TWI751735B - Automatic guided vehicle tracking system and automatic guided vehicle tracking method - Google Patents

Abstract

The present invention provides an automatic guided vehicle tracking system and an automatic guided vehicle tracking method. The automatic guided vehicle tracking system includes: a data collection unit, obtaining a color image and obtaining two dimensional information; a data computing unit, obtaining a three dimensional coordinate of an object in the color image, obtaining a two dimensional coordinate of the object in the two dimensional information, transforming the three dimensional coordinate to a first spherical coordinate of a spherical coordinate system and transforming the two dimensional coordinate to a second spherical coordinate of the spherical coordinate system, generating a third spherical coordinate of the object according to the first spherical coordinate obtained by a color image depth camera and the second spherical coordinate obtained by a Lidar; and a vehicle control unit including a vehicle controller, controlling a vehicle to follow the object according to the third spherical coordinate.

Description

Self-propelled vehicle following system and self-propelled vehicle following method

The present invention relates to a self-propelled vehicle following system and a self-propelled vehicle following method, and in particular to a self-propelled vehicle following system and a self-propelled vehicle that allow a self-propelled vehicle to follow a target in a field with more interference or complicated routes. Walk and follow method.

In recent years, smart retail and logistics e-commerce have flourished, but the labor force is aging, and labor costs such as handling and picking have risen significantly. In order to solve the problem of the shortage of human resources in the logistics industry, the use of human-machine collaborative follow-up bicycles can improve the efficiency of personnel operations and expand the employment age group.

Service-type bicycles for storage can be roughly divided into navigation-type bicycles and follow-up bicycles. Existing self-propelled self-propelled vehicles are mainly used in fields with relatively simple environments, such as logistics warehouses with fixed shelf positions. In areas with more interference or complicated routes, such as logistics warehouses or retail stores with many people and vehicles, when many people work at the same time, it is inevitable that people will stagger. Once the target is blocked, it cannot continue to follow. Therefore, how to make the self-propelled vehicle follow the person correctly is the goal that those skilled in the art should strive for.

In view of this, the present invention provides a self-propelled vehicle following system and a self-propelled vehicle following method, so that the self-propelled vehicle can follow a target in a field with more interference or complicated route.

The present invention provides a self-propelled vehicle following system, including: a data collection unit, including: a color image depth camera, which obtains a color image; and an optical radar, which obtains two-dimensional information; and a data computing unit, including: an image target tracking module, which obtains The three-dimensional coordinates of the target in the color image; the optical radar target tracking module, and obtain the two-dimensional coordinates of the target in the two-dimensional information; the coordinate fusion module, convert the three-dimensional coordinates into spherical coordinates a first spherical coordinate, converting the two-dimensional coordinate into a second spherical coordinate of the spherical coordinate system; and a control command output module, according to the first spherical coordinate obtained based on the color image depth camera and the the second spherical coordinates obtained by the optical radar to generate the third spherical coordinates of the target; and a vehicle body control unit including a vehicle body controller, which controls the vehicle body to follow the said third spherical coordinates according to the third spherical coordinates Target.

The present invention provides a method for following a bicycle, comprising: obtaining a color image by a color image depth camera, and obtaining two-dimensional information by using an optical radar; obtaining three-dimensional coordinates of a target in the color image, and obtaining the two The two-dimensional coordinates of the target in the dimension information; the three-dimensional coordinates are converted into the first spherical coordinates of the spherical coordinate system, and the two-dimensional coordinates are converted into the second spherical coordinates of the spherical coordinate system; the first spherical coordinates obtained by the color image depth camera and the second spherical coordinates obtained based on the optical radar to generate the third spherical coordinates of the target; and the vehicle body controller according to the first spherical coordinates Three spherical coordinates control the body to follow the target.

