EP4139891A1

EP4139891A1 - Method and apparatus for vehicle length estimation

Info

Publication number: EP4139891A1
Application number: EP20937268.9A
Authority: EP
Inventors: Jie Zhao; Jie MIN; Xiao Dong Xu
Original assignee: Siemens Ltd China
Current assignee: Siemens Ltd China
Priority date: 2020-05-29
Filing date: 2020-05-29
Publication date: 2023-03-01
Also published as: EP4139891A4; CN115668297A; WO2021237750A1

Abstract

A method, apparatus, system and computer-readable medium for vehicle length estimation are presented. A method includes: acquiring (S101) an image (30) from a camera (40) which monitors a roadside; obtaining (S102) a bounding box of a vehicle via object recognition; estimating (S103) length of the vehicle using a curve function d=f (x, y), wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image (30); the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.

Description

Method and apparatus for vehicle length estimation

Technical Field

The present invention relates to techniques of image perception, and more particularly to a method, apparatus and computer-readable storage medium for vehicle length estimation.
Background Art
Image perception technology for intelligent traffic system (ITS) is an important research direction, with which vehicle’s length, speed, etc. can be calculated, for regulation of driving or parking vehicles.
Image perception technology usually bases on images captured by cameras. For city, most of monitoring cameras are dome cameras (as shown in FIG. 1) , as they can provide a wide range of view. However, dome cameras can also cause serious distortion in image. For example, the straight line seems to be bent in the image as shown in FIG. 2. In this case, it is tough to calculate the accurate distance from the image.
To calculate the distance in image, the calibration is a necessary step. For traditional method, a trapezoid is calibrated on the image, and each side of the trapezoid is measured in real word. In this way, the distance can be calculated roughly, as shown in FIG. 3. Usually this method is applied to estimate the speed of the vehicle roughly. However, it is assumed that there is no distortion in the image with this method, and it can only provide a rough estimation. This method will cause serious error if there is obvious distortion in image.
Summary of the Invention
In this disclosure, efficient solutions based on curve fitting is proposed to calculate the vehicle length for images with serious distortion.
Embodiments of the present disclosure include methods, apparatuses for object recognition and methods, apparatuses for vehicle length estimation.
According to a first aspect of the present disclosure, a method for vehicle length estimation is presented. The method includes following steps:
- acquiring an image from a camera which monitors a roadside;
- obtaining a bounding box of a vehicle via object recognition;
- estimating length of the vehicle using a curve function d=f (x, y) , wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image ; the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.
According to a second aspect of the present disclosure, an apparatus for vehicle length estimation is presented. The apparatus includes:
- an image acquisition module, configured to acquire an image from a camera which monitors a roadside;
- a vehicle detection module, configure to obtain a bounding box of a vehicle via object recognition;
- an estimation module, configured to estimate length of the vehicle using a curve function d=f (x, y) , wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image ; the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.
According to a third aspect of the present disclosure, an apparatus for vehicle length estimation is presented. The apparatus includes at least one processor; at least one memory, coupled to the at least one processor, configured to execute method according to the first aspect.
According to a fourth aspect of the present disclosure, a computer-readable medium for vehicle length estimation is presented. The computer-readable medium stores computer-executable instructions, wherein the computer-executable instructions when executed cause at least one processor to execute method according to the first aspect.
With solutions provided in the present disclosure, better accuracy can be achieved via preset calibration points, as well as regression algorithm. The proposed novel calibration solution can work well in dome cameras which have serious distortion, and the vehicle length estimation solution can work well for both top view and side view cases, which can reduce the error caused by vehicle height in top view.
Optionally, the view of the camera is a top view, and when estimating length of the vehicle using curve function d=f (x, y) , length of the vehicle can be estimated using curve functions d=f (y) and h=H (d) , wherein h is the pixel length of the vehicle in the image .
Optionally, when estimating length of the vehicle using curve functions d=f (y) and h=H (d) , first, the end point of the vehicle can be calculated based the curve function h=H (d) ; and then the length of the vehicle can be calculated as distance between the end point and the start point of the vehicle.
Optionally, the view of the camera is a side view, and when estimating length of the vehicle using curve function d=f (x, y) , length of the vehicle can be estimated using curve functions d=f (x) .
Optionally, when estimating length of the vehicle using curve functions d=f (x) , start point of the vehicle can be calculated based the curve function d=f (x) of an outer line, wherein the outer line is parallel to the roadside and passing through the start point of the vehicle; and end point of the vehicle can be calculated based the curve function d=f (x) of an inner line, wherein the inner line is parallel to the roadside and passing through the end point of the vehicle; then the length of the vehicle can be calculated as distance between the end point and the start point of the vehicle.
Considering characteristics of top view and side view, curve function d=f (x, y) can be simplified as d=f (x) for side view case and d=f (y) for top view case, which can reduce algorithm complexity and simultaneously achieve relatively accurate fitting result.

