step S5, the evaluation index of the curvature of the clothoid curve lane is the number of turns of the clothoid curve lane and the amplitude of each turn, and the corresponding color tone is marked on the clothoid curve lane of the image according to the amplitude of each turn of the clothoid curve lane.

[0016] Furthermore, in step S6, HOG feature extraction and SVM training are used to perform vehicle recognition.

[0017] Furthermore, in step S6, if the vehicle is traveling in a straight lane, the vehicle distance is the straight-line distance between the vehicle and the nearest vehicle in front, and a corresponding color tone is marked on the nearest vehicle in front of the image according to the straight-line distance.

[0018] Further, in step S6, if the vehicle is traveling in a parabolic lane, the vehicle distance is the Bezier curve distance between the vehicle and the nearest vehicle in front, and a corresponding color tone is marked on the nearest vehicle in front of the image according to the Bezier curve distance.

[0019] Furthermore, in step S6, if the vehicle is traveling in a roundabout lane, the vehicle distance is the B-spline curve distance between the vehicle and the nearest vehicle in front, and a corresponding color tone is marked on the nearest vehicle in front of the image according to the B-spline curve distance.

[0020] A crane vehicle monitoring device, according to the vehicle monitoring method, comprises:

[0021] ADAS acquisition module, image preprocessing module, lane recognition module, lane departure prediction module, vehicle distance prediction module, and collision prediction module.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] The present invention adopts HSV space to identify lanes and road surfaces, avoiding the influence of illumination distribution and foggy weather on the recognition accuracy; when a straight lane is not identified, the shape of the curved lane is judged, the parabolic lane is set as a Bezier curve, the clothoid lane is set as a B-spline curve, and the mathematical model of the corresponding curved lane is constructed according to the algorithmic properties of the Bezier curve and the B-spline curve, thereby improving the accuracy of lane recognition and lane departure prediction; the accuracy of vehicle distance prediction and collision prediction is improved according to the curvature of the curved lane combined with the corresponding mathematical model; the accurate prediction of lane deviation, vehicle distance and collision by the crane vehicle enhances the safety of the crane vehicle during driving, and can better ensure the safety of construction operations, thereby reducing the possibility of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of the overall steps of the crane vehicle monitoring method of the present invention;

[0025] Figure 2 is a flowchart of the steps of image preprocessing of the present invention;

[0026] Figure 3 A flow chart of the steps of a method for monitoring a crane vehicle on a straight lane identified by Hough straight line detection according to the present invention;

[0027] Figure 4 A flow chart of the steps of a method for monitoring a crane vehicle on a straight lane identified in HSV space according to the present invention;

[0028] Figure 5 A flow chart of the steps of a method for monitoring a crane vehicle on a parabolic track of the present invention;

[0029] Figure 6 The present invention is a flowchart of the steps of the monitoring method of a crane vehicle on a roundabout lane. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0031] Embodiment 1

[0032] like Figure 1 , Figure 2 and Figure 3 As shown, the monitoring method of a crane vehicle on a straight lane identified by Hough straight line detection includes the following steps:

[0033] The ADAS camera collects images of the road ahead and preprocesses the collected images. The preprocessing includes scale transformation, perspective transformation, grayscale processing, image filtering, threshold segmentation, and edge detection. The scale transformation includes image scaling and ROI selection. The ROI selection extracts the road surface area in the image based on the HSV space recognition.

[0034] Hough line detection is used to extract straight lanes, and the straight lines are screened by the slope of the straight line and the intersection of the straight line and the horizontal coordinate to identify the straight lanes;

[0035] Lane departure prediction is performed based on the slope of the straight line and the intersection of the straight line and the horizontal axis;

[0036] HOG feature extraction and SVM training are used for vehicle recognition;

[0037] Predict the straight-line distance between the vehicle and the nearest vehicle in front, and mark the nearest vehicle in front of the image with a corresponding color tone according to the straight-line distance;

[0038] A collision prediction is made based on the speed of the vehicle and the nearest vehicle in front.