Based on the above, the self-propelled vehicle following system and the self-propelled vehicle following method of the present invention use a color image depth camera disposed on the vehicle body to obtain a color image, and use an optical radar disposed on the vehicle body to obtain two-dimensional information, and obtain the color image in the color image. The three-dimensional coordinates of the target and the two-dimensional coordinates of the target in the two-dimensional information, and convert the three-dimensional coordinates and the two-dimensional coordinates into the first spherical coordinates and the second spherical coordinates, and then generate the target according to the first spherical coordinates and the second spherical coordinates. 's third spherical coordinates. Finally, the vehicle body controller of the vehicle body can control the vehicle body to follow the target according to the third spherical coordinate. The self-propelled vehicle following system and the self-propelled vehicle following method of the present invention use a color image depth camera to increase the feature that the target can be identified to make up for the problem of insufficient point cloud image features of the optical radar, so as to detect more accurately in a complex environment Target.

The present invention simultaneously uses two following methods, Lidar and visual image, to make up for each other's deficiencies. Specifically, when the target is blocked, the lidar cannot judge the difference between the target and the obstacle, and it is easy to follow the wrong target. The visual image stores the color, shape, texture and other characteristics of the target for the system to match, so that the following car will not misjudge the target, and when the obstacle is removed, it can continue to follow the correct target. The point cloud image generated by the lidar can obtain the distance between the target and the vehicle body and the approximate shape and size of the target, which is easy to be confused with obstacles with similar shapes during the tracking process. The visual image can store the color distribution, shape, texture and other characteristics of the target, so that the process of tracking the target has more information for interpretation. However, the visual field of view of the visual image is narrow, and when the target exceeds the line of sight of the color image depth (RGB-D) camera, it is easy to lose the target. The lidar has a wide field of view, and can continue to track the target beyond the field of view of the camera, so that the following bicycle can continue to track and follow.

FIG. 1 is a block diagram of a bicycle following system according to an embodiment of the present invention.

Referring to FIG. 1 , a bicycle following system according to an embodiment of the present invention includes a data collection unit 110 , a data calculation unit 120 and a vehicle body control unit 130 . The data collection unit 110 includes a color image depth camera 111 and an optical radar 112 disposed on the body of the bicycle. The color image depth camera 111 can obtain color images. The LiDAR 112 can obtain two-dimensional information such as a point cloud image. The data computing unit 120 can perform operations on the color image and the two-dimensional information to output target coordinates for the vehicle body control unit 130 to control the robot operating system (Robot Operating System, ROS) 131 according to the target coordinates to control the vehicle body to follow the target. The data computing unit 120 includes a person detection module 121 , an image target locking module 122 , an image target tracking module 123 , an optical radar target locking module 124 , an optical radar target tracking module 125 , a coordinate fusion module 126 and a control command Output module 127 . In one embodiment, the data computing unit 120 may include a processor and execute the corresponding modules (ie, the personnel detection module 121 , the image target locking module 122 , the image target tracking module 123 , the LiDAR target locking module The software/firmware code of the group 124, the optical radar target tracking module 125, the coordinate fusion module 126 and the control command output module 127). In another embodiment, each module of the data operation unit 120 may be implemented by a hardware circuit. In another embodiment, each module of the data computing unit 120 may also be implemented through a combination of hardware circuits and/or software/firmware codes. The present invention does not limit the implementation of each module of the data computing unit 120 . Each module of the data computing unit 120 will be described in detail below.

In one embodiment, the data computing unit 120 can execute a target person locking program or a person identification tracking program to perform target locking and tracking of the color depth image. The target person program or person identification tracking program can use deep learning methods to obtain the target person image or obtain the image of everyone in the image, and then use machine learning methods (for example, MobileNet-SSD v2 lite object recognition model) to perform image segmentation And obtain features, for example, divide the obtained person image into multiple equal parts, and obtain the first principal component of the multiple equal parts as the features of the image. The data computing unit 120 stores these characteristics when the program of locking the target person is executed. When performing the process of identifying and tracking the person, the data computing unit 120 can input the feature obtained when the target person is locked as the target feature input to perform feature comparison and obtain the target position, and lock the target and follow according to the identification result. The feature extraction method is, for example, RGB histogram (Histogram).

