Automatic zoom method and device for unmanned vehicle

By designing an automatic zoom device on an unmanned vehicle and using the vehicle speed to determine the focal length and pixel size recognition module, the problem of low long-distance target recognition efficiency is solved, and efficient identification and safe driving are achieved.

CN114071020BActive Publication Date: 2024-11-15XIAMEN FOUR FAITH COMM TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111632961.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-28
Publication Date
2024-11-15
Estimated Expiration
2041-12-28

AI Technical Summary

Technical Problem

Existing unmanned vehicles have problems with low recognition efficiency and high misjudgment rate in long-distance target recognition. The cost of lidar is high and the fixed focus camera cannot automatically enlarge and reduce the target, which limits the recognition range.

Method used

Design an automatic zoom device to capture targets through the camera module, identify the pixel size, speed measurement module detects vehicle speed in real time, zoom module calculates focal length based on safe distance, motor walks to realize zooming of camera module, combined with Fast RCNN, YOLOv4, YOLOv5 and other algorithms Perform target recognition.

Benefits of technology

The recognition effect of long-distance targets is improved, the target distance is accurately identified through pixel size, the focal length calculation speed is reduced, millisecond zoom is achieved, and the vehicle's driving safety in dynamic and complex scenarios is enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114071020B_ABST
    Figure CN114071020B_ABST
Patent Text Reader

Abstract

The present invention discloses an automatic zoom method and device for an unmanned vehicle, the device comprising a camera module for photographing a target and obtaining an image of the target; a recognition module for recognizing the target and the pixel size of the target according to the image; a speed measurement module for detecting the speed of the unmanned vehicle in real time and determining a safe distance according to the speed; and a zoom module for obtaining a focal length according to the safe distance and zooming the camera module according to the focal length. The present invention can enable the unmanned vehicle to zoom according to the speed, thereby improving the recognition effect of a distant target.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving technology, and in particular to an automatic zoom method and device for an unmanned driving vehicle. Background Art

[0002] At present, unmanned vehicles mainly use laser radar or fixed-focus cameras to realize the recognition function of long-distance targets. However, laser radar is expensive, and the amount of calculation is large when performing long-distance monitoring, and the recognition efficiency is low. Fixed-focus cameras have a fixed focal length and lack the function of automatically zooming in and out of long-distance targets. They can only have a good recognition effect within a certain distance range, and are prone to misjudgment beyond the range. Summary of the invention

[0003] In order to solve the above problems, the present invention provides an automatic zoom method and device for an unmanned vehicle, which can enable the unmanned vehicle to zoom according to the vehicle speed and improve the recognition effect of distant targets.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] An automatic zoom device for an unmanned vehicle comprises: a camera module, used for photographing a target and obtaining an image of the target; a recognition module, used for recognizing the target and the pixel size of the target according to the image; a speed measurement module, used for detecting the speed of the unmanned vehicle in real time and determining a safe distance according to the speed; and a zoom module, used for obtaining a focal length according to the safe distance and zooming the camera module according to the focal length.

[0006] Preferably, the recognition module is further used to calculate the distance of the target according to the pixel size.

[0007] Preferably, the device further comprises a control module for comparing the safety distance with the distance to the target to control the vehicle speed.

[0008] Preferably, the safety distance is set to be twice the vehicle speed.

[0009] Preferably, the zoom module is loaded with a zoom table, and the zoom module obtains the focal length by comparing the zoom table with the safety distance.

[0010] Preferably, the zoom table includes a coarse focus value and a correction value, the coarse focus value corresponds to the vehicle speed interval, the correction value corresponds to the precise value of the vehicle speed within the interval, and the focal length is the sum of the coarse focus value and the correction value.

[0011] Preferably, the zoom module comprises a motor, and the motor steps to zoom the camera module. The zoom module obtains the number of steps of the motor according to the focal length and the zoom table.

[0012] Preferably, the recognition module adopts one of Fast RCNN, YOLOv4, YOLOv5, and YOLOX algorithms.

[0013] Based on the same inventive concept, the present invention also provides an automatic zoom method for an unmanned vehicle, comprising the following steps: detecting the speed of the unmanned vehicle in real time, and determining a safe distance based on the speed; obtaining a focal length based on the safe distance, and zooming based on the focal length; photographing a target to obtain an image of the target; and identifying the target and the pixel size of the target based on the image.

[0014] The beneficial effects of the present invention are:

[0015] 1. Determine the focal length and zoom according to the vehicle speed to achieve the zoom of the target on the captured image and improve the recognition effect of distant targets;

[0016] 2. By determining the distance of the target based on the pixel size of the target, when there is a target within the safe distance, the pixel size of the target facilitates accurate identification;

[0017] 3. Obtain the focal length by comparing the speed interval and the precise value in the interval with the zoom table, thereby improving the calculation speed of the focal length;

[0018] 4. Accurately control the zoom of the camera module through the movement of the motor. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the pinhole camera model. DETAILED DESCRIPTION

[0020] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention clearer and more understandable, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0021] Embodiment 1

[0022] This embodiment provides an automatic zoom device for an unmanned vehicle, comprising:

[0023] The camera module is used to shoot the target in front of the unmanned vehicle when it is driving and obtain the image of the target.