[0039] The working principle of the monitoring method of a crane vehicle on a straight lane identified by Hough straight line detection based on the first embodiment is:

[0040] The original image has too many pixels, so the calculation amount of target detection is too large, which will take a long time and affect the real-time performance of detection. Image scaling is to reduce pixels, thereby reducing the calculation amount of target detection, thereby reducing time and improving the real-time performance of detection.

[0041] ROI selection is to extract the road surface area in the image, thereby eliminating the interference of useless areas to narrow the detection range and save detection time.

[0042] The color space of the original image is RGB space. Since external light will affect the brightness of the road surface, fog will affect the saturation of the road surface, which will affect the color intensity corresponding to RGB, thereby affecting the accuracy of target extraction. Therefore, it is necessary to convert the RGB space into HSV space, where H is hue, S is saturation, and V is brightness. Only analyzing the H dimension of the image can avoid the interference of saturation and brightness, and thus avoid the influence of external light and fog on image recognition, thereby improving the accuracy of image recognition.

[0043] The purpose of perspective transformation is to convert the road surface area in the image into a top-down view, so as to accurately determine the lane shape and the angle between the vehicle and the lane, and then accurately predict lane departure, vehicle distance and collision.

[0044] The mathematical model of perspective transformation is:

[0045]

[0046] Where: is the coordinate matrix before transformation; is the transformed coordinate matrix; is the transformation matrix; usually a 33 and w and z are set to constant 1.

[0047] For the identified straight lane, lane departure prediction is performed according to the angle between the vehicle's driving direction and the straight lane and the vehicle's current driving speed. The vehicle's driving direction and the angle between the straight lane are determined according to the slope of the straight line. The relative position of the vehicle and the straight lane is determined according to the intersection of the straight line and the horizontal coordinate. The straight lane in the image is marked according to the result of lane departure prediction. The marked color adopts the hue range from cyan to green, then to yellow and finally to red in the H dimension of the HSV color space. When the vehicle has an intersection with the straight lane, the straight lane is marked red. When the vehicle and the straight lane are parallel to each other, the straight lane is marked cyan. If the vehicle and the straight lane are not parallel and there is no intersection, the time when the vehicle deviates from the lane is predicted according to the angle and relative position between the vehicle's driving direction and the straight lane and the vehicle's driving speed. The longer the predicted time is, the closer the hue of the straight lane marking is to cyan, and the shorter the predicted time is, the closer the hue of the straight lane marking is to red.

[0048] Predict the straight-line distance between the vehicle and the nearest vehicle in front, and mark the corresponding color tone on the nearest vehicle in front of the image according to the straight-line distance. The H dimension of the HSV color space is also used to range from cyan to green, then to yellow and finally to red. The longer the predicted distance, the closer the color tone of the nearest vehicle in front is to cyan, and the shorter the predicted distance, the closer the color tone of the nearest vehicle in front is to red.

[0049] The collision is predicted based on the speed of the vehicle and the nearest vehicle in front, and the collision time is calculated based on the predicted distance. The H dimension of the HSV color space is also used to range from cyan to green, then to yellow and finally to red. When it is predicted that the vehicle is about to collide with the nearest vehicle in front, the outline of the nearest vehicle in front is marked as a red rectangle. When it is predicted that the vehicle will not collide with the nearest vehicle in front, the outline of the nearest vehicle in front is marked as a cyan rectangle. The longer the predicted time, the closer the hue of the outline rectangle of the nearest vehicle in front is to cyan, and the shorter the predicted distance, the closer the hue of the outline rectangle of the nearest vehicle in front is to red.

[0050] Embodiment 2

[0051] like Figure 1 , Figure 2 and Figure 4As shown, the monitoring method of the crane vehicle on the straight lane identified in the HSV space includes the following steps:

[0052] The ADAS camera collects images of the road ahead and preprocesses the collected images. The preprocessing includes scale transformation, perspective transformation, grayscale processing, image filtering, threshold segmentation, and edge detection. The scale transformation includes image scaling and ROI selection. The ROI selection extracts the road surface area in the image based on the HSV space recognition.