On the other hand, the data computing unit 120 can also execute a lidar algorithm to perform target locking and tracking of the lidar 2D information. The Lidar algorithm can obtain environmental information (eg, point cloud images) through optical sensors through two-dimensional Lidar. The information is a collection of distance and angle information from the surrounding objects, centered on the LiDAR. In order to avoid being lost due to the change of the size of the object in the picture, the present invention can use the CSRT tracking algorithm to track the object. The CSRT tracking algorithm uses a Discriminative Correlation Filter with Channel and Spatial Reliability (DCF-CSR) algorithm to adjust the filter to ensure that objects can still be tracked when scaled. The CSRT tracking algorithm can calculate the Histogram of Oriented Gradient (HOG) feature and Colornames feature of the selected area, and compare it with the previous frame to determine the current position of the object.

FIG. 2 is a block diagram of a person detection module according to an embodiment of the present invention.

Referring to FIG. 2 , the person detection module 121 can receive the input of the color image 210 , and analyze the color image 210 by the deep learning object detection model 230 to output the coordinates of the bounding frame of the person and the frame image 220 . The deep learning object detection model 230 is, for example, a Single Shot Multibox Detector (SSD). The deep learning object detection model 230 can receive the color image 210 through the classifier 231 and then perform feature extraction through a plurality of convolutional layers 232 to perform the person detection function 233 , and finally output the person whole body bounding frame coordinates and the frame image 220 .

3 is a block diagram of an image target locking module according to an embodiment of the present invention.

Referring to FIG. 3 , the image target locking module 122 will receive the coordinates of the human bounding frame and the frame image 220 from the human identification module 121 , and within the field of view of the color image depth camera 111 , lock the bounding box with the largest area It is a target ( S301 ) and other non-target persons detected by the person detection module 121 are ignored. Next, the image target locking module 122 extracts image features for the bounding box with the largest area and stores it as the target image feature (S302) to output the target image feature 310, and stores the bounding box with the largest area as the target image bounding box (S303). ) to output the target bounding box 320 (or referred to as the first target bounding box).

4 is a block diagram of an image target tracking module according to an embodiment of the present invention.

，其中 x及 y為定界框的基準座標(例如，中心座標或角落座標)， w及 h分別為定界框的寬度及高度。 Please refer to FIG. 4 , the image target tracking module 123 can execute the target tracking module to execute the image target tracking program to receive the target image feature 310 , the target bounding frame 320 , the coordinates of the personnel bounding frame and the frame image 220 to calculate the personnel boundary The coverage of the box and the target bounding box (S401). The image target tracking module 123 also performs the feature matching of the whole body image of the person with the frame image and the target image feature 310 (S402). The image target tracking module 123 then receives the pixel depth information 410 of the color image 210 to locate the 3D center position of the target ( S403 ) to output the 3D coordinates 420 . For example, when the target bounding box is represented by ( x ₁ , y ₁ , w ₁ , h ₁ ) and the person bounding box is represented by ( x ₂ , y ₂ , w ₂ , h ₂ ), the coverage rate (Coverage) Rate)

, where x and y are the reference coordinates (eg, center coordinates or corner coordinates) of the bounding box, and w and h are the width and height of the bounding box, respectively.

FIG. 5 is a block diagram of an optical radar target locking module according to an embodiment of the present invention.

Referring to FIG. 5 , the lidar target locking module 124 can receive the two-dimensional information 510 (eg, a two-dimensional point cloud image) of the lidar 112 to lock the bounding box of predetermined coordinates as the target bounding box (or referred to as the second target bounding box) (S501), initialize the feature tracker (S502) and output the feature tracker 520. Feature tracker 520 is, for example, a leg feature tracker.

FIG. 6 is a block diagram of an optical radar target tracking module according to an embodiment of the present invention.

Referring to FIG. 6 , the LiDAR target tracking module 125 can execute a point cloud target tracking program to receive the 2D information 510 and the feature tracker 520 and track the target according to the 2D information 510 and the feature tracker 520 ( S601 ), and locate the target The two-dimensional center position of the target ( S602 ) to output the two-dimensional coordinates 610 of the target.