[0024] Since the focal length of the camera module is determined according to the safety distance, in shooting scenarios where the target and the unmanned vehicle are at different safety distances, the camera module achieves the same zoom effect on the image obtained by shooting at the corresponding focal length, which can improve the recognition effect of distant targets.

[0025] The recognition module is used to recognize the target and the pixel size of the target based on the image.

[0026] The recognition module is loaded with a model of a lightweight outdoor comparison recognition algorithm, which can use one of the Fast RCNN, YOLOv4, YOLOv5, and YOLOX algorithms. In theory, this type of model can recognize targets with a pixel size of 32*32. Considering possible complex scenarios, such as various weather factors, the pixel size of the target that reaches a safe distance in the captured image is set to 64*64, which can achieve a high recognition rate.

[0027] This type of model can identify the type of target, such as pedestrians or vehicles, through three consecutive frames of images. When a pedestrian is identified in front of the same lane, the recognition module notifies the control module to control the driverless vehicle to brake and slow down.

[0028] For distant targets, individual height differences have little effect on the pixel size of the target. Therefore, the recognition module uses preset values ​​to calculate the heights of different types of targets. For example, the preset value for the height of a pedestrian is a low value of 1.2m.

[0029] See attached Figure 1 According to the pinhole camera model, s / f=X / x=f*X*a / u, where s is the distance of the target, f is the focal length, X is the height of the target, x is the height of the target on the image, u is the size of the pixel, and a is the number of pixels per unit height.

[0030] Under the premise that the focal length, the height of the target and the height of each pixel (resolution of the camera module) are determined, the distance of the target is inversely proportional to the pixel size of the target. Therefore, the recognition module can calculate the distance between the target and the unmanned vehicle based on the pixel size of the target in the captured image.

[0031] The speed measurement module is used to detect the speed of the driverless vehicle in real time and determine the safe distance based on the speed.

[0032] The safety distance described in this embodiment is the distance at which the driverless vehicle can achieve safe braking before reaching the target, so as to avoid traffic accidents and improve driving safety. After multiple experiments, the size of the safety distance (in meters) is set to be twice the size of the vehicle speed (in kilometers per hour).

[0033] The zoom module is used to obtain the focal length according to the safety distance and zoom the camera module according to the focal length.

[0034] See attached Figure 1 According to the pinhole camera model, f = S*x / X = v*u / (a*X), where S is the safety distance and v is the vehicle speed.

[0035] In this embodiment, the zoom module is loaded with a zoom table, and the zoom module obtains the focal length according to the safety distance by comparing the zoom table, thereby achieving millisecond-level zoom and identifying distant targets.

[0036] The zoom of the camera module is achieved by moving the internal lens combination spacing. In this embodiment, the zoom module includes a motor, and the accurate adjustment of the lens combination spacing is achieved by the motor running. The number of steps required for the motor to change the corresponding focal length is calculated in advance and recorded in the zoom table. When the zoom module is working, the number of steps can be directly obtained according to the vehicle speed.

[0037] Furthermore, in this embodiment, the zoom table includes a coarse focus value and a correction value. The focal length is the sum of the coarse focus value and the correction value. Referring to Table 1, the coarse focus table in the zoom table is taken as an example. The coarse focus value corresponds to the vehicle speed interval. According to the vehicle speed interval, the coarse focus value and its corresponding number of steps are obtained, and then the correction value and its corresponding number of steps are obtained according to the precise value of the vehicle speed in the interval. The sum of the two is the number of steps that the motor actually needs to run.

[0038] Among them, the coarse focus value and the correction value can be obtained through the product design value provided by the camera module manufacturer. The correction value further includes the deviation value between the actual test value and the product design value (ie, the theoretical correction value). For example, when the vehicle speed is 43km / h, the number of steps corresponding to the coarse focus value is 600, the theoretical correction value is +3, and when the measured statistical speed range is 40-50km / h, the deviation value is +1, that is, the number of corrected steps obtained by comparing the correction table in the zoom table is +4. Therefore, the actual number of steps of the motor operation is 604.

[0039] Table 1 Coarse focus table

[0040] Speed ​​range km / h Safety distance m Coarse focus value mm Steps 10 10 6 280 20 20 12 470 30 30 16 530 40 40 25 600 50 50 30 610

[0041] The control module is used to compare the safety distance with the target distance to control the vehicle speed.