[0053] Hough line detection is used to extract straight lanes, and the lines are screened by the slope of the line and the intersection of the line and the horizontal coordinate. No straight lanes are identified;

[0054] The lanes are identified using HSV space and their shapes are judged as straight lines;

[0055] Lane departure prediction is performed based on the slope of the straight line and the intersection of the straight line and the horizontal axis;

[0056] HOG feature extraction and SVM training are used for vehicle recognition;

[0057] Predict the straight-line distance between the vehicle and the nearest vehicle in front, and mark the nearest vehicle in front of the image with a corresponding color tone according to the straight-line distance;

[0058] A collision prediction is made based on the speed of the vehicle and the nearest vehicle in front.

[0059] The working principle of the monitoring method of a crane vehicle on a straight lane of HSV space recognition based on the second embodiment is:

[0060] Hough line detection is used to extract straight lanes, and the lines are screened by the slope of the line and the intersection of the line and the horizontal axis. No straight lanes are identified.

[0061] The lane is identified through the H dimension of the HSV space. According to the color pixel distribution of the lane, if the lane can be fitted into a straight line through the least squares method, the lane is determined to be a straight lane, and the straight lane is analyzed based on the fitted straight line. The following lane departure prediction, vehicle distance prediction and collision prediction methods are the same as those in the embodiment.

[0062] Embodiment 3

[0063] like Figure 1 , Figure 2 and Figure 5 As shown, the monitoring method of a crane vehicle on a parabolic track includes the following steps:

[0064] The ADAS camera collects images of the road ahead and preprocesses the collected images. The preprocessing includes scale transformation, perspective transformation, grayscale processing, image filtering, threshold segmentation, and edge detection. The scale transformation includes image scaling and ROI selection. The ROI selection extracts the road surface area in the image based on the HSV space recognition.

[0065] Hough line detection is used to extract straight lanes, and the lines are screened by the slope of the line and the intersection of the line and the horizontal coordinate. No straight lanes are identified;

[0066] The lanes are identified using HSV space and their shapes are judged as parabolas;

[0067] Set the parabolic lane to a Bezier curve;

[0068] Obtain the control points of the Bezier curve, use the control points as evaluation indicators of the curvature of the parabolic lane and perform lane departure prediction, wherein the evaluation indicator of the curvature of the parabolic lane is the turning amplitude of the parabolic lane, and mark the corresponding color tone on the parabolic lane of the image according to the turning amplitude of the parabolic lane;

[0069] HOG feature extraction and SVM training are used for vehicle recognition;

[0070] Predicting the Bezier curve distance between the vehicle and the nearest vehicle in front, and marking the nearest vehicle in front of the image with a corresponding color tone according to the Bezier curve distance;

[0071] A collision prediction is made based on the speed of the vehicle and the nearest vehicle in front.

[0072] The working principle of the monitoring method of a crane vehicle on a parabolic track based on the third embodiment is:

[0073] Hough line detection is used to extract straight lanes, and the lines are screened by the slope of the line and the intersection of the line and the horizontal axis. No straight lanes are identified.

[0074] The lane is identified through the H dimension of the HSV space. According to the color pixel distribution of the lane, if the lane cannot be fitted into a straight line through the least squares method, the lane is judged to be a curved lane, and the inner side curve of the curved lane is extracted. The leftmost and rightmost coordinate points of the curve are recorded, and a vertical line is set in the image. The horizontal coordinate of the vertical line is set to the mean of the horizontal coordinates of the leftmost and rightmost coordinate points. If the number of intersections between the vertical line and the curve is 1 or 2, the curved lane is judged to be a parabolic lane.

[0075] Suppose the bottom point of the curve is the starting point, and the curve is rotated 45° clockwise with the starting point as the rotation axis using the polar coordinate equation. The polar coordinate equations are x=r·cos(θ) and y=r·sin(θ), where r is the straight-line distance from any point on the curve to the starting point, and θ is 45°. Suppose the horizontal and vertical coordinates of the starting point are both 0, then x and y are the horizontal and vertical coordinate values ​​of any point on the curve after rotation, and a Bezier curve model is constructed for the rotated curve.

[0076] The mathematical model of the Bezier curve is:

[0077]

[0078] Where: is the Bessel basis function; p i is the control point of the Bezier curve; u is the value of the Bezier curve before transformation; P(u) is the value of the Bezier curve after transformation.