Please refer to FIG. 1 again, the coordinate fusion module 126 discards the coordinate value corresponding to the height in the three-dimensional coordinate 420 of the camera coordinate system to convert the three-dimensional coordinate 420 into the first coordinate of the vehicle body coordinate system corresponding to the vehicle body, and then converts the first coordinate is the first spherical coordinate, and the two-dimensional coordinate 610 of the optical radar coordinate system is converted into the second coordinate of the vehicle body coordinate system, and then the second coordinate is converted into the second spherical coordinate. The following will describe the conversion process of the camera coordinate system and the optical radar coordinate system into the vehicle body coordinate system with reference to FIG. 7 .

FIG. 7 is a schematic diagram of the coordinate relationship between a camera coordinate system, an optical radar coordinate system, and a vehicle body coordinate system according to an embodiment of the present invention.

Referring to FIG. 7 , the vehicle body determines the position of the person from the perspective of the vehicle body coordinate system 710 , so the coordinate fusion module 126 converts the data of the camera coordinate system 720 and the optical radar coordinate system 730 to the vehicle body coordinate system 710 to show.

。若車體座標系710與攝影機座標系720之間存在旋轉關係

及平移關係

，則攝影機座標系720的資料點在車體座標系710的第一座標為

，

，其中

,

。 For example, suppose the data point coordinates of the camera coordinate system 720 are

. If there is a rotational relationship between the vehicle body coordinate system 710 and the camera coordinate system 720

and translation relationship

, then the first coordinate of the data point of the camera coordinate system 720 in the vehicle body coordinate system 710 is

,

,in

,

.

。若車體座標系710與光學雷達座標系730之間存在旋轉關係

平移關係

，則光學雷達座標系730的資料點在車體座標系710的第二座標為

，

，其中

,

。 Similarly, suppose the data point coordinates of the lidar coordinate system 730 are

. If there is a rotational relationship between the vehicle body coordinate system 710 and the optical radar coordinate system 730

Translation relationship

, then the second coordinate of the data point of the optical radar coordinate system 730 in the vehicle body coordinate system 710 is

,

,in

,

.

，

，也就是將第一座標及第二座標的笛卡兒座標( x, y)轉換為第一球座標及第二球座標的極座標(

,

)。 Next, the coordinate fusion module 126 can use the following formula to convert the first coordinate and the second coordinate into the first spherical coordinate and the second spherical coordinate, respectively:

,

, that is to convert the Cartesian coordinates ( x , y ) of the first and second coordinates into the polar coordinates of the first and second spherical coordinates (

,

).

FIG. 8 is a flowchart of a third spherical coordinate of an output target of a control command output module according to an embodiment of the present invention.

,

)」801及「B：基於光學雷達判斷的第二球座標(

,

)」802會被輸入控制指令輸出模組127來進行目標位置判斷(S810)。若A及B皆有值(S811)，且

及

相差大於門檻值(S812)，取A為第三球座標(S813)，因為彩色影像深度攝影機111在沒跟丟目標時準確率比光學雷達112高。 Please refer to Fig. 8, "A: The first spherical coordinate determined based on the color image (

,

)" 801 and "B: The second spherical coordinate determined based on the lidar (

,

)" 802 will be input to the control command output module 127 to determine the target position (S810). If both A and B have values (S811), and

and

If the difference is greater than the threshold value ( S812 ), A is taken as the third spherical coordinate ( S813 ), because the color image depth camera 111 has a higher accuracy than the optical radar 112 when it does not follow the target.

及

相差小於等於門檻值(S814)，根據

及

的算術平均值及

及

較小者獲得第三球座標(S815)，因為具有較小半徑值的輸出座標離車體較近，有較大的可能性為要追蹤的目標。 if both A and B have values, and

and

The difference is less than or equal to the threshold value (S814), according to

and

The arithmetic mean of and

and

The smaller one obtains the third spherical coordinate (S815), because the output coordinate with the smaller radius value is closer to the vehicle body, and has a greater possibility of being the target to be tracked.