Brief Description of the Drawings

The above mentioned attributes and other features and advantages of the present technique and the manner of attaining them will become more apparent and the present technique itself will be better understood by reference to the following description of embodiments of the present technique taken in conjunction with the accompanying drawings, wherein:
FIG. 1 depicts a dome camera.
FIG. 2 depicts serious distortion in dome camera.
FIG. 3 depicts trapezoid calibration.
FIG. 4 depicts a block diagram of an apparatus for vehicle length estimation in accordance with one embodiment of the present disclosure.
FIG. 5 depicts calibration methodology deployed in accordance with one embodiment of the present disclosure.
FIG. 6 depicts curve regression in accordance with one embodiment of the present disclosure.
FIG. 7 depicts vehicle length estimation for top view in accordance with one embodiment of the present disclosure.
FIG. 8 depicts vehicle length estimation for side view in accordance with one embodiment of the present disclosure.
FIG. 9 depicts a flow diagram of a method for vehicle length estimation in accordance with one embodiment of the present disclosure.
FIG. 10 depicts an example procedure in accordance with one embodiment of the present disclosure.
FIG. 11 depicts the vehicle length estimation result in accordance with one embodiment of the present disclosure.
Reference Numbers:
10, an apparatus for vehicle length estimation
101, at least one memory
102, at least one processor
103, a communication module
20, a vehicle length estimation program
201, an image acquisition module
202, a vehicle detection module
203, an estimation module
30, image taken by a camera 40
40, a camera monitoring a roadside
50, server executing calibration process
60, coefficients of curve functions d=f (y) and h=H (d) for a top view case
61, coefficients of curve function d=f (x) for a side view case
100, a method for vehicle length estimation
S101～S103, steps of method 100
Detailed Description of Example Embodiments
Hereinafter, above-mentioned and other features of the present technique are described in detail. Various embodiments are described with reference to the drawing, where like reference numerals are used to refer to like elements throughout. In the following description, for purpose of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more embodiments. It may be noted that the illustrated embodiments are intended to explain, and not to limit the invention. It may be evident that such embodiments may be practiced without these specific details.
When introducing elements of various embodiments of the present disclosure, the articles “a” , “an” , “the” and “said” are intended to mean that there are one or more of the elements. The terms “comprising” , “including” and “having” are intended to be inclusive and mean that there may be additional elements other than the listed elements.
Vehicle length estimation solutions are proposed in this disclosure, which can be used to calculate the vehicle length at roadside parking space. The first step is calibration which can be based on curving fitting. After calibration, the curve function can be calculated. The vehicle bounding box can be obtained with object detection technology. The vehicle length can be estimated according to the camera views: top view and side view. Now the present disclosure will be described hereinafter in details by referring to FIG. 4 to FIG. 11.
Calibration
Now referring to FIG. 5, for a specific dome camera, for calibration purpose, calibration points P _i , i=0, 1, 2…, can be firstly deployed along roadside which the dome camera is monitoring, P ₀, as the starting point, can be called as “Reference point” . A ruler can be placed along the roadside, beside these calibration points, to mark real distance d _i of a specific calibration point P _i from the reference point P ₀. For simplicity of calculation, for the reference point P ₀, d=0. These reference points can be deployed evenly along the roadside, or otherwise, as long as there is a ruler which can measure real distance of a specific calibration point from the reference point.
For a specific calibration point P _i (d _i, h _i, x _i, y _i) , d _i is the real distance from reference point P ₀ to P _i in meter; h _i is the pixel distance, which can be imagined as the pole with fixed length standing on the calibration point. The x _i and y _i are the pixel coordinates in image coordinate system which means it is located in the x _ith column and y _ith row.
For vehicle’s length calculation, with the head point and the rear point known, the real length of the vehicle can be calculated. However, with distortion brought in by the dome camera, length of a vehicle can’ t be calculated directly via pixel coordinate of the rear and head of the vehicle in image. If we can find relationship between real position and pixel coordinate, then distortion can be eliminated during calculation of vehicle length. Here, curve function d=f (x, y) denoting functional relationship of real distance of a point to reference point with pixel coordinates x, y of the point, and it will be worked out via calibration.