[0079] Normalize the horizontal and vertical coordinates of the rotated curve, select N points on the curve, the corresponding horizontal coordinates are 0, 1 / N, 2 / N, ..., (N-1) / N, 1, record the vertical coordinate values ​​corresponding to the N points, take the horizontal coordinate as the value before the Bezier curve transformation, take the vertical coordinate as the value after the Bezier curve transformation, and obtain the control points p0, p1, p2, ..., p N-1 、p N . Calculate max(|p n -n / N|), where n=0, 1, 2, ..., N-1, N. max(|p n -n / N|) is larger, the greater the turning range of the parabolic lane is, and max(|p n The smaller the turning angle of the parabolic lane, the smaller the turning angle of the parabolic lane. The corresponding hue is marked on the parabolic lane of the image according to the turning angle of the parabolic lane. The marked color adopts the hue range from cyan to green to yellow and finally to red in the H dimension of the HSV color space. When max(|p n -n / N|)=1, then the parabolic lane is marked in red. n -n / N|)=0, the straight lane is marked as cyan, max(|p n The closer the parabolic lane marking is to cyan, the closer the color of the parabolic lane marking is to cyan. n The closer -n / N|) is to 1, the closer the hue of the parabolic lane marking is to red.

[0080] The time is calculated based on the vehicle's current speed and the distance between the vehicle and the intersection of the straight line and the curve in the driving direction, and this is used as the result of lane departure prediction. The inner side of the curved lane in the image is marked based on the result of lane departure prediction. The marked color uses the hue range from cyan to green, then to yellow and finally to red in the H dimension of the HSV color space. When the vehicle has an intersection with the curved lane, the inner side of the curved lane is marked in red. If the vehicle has no intersection with the curved lane, the longer the predicted time is, the closer the hue of the mark on the inner side of the curved lane is to cyan, and the shorter the predicted time is, the closer the hue of the mark on the inner side of the curved lane is to red.

[0081] Predict the Bezier curve distance between the vehicle and the nearest vehicle in front, and mark the corresponding color tone on the nearest vehicle in front of the image according to the Bezier curve distance. According to the position corresponding to the Bezier curve of the nearest vehicle in front, the Bezier curve part between the vehicle and the vehicle is divided into K small segments. The lengths of K small segments are calculated and summed on the Bezier curve to be used as the Bezier curve distance. The H dimension of the HSV color space is also used to range the color tone from cyan to green to yellow and finally to red. The longer the predicted distance is, the closer the color tone of the nearest vehicle in front is to cyan, and the shorter the predicted distance is, the closer the color tone of the nearest vehicle in front is to red.

[0082] The collision prediction method is the same as that in the embodiment.

[0083] Embodiment 4

[0084] like Figure 1 , Figure 2 and Figure 6 As shown, the monitoring method of a crane vehicle on a roundabout lane includes the following steps:

[0085] The ADAS camera collects images of the road ahead and preprocesses the collected images. The preprocessing includes scale transformation, perspective transformation, grayscale processing, image filtering, threshold segmentation, and edge detection. The scale transformation includes image scaling and ROI selection. The ROI selection extracts the road surface area in the image based on the HSV space recognition.

[0086] Hough line detection is used to extract straight lanes, and the lines are screened by the slope of the line and the intersection of the line and the horizontal coordinate. No straight lanes are identified;

[0087] The lanes are identified using HSV space and their shapes are judged as clothoid curves;

[0088] Set the clothoid-shaped lane to a B-spline curve;

[0089] Obtain the control points of the B-spline curve, use the control points as evaluation indicators of the curvature of the clothoid lane and perform lane departure prediction, wherein the evaluation indicator of the curvature of the clothoid lane is the number of turns of the clothoid lane and the amplitude of each turn, and mark the corresponding color tone on the clothoid lane of the image according to the amplitude of each turn of the clothoid lane;

[0090] HOG feature extraction and SVM training are used for vehicle recognition;

[0091] Predict the B-spline curve distance between the vehicle and the nearest vehicle in front, and mark the corresponding color tone on the nearest vehicle in front of the image according to the B-spline curve distance;

[0092] A collision prediction is made based on the speed of the vehicle and the nearest vehicle in front.