If one of A and B is a null value (S816), that is, when the target is not tracked from one of the color image depth camera 111 and the optical radar 112, the value of A and B is taken as the third spherical coordinate ( S817).

If both A and B are null (S818), the third spherical coordinate is not output (S819), and the target tracking program is re-executed (S820).

To sum up, the self-propelled vehicle following system and the self-propelled vehicle following method of the present invention use the color image depth camera disposed on the vehicle body to obtain color images, and use the optical radar disposed on the vehicle body to obtain two-dimensional information, and obtain color images. The three-dimensional coordinates of the target in the image and the two-dimensional coordinates of the target in the two-dimensional information, and the three-dimensional coordinates and the two-dimensional coordinates are converted into the first spherical coordinates and the second spherical coordinates, and then according to the first spherical coordinates and the second spherical coordinates A third spherical coordinate of the target is generated. Finally, the vehicle body controller of the vehicle body can control the vehicle body to follow the target according to the third spherical coordinate. The self-propelled vehicle following system and the self-propelled vehicle following method of the present invention use a color image depth camera to increase the feature that the target can be identified to make up for the problem of insufficient point cloud image features of the optical radar, so as to detect more accurately in a complex environment Target. In addition, the system can also memorize the characteristics of the target and continue to follow even if the target is obscured.

Although the present invention has been disclosed above by the embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, The protection scope of the present invention shall be determined by the scope of the appended patent application.

,

)」 802:「B：基於光學雷達判斷的第二球座標(

,

)」 S810~S820:步驟 110: Data Collection Unit 111: Color Image Depth Camera 112: Optical Radar 120: Data Operation Unit 121: Person Detection Module 122: Image Target Locking Module 123: Image Target Tracking Module 124: Optical Radar Target Locking Module 125 : Optical radar target tracking module 126 : Coordinate fusion module 127 : Control command output module 130 : Vehicle body control unit 131 : Vehicle body control robot operating system 210 : Color image 220 : Coordinates and frame shadows of human body bounding frame 230 : Deep learning object detection model 231: Classifier 232: Convolution layer 233: Person detection function 310: Target image feature 320: Target bounding box S301~S303: Step 410: Pixel depth information 420: 3D coordinates S401~S403: Step 510: Two-dimensional information 520: Feature tracker S501~S502: Step 610: Two-dimensional coordinates S601~S602: Step 710: Vehicle body coordinate system 720: Camera coordinate system 730: Optical radar coordinate system 801: "A: based on color The first spherical coordinate for image judgment (

,

)" 802: "B: The second spherical coordinate determined based on the lidar (

,

)” S810~S820: Steps

FIG. 1 is a block diagram of a bicycle following system according to an embodiment of the present invention. FIG. 2 is a block diagram of a person detection module according to an embodiment of the present invention. 3 is a block diagram of an image target locking module according to an embodiment of the present invention. 4 is a block diagram of an image target tracking module according to an embodiment of the present invention. FIG. 5 is a block diagram of an optical radar target locking module according to an embodiment of the present invention. FIG. 6 is a block diagram of an optical radar target tracking module according to an embodiment of the present invention. FIG. 7 is a schematic diagram of the coordinate relationship between a camera coordinate system, an optical radar coordinate system, and a vehicle body coordinate system according to an embodiment of the present invention. FIG. 8 is a flowchart of a third spherical coordinate of an output target of a control command output module according to an embodiment of the present invention.