For each calibration point P _i, coordinates d _i can be got easily with the ruler; x _i, y _i can also be got easily from image captured. With multiple calibration points and corresponding measured d _i, x _i, y _i, the curve function f (x, y) can be obtained.
Here, during the process of obtaining f (x, y) , we use 1.5m for height calibration, as most of the small cars are 1.5m high. Then h _i can be assigned, for example, 22 pixels, corresponding to 1.5m in the real distance.
For the case in FIG. 5 (top view) , the parking space extends in y direction which means the coordinate y _i contributes much more than x _i. Under this assumption, the curve function can be simplified as d=f (y) . If the polynomial of degree 3 is applied to regress the curve function, the function can be written as:
d=f (y) =P ₁·y ³+P ₂·y ²+P ₃·y+P ₄ (1)
d and y can be obtained directly from the calibration point. The regression curve can be calculated as FIG. 6. The coefficients calculated out for the example shown in FIG. 6 are:
P ₁=1.293e-08
P ₂=4.647e-05
P ₃=0.06438e-08
P ₄=36.12 (1’ )
Similarly, for the case in FIG. 8 (side view) , the parking space extends in x direction which means the coordinate x _i contributes much more than y _i. Under this assumption, the curve function can be written as:
d=f (x) =P ₁·x ³+P ₂·x ²+P ₃·x+P ₄ (2)
Usually, there are two kinds of views for a dome camera, they are top view, as shown in FIG. 7, and side view, as shown in FIG. 8. For the top view case, the height of the bounding box contains both height and length component of the vehicle. Generally, h changes with d, the larger is d, the smaller is h. however, the functional relationship between h and d is not linear and is also influenced by distortion introduced by dome camera. So here, With the same curve fitting method, the function between h and d can be obtained as:
h=H (d) =Q ₁·d ³+Q ₂·d ²+Q ₃·d+Q ₄ (3)
To be mentioned that, the functions got above are specific for a dome camera. For different cameras monitor different streets, the installation locations, angles are different, then the coefficients Q ₁, Q ₂, Q ₃, Q ₄ are different accordingly.
Once functions are got via above curve regression, they can be pre-stored in storage memories the dome camera is connected to, and when length estimation is to be executed by processors, these functions can be called for calculation.
To be mentioned that, here a dome camera is only an example for cameras with distortion on images. The solutions provided in the present disclosure can also be applicable to any kind of cameras with image distortion problems.
Vehicle length estimation
Images will be taken via the dome camera, then vehicle length can be estimated based on the taken images and above functions got during calibration process.
For top view and side view cases, different calculation methods can be applied.
Top view
For vehicle length estimation from top view, it cannot calculate the distance between line a and line c from bounding box directly (shown in FIG. 7) , as the height of the bounding box contains both height and length component of the vehicle. The distance between line a and line c is the height of the bounding box, and it is much longer than the actual vehicle length. The distance between line a and line b can be regarded as the actual vehicle length which needs to figure out, while the distance between line b and line c is caused by the vehicle height which should be removed. Therefore, the task is to find out the y coordinate of line a and line b, then the vehicle length is L=d _b-d _a=f (y _b) -f (y _a) .
To determine the position or y coordinate of line b, it will make use of the above mentioned height calibration information h=f (d) , where d=d _c=f (y _c) (for d _b is unknown, there the d _c of is used, of course, d _a can also be used here) . In the formula h=H (d) , h is the pixel length of a vehicle in image, so the y coordinate of line b can be calculated.
During the calibration process, we use 1.5m for height calibration, as most of the small cars are 1.5m high. However, for different vehicle types the vehicle heights can be different. For example, the height of a bus can be about 2.5m, therefore the pixel height of bus in image is
If the vehicle bounding box is (x _i, y _i, w _i, h _i) (x _i, y _i are the pixel coordinates of left top corner of the bounding box, w _i is the pixel width of the bounding box and h _i is the pixel height of the bounding box) , the distance function is d=f (y) =P ₁ y ³+P ₂·y ²+P ₃·y+P ₄ (1) , the height function is h=H (d) =Q ₁·d ³+Q ₂. d ²+Q ₃·d+Q ₄ (3) , taking all of these into consideration:
d _start=f (y _i+h _i) , the vehicle starting position (4)
the vehicle height in pixel (5)
Here, h _type takes into vehicle type, which can be calculated out based on a car’s common value, that is 1.5 meters.
d _end=f (y _i+h _type) , the vehicle end position (6)
L=|d _end-d _start|, the vehicle length (7)
Then, the length of vehicle L can be calculated out, eliminating distortion brought in by dome camera and influence of height by the top view.
Side view
For vehicle length estimation from side view, it is relatively simple in comparison to top view. In previous conclusion, the curve function for side view can be written as formula (2) . In this case, two regression lines are needed, the outer line f ₁ (x) and the inner line f ₂ (x) , which are two lines in parallel with the roadside, one passes through the start point of the vehicle and the other passes through the end point of the vehicle. If the vehicle bounding box is (x _i, y _i, w _i, h _i) , then the starting position is the crossing point P ₁, and the end position is P ₂. Therefore, the vehicle length can be calculated as:
d _start=f ₁ (x _i+h _i) , the vehicle starting position (8)
d _end=f ₂ (x _i) , the vehicle end position (9)
L=|d _end-d _start|, the vehicle length (10)
FIG. 4 depicts a block diagrams of an apparatus in accordance with one embodiment of the present disclosure. The apparatus 10 for vehicle length estimation presented in the present disclosure can be implemented as a network of computer processors, to execute following method 100 for vehicle length estimation presented in the present disclosure. the apparatus 10 can also be a single computer, as shown in FIG. 4, including at least one memory 101, which includes computer-readable medium, such as a random access memory (RAM) . The apparatus 10 also includes at least one processor 102, coupled with the at least one memory 101. Computer-executable instructions are stored in the at least one memory 101, and when executed by the at least one processor 102, can cause the at least one processor 102 to perform the steps described herein. The at least one processor 102 may include a microprocessor, an application specific integrated circuit (ASIC) , a digital signal processor (DSP) , a central processing unit (CPU) , a graphics processing unit (GPU) , state machines, etc. embodiments of computer-readable medium include, but not limited to a floppy disk, CD-ROM, magnetic disk, memory chip, ROM, RAM, an ASIC, a configured processor, all optical media, all magnetic tape or other magnetic media, or any other medium from which a computer processor can read instructions. Also, various other forms of computer-readable medium may transmit or carry instructions to a computer, including a router, private or public network, or other transmission device or channel, both wired and wireless. The instructions may include code from any computer-programming language, including, for example, C, C++, C#, Visual Basic, Java, and JavaScript.
The at least one memory 101 shown in FIG. 4 can contain a vehicle length estimation program 20, when executed by the at least one processor 102, causing the at least one processor 102 to execute the method 100 for vehicle length estimation presented in the present disclosure. Images 30 of vehicles can also be stored in the at least one memory 101. These data can be received via a communication module 103 of the apparatus 10. A dome camera 40 can be connected to the apparatus 10 via the communication module 102. The dome camera 40 can take images of the roadside it monitors and send taken images to the apparatus 10 for further processing.
The above mentioned calibration process can be executed on server 50, which output coefficients of the curve functions d=f (y) in equation (1) and h=H (d) in equation (3) to apparatus 10, received via the communication module 103 and stored in the at least one memory 101 as coefficients 60 for a top view case. Similarly, the server 50 can output coefficients of the functions d=f (x) in equation (2) to apparatus 10, received via the communication module 103 and stored in the at least one memory 101 as coefficients 61 for a side view case.
However, the calibration process can also be executed on the apparatus 10, which depends on device configuration and processing competence. In such a case, the calibration program can be part of the vehicle length estimation program 20 and can be pre-stored in the at least memory 101.
The vehicle length estimation program 20 can include:
- an image acquisition module 201, configured to acquire an image 30 from a camera 40 which monitors a roadside;
- a vehicle detection module 202, configure to obtain a bounding box of a vehicle via object recognition;
- an estimation module 203, configured to estimate length of the vehicle using a curve function d=f (x, y) , wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image 30; the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.
Optionally, if the view of the camera 40 is a top view, then the estimation module 203 can be further configured to: estimate length of the vehicle using curve functions d=f (y) and h=H (d) , wherein h is the pixel length of the vehicle in the image 30. And optionally, the estimation module 203 can calculate the end point of the vehicle based the curve function h=H (d) ; and calculate the length of the vehicle as distance between the end point and the start point of the vehicle.
Optionally, if the view of the camera 40 is a side view, then the estimation module 203 can be further configured to: estimate length of the vehicle using curve functions d=f (x) . And optionally, the estimation module 203 can calculate start point of the vehicle based the curve function d=f (x) of an outer line, wherein the outer line is parallel to the roadside and passing through the start point of the vehicle; calculate end point of the vehicle based the curve function d=f (x) of an inner line, wherein the inner line is parallel to the roadside and passing through the end point of the vehicle; and then calculate the length of the vehicle as distance between the end point and the start point of the vehicle.