[0042] Since driving events are a continuous dynamic process, after the target enters the safe distance, the distance from the unmanned vehicle may change at any time, or leave the front of the vehicle. Therefore, it is necessary to monitor the target in real time and control the speed of the unmanned vehicle accordingly to achieve the purpose of smooth driving.

[0043] When the target enters the safe distance, its distance can be monitored by its pixel size. Especially when the target is coming head-on, it is necessary to increase the braking intensity and provide early warning and protection measures for the people in the car.

[0044] When there is no target within a safe distance or a distance that the recognition module can accurately identify, the control module can gradually restore the vehicle speed to an appropriate speed to improve the smoothness of the journey.

[0045] The device described in this embodiment can realize real-time monitoring of distant targets and has good recognition effect, and is better suitable for dynamic and complex driving scenes.

[0046] Embodiment 2

[0047] This embodiment provides an automatic zoom method for an unmanned vehicle, comprising the following steps:

[0048] S1. Detect the speed of the driverless vehicle in real time and determine the safe distance based on the speed.

[0049] This can better adjust the monitoring distance of the target and the distance the vehicle uses to respond according to the vehicle speed in different driving scenarios.

[0050] S2. Get the focal length according to the safety distance and zoom according to the focal length.

[0051] The method described in this embodiment performs calculation based on a pinhole camera model.

[0052] The zoom operation is performed by the motor stepping, and the zoom focal length and the number of motor steps are obtained by comparing the zoom table. The vehicle speed is divided into intervals, and each interval corresponds to a coarse focus value. Therefore, it is only necessary to correct the coarse focus value by the precise value of the vehicle speed to obtain the correct focal length, which can reduce the amount of data calculation and achieve a zoom speed of milliseconds.

[0053] S3. Photograph the target and obtain an image of the target.

[0054] Take photos of the road conditions in front of the driverless vehicle, especially in the same lane.

[0055] S4. Identify the target according to the image, and the pixel size of the target.

[0056] The target in the image is identified through the outdoor comparison recognition algorithm to obtain the target type and pixel size. When the pixel size of the target reaches the threshold, it is determined that the target is within the safe distance of the unmanned vehicle and the distance to the target is calculated in real time.

[0057] S5. Compare the safe distance with the target distance to control the vehicle speed.

[0058] When a target appears within a safe distance, brake and slow down in time to avoid traffic accidents.

[0059] The above description shows and describes the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the invention, through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not depart from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. An automatic zoom device for an unmanned vehicle, It is characterized in that include: A camera module, used to shoot a target and obtain an image of the target; A recognition module, used to recognize the target and the pixel size of the target according to the image; the recognition module is also used to calculate the distance of the target according to the pixel size; the recognition module adopts one of Fast RCNN, YOLOv4, YOLOv5, and YOLOX algorithms to recognize the type of the target through three consecutive frames of images, and the recognition module uses preset values ​​to calculate the heights of different types of targets; A speed measurement module, used to detect the speed of the unmanned vehicle in real time and determine the safe distance according to the speed; A zoom module is used to obtain a focal length according to the safety distance, and to zoom the camera module according to the focal length; the zoom module is loaded with a zoom table, and the zoom module obtains the focal length by comparing the zoom table with the safety distance; the zoom table includes a coarse focus value and a correction value, the coarse focus value corresponds to the interval of the vehicle speed, the correction value corresponds to the precise value of the vehicle speed within the interval, and the focal length is the sum of the coarse focus value and the correction value; the zoom module includes a motor, the motor steps to zoom the camera module, and the zoom module obtains the number of steps of the motor by comparing the focal length with the zoom table; The system also includes a control module for comparing the safety distance with the target distance to control the vehicle speed.

2. The automatic zoom device for an unmanned vehicle according to claim 1, It is characterized in that The size of the safety distance is set to be one times the size of the vehicle speed.

3. An automatic zoom method for an unmanned vehicle, It is characterized in that The following steps are involved: detecting the speed of the unmanned vehicle in real time and determining a safe distance according to the speed; Obtaining a focal length according to the safety distance, and zooming according to the focal length; The focal length is obtained by comparing the safety distance with a zoom table; the zoom table includes a coarse focus value and a correction value, the coarse focus value corresponds to the interval of the vehicle speed, the correction value corresponds to the precise value of the vehicle speed in the interval, and the focal length is the sum of the coarse focus value and the correction value; the zoom module includes a motor, the motor steps to zoom the camera module, and the zoom module obtains the number of steps of the motor by comparing the focal length with the zoom table; photographing a target to obtain an image of the target; Identify the target and the pixel size of the target according to the image; calculate the distance of the target according to the pixel size; the recognition module adopts one of Fast RCNN, YOLOv4, YOLOv5, and YOLOX algorithms to identify the type of the target through three consecutive frames of images, and the recognition module calculates the height of different types of targets using preset values; The safety distance is compared with the distance to the target to control the vehicle speed.