110: Data Collection Unit

111: Color Image Depth Camera

112: LiDAR

120: Data operation unit

121: Personnel Detection Module

122: Image Target Locking Module

123: Image Target Tracking Module

124: LiDAR target locking module

125: LiDAR target tracking module

126: Coordinate fusion module

127: Control command output module

130: Body control unit

131: Body Control Robot Operating System

Claims (20)

A self-propelled vehicle following system includes: a data collection unit, including: a color image depth camera, which obtains a color image; and an optical radar, which obtains two-dimensional information; and a data operation unit, including: a person detection module, which receives After the color image of the color image depth camera is analyzed, the color image is analyzed by the deep learning object detection model, and the coordinates of at least one personnel bounding box and at least one frame image are output; and the image target locking module receives the The coordinates of the at least one personnel bounding box and the image in the at least one box are used to lock the at least one bounding box with the largest area as the first target bounding box, and extract the image from the first target bounding box The feature is stored as a target image feature, and the first target bounding frame and the target image feature are output; the image target tracking module obtains the three-dimensional coordinates of the target in the color image by the image target locking module; The optical radar target tracking module obtains the two-dimensional coordinates of the target in the two-dimensional information; the coordinate fusion module converts the three-dimensional coordinates into the first spherical coordinates of the spherical coordinate system, and converts the two-dimensional coordinates into the first spherical coordinates of the spherical coordinate system. Converting into a second spherical coordinate of the spherical coordinate system; and a control command output module, according to the depth camera based on the color image The obtained first spherical coordinate and the second spherical coordinate obtained based on the optical radar generate a third spherical coordinate of the target; and a vehicle body control unit, including a vehicle body controller, according to the third spherical coordinate Three spherical coordinates control the body to follow the target. The self-propelled vehicle following system according to claim 1, further comprising: an optical radar target locking module, which receives the two-dimensional information from the optical radar, and is used to lock the bounding frame of the predetermined coordinates for the second target. bounding box, initialize the feature tracker and output the feature tracker. The self-propelled vehicle following system according to claim 1, wherein the target tracking module executes an image target tracking program and a point cloud target tracking program, and the image target tracking program is based on the coordinates of the at least one personnel bounding box, The at least one frame image, the first target bounding frame, and the target image feature, calculate a coverage ratio of the at least one personnel bounding frame and the target bounding frame, and calculate the at least one The frame image and the target image feature are matched with the whole body image feature of the person, and the 3D coordinates are output according to the pixel depth information of the color image, and the point cloud target tracking program is tracked according to the 2D information and the feature. to track the target and locate the two-dimensional coordinates of the target. The self-propelled vehicle following system according to claim 1, wherein the coordinate fusion module discards the coordinate value corresponding to the height in the three-dimensional coordinate to convert the three-dimensional coordinate into the first coordinate of the vehicle body coordinate system and convert the coordinate value of the three-dimensional coordinate into the first coordinate of the vehicle body coordinate system The first coordinates are converted to the first spherical coordinates, and converting the two-dimensional coordinates into second coordinates of the vehicle body coordinate system and converting the second coordinates into the second spherical coordinates. The self-propelled vehicle following system according to claim 1, wherein when both the first spherical coordinate and the second spherical coordinate have values and the angle difference between the first spherical coordinate and the second spherical coordinate is greater than one When the threshold value is set, the first spherical coordinate is taken as the third spherical coordinate; and when both the first spherical coordinate and the second spherical coordinate have values and the first spherical coordinate and the second spherical coordinate have values When the angle difference of the coordinates is not greater than the threshold value, according to the arithmetic mean of the angles of the first spherical coordinates and the second spherical coordinates and the smaller radius of the first spherical coordinates and the second spherical coordinates value to obtain the third spherical coordinate. The self-propelled vehicle following system according to claim 5, wherein when one of the first spherical coordinate and the second spherical coordinate is a null value, the first spherical coordinate and the second spherical coordinate are taken The value of is the third spherical coordinate. The self-propelled vehicle following system according to claim 6, wherein when both the first spherical coordinate and the second spherical coordinate are null values, a target tracking procedure is re-executed. The self-propelled vehicle following system of claim 1, wherein the two-dimensional