Details of each module’s processing can be referred to above mentioned “calibration” process and “vehicle length estimation” process.
Although the image acquisition module 201, the vehicle detection module 202, and the estimation module 203 are described above as software modules of the vehicle length estimation program 20. Also, they can be implemented via hardware, such as ASIC chips. They can be integrated into one chip, or separately implemented and electrically connected.
It should be mentioned that the present disclosure may include apparatuses having different architecture than shown in FIG. 4. The architecture above is merely exemplary and used to explain the exemplary method 100 shown in FIG. 9.
Various methods in accordance with the present disclosure may be carried out. One exemplary method 100 according to the present disclosure includes following steps:
S101: acquiring an image 30 from a camera 40 which monitors a roadside;
S102: obtaining a bounding box of a vehicle via object recognition;
S103: estimating length of the vehicle using a curve function d=f (x, y) , wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image 30; the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.
Optionally, if the view of the camera 40 is a top view, in the step S103, length of the vehicle can be calculated using curve functions d=f (y) and h=H (d) , wherein h is the pixel length of the vehicle in the image 30. And furthermore, firstly, the end point of the vehicle can be calculated based the curve function h=H (d) , then the length of the vehicle can be calculated as distance between the end point and the start point of the vehicle.
Optionally, if the view of the camera 40 is a side view, in the step S103, length of the vehicle can be estimated using curve functions d=f (x) . Furthermore, first start point of the vehicle can be calculated based the curve function d=f (x) of an outer line, wherein the outer line is parallel to the roadside and passing through the start point of the vehicle; and end point of the vehicle can be calculated based the curve function d=f (x) of an inner line, wherein the inner line is parallel to the roadside and passing through the end point of the vehicle; then the length of the vehicle can be calculated as distance between the end point and the start point of the vehicle.
Details of each step can be referred to above mentioned “calibration” process and “vehicle length estimation” process.
Now referring to FIG. 10 and FIG. 11, an example for process of calibration and vehicle length estimation, and corresponding application and result will be described.
According to FIG. 10, firstly, calibration is executed based on curving fitting. After calibration, the curve regression function (s) (d=f (x, y) , h=H (d) ) can be calculated. The vehicle’s bounding box can be obtained with object detection, and with the input of bounding box and vehicle type, the vehicle length can be estimated according to the camera view, i.e., top view and side view.
The performance and result can be seen in FIG. 11. The location of the target vehicle can be seen on the map (referring to the left second picture on top) , also detailed information of the vehicle and its parking action can be seen on the left first picture on top, the vehicle length is 5 meter in this case. On bottom are four pictures depicting the car’s parking action. On the right picture, each point corresponds to a parking lot, and different colors indicate different parking statuses.
A computer-readable medium is also provided in the present disclosure, storing computer-executable instructions, which upon execution by a computer, enables the computer to execute any of the methods presented in this disclosure.
A computer program, which is being executed by at least one processor and performs any of the methods presented in this disclosure.
With the solutions provided in the present disclosure, better accuracy can be achieved via preset calibration points, as well as regression algorithm. The proposed novel calibration solution can work well in dome cameras which have serious distortion, and the vehicle length estimation solution can work well for both top view and side view cases, which can reduce the error caused by vehicle height in top view.
Considering characteristics of top view and side view, curve function d=f (x, y) can be simplified as d=f (x) for side view case and d=f (y) for top view case, which can reduce algorithm complexity and simultaneously achieve relatively accurate fitting result.
While the present technique has been described in detail with reference to certain embodiments, it should be appreciated that the present technique is not limited to those precise embodiments. Rather, in view of the present disclosure which describes exemplary modes for practicing the invention, many modifications and variations would present themselves, to those skilled in the art without departing from the scope and spirit of this invention. The scope of the invention is, therefore, indicated by the following claims rather than by the foregoing description. All changes, modifications, and variations coming within the meaning and range of equivalency of the claims are to be considered within their scope.