[0093] The working principle of the monitoring method of a crane vehicle on a roundabout lane based on the fourth embodiment is as follows:

[0094] Hough line detection is used to extract straight lanes, and the lines are screened by the slope of the line and the intersection of the line and the horizontal axis. No straight lanes are identified.

[0095] The lane is identified through the H dimension of the HSV space. According to the color pixel distribution of the lane, if the lane cannot be fitted into a straight line through the least squares method, the lane is judged to be a curved lane, and the inner side curve of the curved lane is extracted. The leftmost and rightmost coordinate points of the curve are recorded, and a vertical line is set in the image. The horizontal coordinate of the vertical line is set to the mean of the horizontal coordinates of the leftmost and rightmost coordinate points. If the number of intersections between the vertical line and the curve is greater than or equal to 3, the curved lane is judged to be a roundabout lane.

[0096] Assume that the bottom point of the curve is the starting point, and the curve is rotated 45° clockwise with the starting point as the rotation axis through the polar coordinate equation. The polar coordinate equations are x=r·cos(θ) and y=r·sin(θ), r is the straight-line distance from any point on the curve to the starting point, θ is 45°, and the horizontal and vertical coordinates of the starting point are both 0. Then x and y are the horizontal and vertical coordinates of any point on the curve after rotation. The B-spline curve model is constructed for the rotated curve. Compared with Bezier curves, B-spline curves are more suitable for segmented curves, so the clothoid curve is analyzed by B-spline curves.

[0097] The mathematical model of the B-spline curve is:

[0098]

[0099] Where: P i,n(t) is the value of the i-th segment after the n-th B-spline curve transformation, where i = 1, 2, 3, ..., m; t is the value before the B-spline curve transformation, and for each point of the curve segment, 0 ≤ t ≤ 1 is satisfied, t = 0 represents the beginning node of the i-th segment, and t = 1 represents the end node of the i-th segment; p i+k-1 is the control point of the i-th segment of the B-spline curve; F k,n (t) is a piecewise mixing function of an n-order B-spline curve, where 0≤t≤1, k=0,1,2,...,n.

[0100] Normalize the horizontal and vertical coordinates of the rotated curve, select N points on the curve, the corresponding horizontal coordinates are 0, 1 / N, 2 / N, ..., (N-1) / N, 1, record the vertical coordinate values ​​corresponding to the N points, take the horizontal coordinate as the value before the B-spline curve transformation, take the vertical coordinate as the value after the B-spline curve transformation, and obtain the control points p0, p1, p2, ..., p N-1 、p N According to p n The number of turns of the roundabout lane is determined by the number of positive and negative changes of -n / N. For each control point corresponding to the curve segment of the turn, calculate max(|p n -n / N|), where n=0, 1, 2, ..., N-1, N. max(|p n -n / N|) is larger, the greater the turning amplitude of the current clothoid curve segment is, and max(|p n The smaller the value of -n / N|), the smaller the turning amplitude of the current clothoid curve segment. According to the turning amplitude of the clothoid curve lane, the corresponding hue is marked on the current clothoid curve segment lane in the image. The marked color adopts the hue range from cyan to green to yellow and finally to red in the H dimension of the HSV color space. When max(|p n -n / N|)=1, the current clothoid curve lane is marked as red. n -n / N|)=0, then the current clothoid curve lane is marked as cyan, max(|p n The closer the value of max(|p n The closer -n / N|) is to 1, the closer the hue of the current clothoid curve segment lane marking is to red.

[0101] The lane departure prediction method is the same as that of the third embodiment.

[0102] Predict the B-spline distance between the vehicle and the nearest vehicle in front, and mark the corresponding color tone on the nearest vehicle in front of the image according to the B-spline distance. According to the position of the B-spline curve corresponding to the nearest vehicle in front, the B-spline curve part between the vehicle and the vehicle is divided into K small segments. The lengths of the K small segments are calculated and summed on the B-spline curve to be used as the B-spline distance. The H dimension of the HSV color space is also used to range the color tone from cyan to green to yellow and finally to red. The longer the predicted distance is, the closer the color tone of the nearest vehicle in front is to cyan, and the shorter the predicted distance is, the closer the color tone of the nearest vehicle in front is to red.