information includes a point cloud image. A method for following a bicycle, comprising: obtaining a color image by a color image depth camera, and obtaining two-dimensional information by using an optical radar; after receiving the color image from the color image depth camera, converting the color image Through deep learning object detection model analysis, output at least one person Coordinates of the bounding box and at least one frame image; receiving the coordinates of the at least one personnel bounding frame and the at least one frame image, so as to lock the at least one bounding frame with the largest area to delimit the first target frame, and extract image features from the first target bounding box and store them as target image features, output the first target bounding box and the target image features; The three-dimensional coordinates of the target in the color image are obtained from the target image feature, and the two-dimensional coordinates of the target in the two-dimensional information are obtained from the two-dimensional information; the three-dimensional coordinates are converted into spherical coordinates. first spherical coordinates, and convert the two-dimensional coordinates into second spherical coordinates of the spherical coordinate system; according to the first spherical coordinates obtained based on the color image depth camera and the the second spherical coordinate to generate the third spherical coordinate of the target; and the vehicle body controller controls the vehicle body according to the third spherical coordinate to follow the target. The self-propelled vehicle following method according to claim 9, further comprising: receiving the two-dimensional information from the optical radar, locking a bounding frame of predetermined coordinates as a second target bounding frame, and initializing a feature tracker and output the feature tracker. The self-propelled vehicle following method according to claim 9, further comprising: executing an image target tracking program and a point cloud target tracking program, wherein the image target tracking program is based on the coordinates of the at least one personnel bounding box, the at least one frame image, the first target bounding box, the target image feature feature, calculate a coverage ratio of the at least one person bounding box and the target bounding box, and match the at least one frame image and the target image feature with the human whole body image feature, and according to the color The pixel depth information of the image outputs the 3D coordinates, and the point cloud object tracking program tracks the object and locates the 2D coordinates of the object according to the 2D information and the feature tracker. The self-propelled vehicle following method according to claim 9, further comprising: discarding the coordinate value corresponding to the height in the three-dimensional coordinate to convert the three-dimensional coordinate into the first coordinate of the vehicle body coordinate system and converting the first coordinate Converting to the first spherical coordinates, converting the two-dimensional coordinates to the second coordinates of the vehicle body coordinate system, and converting the second coordinates to the second spherical coordinates. The self-propelled vehicle following method according to claim 9, wherein when both the first spherical coordinate and the second spherical coordinate have values and the angle difference between the first spherical coordinate and the second spherical coordinate is greater than one When the threshold value is set, the first spherical coordinate is taken as the third spherical coordinate; and when both the first spherical coordinate and the second spherical coordinate have values and the first spherical coordinate and the second spherical coordinate have values When the angle difference of the coordinates is not greater than the threshold value, according to the arithmetic mean of the angles of the first spherical coordinates and the second spherical coordinates and the smaller radius of the first spherical coordinates and the second spherical coordinates value to obtain the third spherical coordinate. The self-propelled vehicle following method according to claim 13, wherein when one of the first spherical coordinate and the second spherical coordinate is a null value, the first spherical coordinate and the second spherical coordinate are taken The value of is the third spherical coordinate. The self-propelled vehicle following method according to claim 14, wherein when both the first spherical coordinate and the second spherical coordinate are null values, a target tracking procedure is re-executed. The self-propelled vehicle following method according to claim 9, wherein the two-dimensional information includes a point cloud image. A self-propelled vehicle following system includes: a data collection unit, including: a color image depth camera to obtain a color image; and an optical radar to obtain two-dimensional information; and a data operation unit, including: an image target tracking module, to obtain the color image The three-dimensional coordinates of the target in the image; the optical radar target tracking module obtains the two-dimensional coordinates of the target in the two-dimensional information; the coordinate fusion module converts the three-dimensional coordinates into the first spherical coordinates of the spherical coordinate system , and convert the two-dimensional coordinates into the second spherical coordinates of the spherical coordinate system, wherein the coordinate fusion module discards the coordinate value corresponding to the height in the three-dimensional coordinates to convert the three-dimensional coordinates into vehicle body coordinates The first coordinate of the vehicle body coordinate system, and the first