Claims

A method (100) for vehicle length estimation, comprising:

- acquiring (S101) an image (30) from a camera (40) which monitors a roadside;

- obtaining (S102) a bounding box of a vehicle via object recognition;

- estimating (S103) length of the vehicle using a curve function d=f (x, y) , wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image (30) ; the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.
the method (100) according to claim 1, wherein the view of the camera (40) is a top view, and estimating (S103) length of the vehicle using curve function d=f (x, y) further comprises:

- estimating (S103) length of the vehicle using curve functions d=f (y) and h=H(d) , wherein h is the pixel length of the vehicle in the image (30) .
the method (100) according to claim 2, wherein estimating (S103) length of the vehicle using curve functions d=f (y) and h=H (d) further comprises:

- calculating the end point of the vehicle based the curve function h=H (d) ;

- calculating the length of the vehicle as distance between the end point and the start point of the vehicle.
the method (100) according to claim 2, wherein the view of the camera (40) is a side view, and estimating (S103) length of the vehicle using curve function d=f (x, y) further comprises:

- estimating (S103) length of the vehicle using curve functions d=f (x) .
the method (100) according to claim 4, wherein estimating (S103) length of the vehicle using curve functions d=f (x) , further comprises:

- calculating start point of the vehicle based the curve function d=f (x) of an outer line, wherein the outer line is parallel to the roadside and passing through the start point of the vehicle;

- calculating end point of the vehicle based the curve function d=f (x) of an inner line, wherein the inner line is parallel to the roadside and passing through the end point of the vehicle;

- calculating the length of the vehicle as distance between the end point and the start point of the vehicle.
An apparatus (10) for vehicle length estimation, comprising:

- an image acquisition module (201) , configured to acquire an image (30) from a camera (40) which monitors a roadside;

- a vehicle detection module (202) , configure to obtain a bounding box of a vehicle via object recognition;

- an estimation module (203) , configured to estimate length of the vehicle using a curve function d=f (x, y) , wherein d is the real distance from a specific point to a reference point at the roadside, x and y are pixel coordinates of the specific point on the image (30) ; the curve function is pre-calculated based on at least two preset calibration points including the reference point along the roadside.
the apparatus (10) according to claim 6, wherein the view of the camera (40) is a top view, and the estimation module (203) is further configured to:

- estimate length of the vehicle using curve functions d=f (y) and h=H (d) , wherein h is the pixel length of the vehicle in the image (30) .
the apparatus (10) according to claim 7, wherein the estimation module (203) is further configured to:

- calculate the end point of the vehicle based the curve function h=H (d) ;

- calculate the length of the vehicle as distance between the end point and the start point of the vehicle.
the apparatus (10) according to claim 6, wherein the view of the camera (40) is a side view, and the estimation module (203) is further configured to:

- estimate length of the vehicle using curve functions d=f (x) .
the apparatus (10) according to claim 9, wherein the estimation module (203) is further configured to:

- calculate start point of the vehicle based the curve function d=f (x) of an outer line, wherein the outer line is parallel to the roadside and passing through the start point of the vehicle;

- calculate end point of the vehicle based the curve function d=f (x) of an inner line, wherein the inner line is parallel to the roadside and passing through the end point of the vehicle;

- calculate the length of the vehicle as distance between the end point and the start point of the vehicle.
An apparatus (10) for vehicle length estimation, comprising:

- at least one processor (102) ;

- at least one memory (101) , coupled to the at least one processor (102) , configured to execute method according to any of claims 1～5.
A computer-readable medium for vehicle length estimation, storing computer-executable instructions, wherein the computer-executable instructions when executed cause at least one processor to execute method according to any of claims 1～5.