[0103] The collision prediction method is the same as that in the embodiment.

[0104] The above-mentioned specific embodiments are only several preferred embodiments of the present invention. Based on the technical solutions of the present invention and the relevant inspirations of the above-mentioned embodiments, those skilled in the art can make various alternative improvements and combinations to the above-mentioned specific embodiments.

Claims

1. A method for monitoring a lifting vehicle, characterized in that: include: S1, ADAS camera collects images of the road conditions ahead and pre-processes the collected images; S2, using Hough line detection to extract straight lanes, and screening the straight lines by the slope of the straight line and the intersection of the straight line and the horizontal coordinate to identify the straight lanes; S3. If S2 does not identify a straight lane, the HSV space is used to identify the curved lane and its shape is judged; if the judgment result is a parabolic lane, the inner side of the parabolic lane is set as a Bezier curve, and S4 is performed; if the judgment result is a clothoid lane, the inner side of the clothoid lane is set as a B-spline curve, and S5 is performed; if the judgment result is a straight lane or S2 identifies a straight lane, the lane departure is predicted according to the slope of the straight line and the intersection of the straight line and the horizontal coordinate, and S6 is performed; S4, obtaining the control points of the Bezier curve, using the control points as evaluation indicators of the arc of the parabolic lane and predicting lane departure, and then proceeding to S6; S5, obtaining the control points of the B-spline curve, using the control points as evaluation indicators of the arc of the clothoid lane and performing lane departure prediction, and then proceeding to S6; S6. Predict vehicle distance and collision.

2. A method for monitoring a lifting vehicle according to claim 1, characterized in that: In step S1, the preprocessing includes scale transformation, perspective transformation, grayscale processing, image filtering, threshold segmentation, and edge detection, wherein the scale transformation includes image scaling and ROI selection.

3. A method for monitoring a lifting vehicle according to claim 2, characterized in that: The ROI selection is performed based on the HSV space recognition image to extract the road surface area.

4. A method for monitoring a lifting vehicle according to claim 1, characterized in that: In step S4, the evaluation index of the curvature of the parabolic lane is the turning amplitude of the parabolic lane, and a corresponding color tone is marked on the parabolic lane of the image according to the turning amplitude of the parabolic lane.

5. A method for monitoring a lifting vehicle according to claim 1, characterized in that: In step S5, the evaluation index of the arc of the clothoid curve lane is the number of turns of the clothoid curve lane and the amplitude of each turn, and the corresponding color tone is marked on the clothoid curve lane in the image according to the amplitude of each turn of the clothoid curve lane.

6. A crane vehicle monitoring method according to claim 1, characterized in that: In step S6, HOG feature extraction and SVM training are used to perform vehicle recognition.

7. A crane vehicle monitoring method according to claim 1, characterized in that: In step S6, if the vehicle is traveling in a straight lane, the vehicle distance is the straight-line distance between the vehicle and the nearest vehicle in front, and the nearest vehicle in front of the image is marked with a corresponding color tone according to the straight-line distance.

8. A method for monitoring a lifting vehicle according to claim 1, characterized in that: In step S6, if the vehicle is traveling in a parabolic lane, the vehicle distance is the Bezier curve distance between the vehicle and the nearest vehicle in front, and the corresponding color tone is marked on the nearest vehicle in front of the image according to the Bezier curve distance.

9. A method for monitoring a lifting vehicle according to claim 1, characterized in that: In step S6, if the vehicle is traveling in a roundabout lane, the vehicle distance is the B-spline distance between the vehicle and the nearest vehicle in front, and the corresponding color tone is marked on the nearest vehicle in front of the image according to the B-spline distance.

10. A crane vehicle monitoring device, according to a crane vehicle monitoring method according to any one of claims 1 to 9, characterized in that: include: ADAS acquisition module, image preprocessing module, lane recognition module, lane departure prediction module, vehicle distance prediction module, and collision prediction module.

Images