coordinate is converted into the first spherical coordinate, the two-dimensional coordinate is converted into the second coordinate of the vehicle body coordinate system, and the second coordinate is converted into the second spherical coordinates; and a control command output module, according to the first spherical coordinates obtained based on the color image depth camera and the second spherical coordinates obtained based on the optical radar the target to generate a third spherical coordinate of the target; and a vehicle body control unit, including a vehicle body controller, which controls the vehicle body to follow the target according to the third spherical coordinate. A self-propelled vehicle following system includes: a data collection unit, including: a color image depth camera to obtain a color image; and an optical radar to obtain two-dimensional information; and a data operation unit, including: an image target tracking module, to obtain the color image The three-dimensional coordinates of the target in the image; the optical radar target tracking module obtains the two-dimensional coordinates of the target in the two-dimensional information; the coordinate fusion module converts the three-dimensional coordinates into the first spherical coordinates of the spherical coordinate system , and convert the two-dimensional coordinates into the second spherical coordinates of the spherical coordinate system; and a control command output module, according to the first spherical coordinates obtained based on the color image depth camera and based on the optical radar the obtained second spherical coordinate to generate the third spherical coordinate of the target, wherein when the first spherical coordinate and the second spherical coordinate both have values and the first spherical coordinate and the second spherical coordinate When the angle difference between the spherical coordinates is greater than a threshold value, the first spherical coordinate is taken as the third spherical coordinate, and when both the first spherical coordinate and the second spherical coordinate have values and the first spherical coordinate When the angle difference between the spherical coordinate and the second spherical coordinate is not greater than the threshold value, according to the The third spherical coordinate is obtained from the arithmetic mean value of the angle between the first spherical coordinate and the second spherical coordinate and the smaller radius value of the first spherical coordinate and the second spherical coordinate; and the vehicle body control unit, A vehicle body controller is included to control the vehicle body to follow the target according to the third spherical coordinate. A method for following a self-propelled vehicle, comprising: obtaining a color image by a color image depth camera, and obtaining two-dimensional information by using an optical radar; obtaining three-dimensional coordinates of a target in the color image, and obtaining the two-dimensional information. The two-dimensional coordinates of the target; convert the three-dimensional coordinates into the first spherical coordinates of the spherical coordinate system, and convert the two-dimensional coordinates into the second spherical coordinates of the spherical coordinate system, wherein the three-dimensional coordinates are discarded to convert the three-dimensional coordinates to the first coordinates of the vehicle body coordinate system, convert the first coordinates to the first spherical coordinates, and convert the two-dimensional coordinates to the the second coordinates of the vehicle body coordinate system, and convert the second coordinates into the second spherical coordinates; according to the first spherical coordinates obtained based on the color image depth camera and the first spherical coordinates obtained based on the optical radar the second spherical coordinate to generate the third spherical coordinate of the target; and the vehicle body controller controls the vehicle body according to the third spherical coordinate to follow the target. A method for following a bicycle, comprising: obtaining a color image by a color image depth camera, and obtaining a color image by an optical radar, obtaining two-dimensional information; obtaining the three-dimensional coordinates of the target in the color image, and obtaining the two-dimensional coordinates of the target in the two-dimensional information; converting the three-dimensional coordinates into the first spherical coordinates of a spherical coordinate system, and converting the two-dimensional coordinates into the second spherical coordinates of the spherical coordinate system; according to the first spherical coordinates obtained based on the color image depth camera and the second spherical coordinates obtained based on the optical radar, to generate the third spherical coordinate of the target, wherein when both the first spherical coordinate and the second spherical coordinate have values and the angle difference between the first spherical coordinate and the second spherical coordinate is greater than a threshold value , take the first spherical coordinate as the third spherical coordinate, and when both the first spherical coordinate and the second spherical coordinate have values and the difference between the first spherical coordinate and the second spherical coordinate When the angle difference is not greater than the threshold value, it is obtained according to the arithmetic mean of the angles of the first spherical coordinate and the second spherical coordinate and the smaller radius value of the first spherical coordinate and the second spherical coordinate the third spherical coordinate; and controlling the vehicle body according to the third spherical coordinate by the vehicle body controller to follow the target.

Images