CN201188667Y - A binocular stereo camera with automatic adjustment of baseline length - Google Patents

Abstract

The utility model relates to a binocular stereoscopic video camera with automatic baseline length adjustment, which is used for robot guidance, three-dimensional reconstruction and the like, and belongs to the field of machine vision. The binocular stereoscopic video camera comprises a stereoscopic imaging part, a stereoscopic vision information processing part and an electric motor driving part. The stereoscopic imaging part comprises two CCD image sensors which are installed on two parallel slide rods, wherein, one CCD image sensor is connected with the electric motor driving part and can slide along the slide rods on a bracket due to the guidance of the electric motor driving part, and the other CCD image sensor is fixedly connected with the slide rods. The stereoscopic vision information processing part is an embedded type image processor which is connected with the two CCD image sensors, and the embedded type image processor is connected with the electric motor driving part. Compared with a binocular stereoscopic video camera with fixed baseline length, utility model adds the degree of freedom for baseline length adjustment, and the baseline length can be changed in use so as to meet the requirements of different assignments for matching or tracking and the like.

Description

A binocular stereo camera with automatic adjustment of baseline length

technical field

The utility model relates to a binocular stereo camera with automatic adjustment of baseline length, which is used for robot navigation, three-dimensional reconstruction, etc., and belongs to the field of machine vision.

Background technique

Stereo vision technology has been widely used in mobile robot navigation, multi-target tracking, 3D measurement and 3D reconstruction. At present, there are many kinds of stereo cameras used in robots, such as the Bumblebee series stereo cameras produced by Point Gray Research Inc of Canada. Bumblebee2 is composed of two side-by-side CCD image sensors, and the lens can be a focal length of 3.8mm or 6mm. The micro-lens, with a baseline length of 12cm, is connected to the PC through 1394 lines, and image matching and other algorithms are implemented on the PC, which can build a real-time depth map. American Videre Design company produces a binocular stereo camera whose baseline can be adjusted manually, such as the STH-MDCS3-VAR stereo camera, which consists of two side-by-side CMOS image sensors, the baseline length can be adjusted between 5-20cm, and the lens can be selected Various C/CS interface standard lenses, image matching and other algorithms are implemented on the PC, and real-time depth maps can be obtained. Another kind of STH-MDCS3-VARX camera, the baseline length can be between 18-60cm. The step length is adjusted by 7cm, and other parameters are similar to those of STH-MDCS3-VAR. The invention patent of Wang Yaonan, Yu Hongshan, Liang Ang and Duan Feng of Hunan University (patent application number: 03124756.3)---human-like multi-degree-of-freedom stereoscopic binocular vision device, its main purpose is to automatically obtain three-dimensional environmental information, with a camera Horizontal rotation, pitching action, four degrees of freedom adjustment of the vergence of the two cameras, fast and dynamic interactive adjustment of each moving part can be realized through the computer control system, and the movement function similar to human eyes and head has been realized. intelligence.

At present, the binocular stereo camera mainly has the following problems:

The distance between the two cameras of the binocular stereo camera, that is, the baseline length cannot be adjusted automatically. As mentioned above, the baseline length of Bumblebee2 produced by Canadian Gray Point Company (Point Gray Research Inc) is fixed, and Videre Design Company Although the baseline length of the produced STH-MDCS3-VAR(X) can be adjusted, this adjustment is done before use and cannot be adjusted during work. Invention patent of Hunan University Wang Yaonan et al.---Human-like multi-degree-of-freedom stereoscopic binocular vision device. Although the two cameras have the function of automatic vergence adjustment, this adjustment is formed by the rotation of the camera without changing the baseline. Length, and after the camera rotates, it will bring great difficulty to algorithms such as image matching and depth calculation.

Utility model content:

The purpose of this utility model is to overcome the above-mentioned defects of the prior art, and provide a binocular stereo camera whose baseline length can be automatically adjusted. The utility model has high-precision servo control performance, and can precisely adjust the length of the baseline, thereby changing two The size of the common visual range of the cameras can effectively calculate the depth information of the image, and can be applied to robot navigation, target tracking, 3D reconstruction and other fields.

In order to achieve the above object, the utility model adopts the following technical solutions. It includes a stereo imaging part, a stereo vision information processing part and a motor transmission part. The stereoscopic imaging part includes two CCD image sensors, and the two CCD image sensors are installed on two parallel sliding rods on the bracket, and one of the CCD image sensors is connected with the motor transmission part, and can be guided along the sliding rods on the bracket. Sliding, another CCD image sensor is fixedly connected with the sliding rod. The stereo vision information processing part is an embedded image processor connected with two CCD image sensors to receive and process image information from the two CCD image sensors; the embedded image processor is connected with the motor transmission part.

The motor transmission part includes a motor, a motion control card for controlling the motion of the motor, and a belt. The belt is fixedly connected to one of the CCD image sensors through a connection block. The motion control card is connected to the motor to drive the motion of the motor. Motion control The card is also connected with the embedded image processor of the stereo vision information processing part.

It also includes a host computer, which is respectively connected with the stereo vision information processing part and the motion control card of the motor transmission part.

The motor is a DC servo motor with a motor encoder.

The binocular camera of the utility model always keeps the optical axes of the two cameras parallel, and the image planes of the two cameras are on the same plane, which satisfies the most basic binocular three-dimensional geometric relationship. depth $z=\frac{\mathrm{fb}}{d},$ Where f is the focal length, b is the baseline distance, and d is the disparity obtained by matching the left and right images. When the baseline length is changed, only the value of b in the formula needs to be changed. At the same time, it can be seen from the above depth calculation formula that in order to improve the accuracy of the depth value, the baseline length b can be increased. When the baseline length increases, the visual range of the two cameras increases, which is conducive to the search and tracking of objects, but the common The visible range is reduced, and the disparity value corresponding to the scene point is increased, which increases the difficulty of image matching. According to the above principles, the baseline length can be changed in real time according to the needs of tasks at different times and places. In the utility model, the adjustment range of the baseline length is 5-20cm.

The utility model completes multiple tasks by acquiring the information of two image sensors. For example, the three-dimensional information of space objects can be obtained, and a certain target in space can be tracked. The working principle of acquiring three-dimensional information of space objects is: the optical axis of one of the image sensors is always parallel to the optical axis of the other image sensor during the movement, the image planes of the two sensors are located on the same plane, and a point in space is at The projection positions on the two image planes are different, and the coordinate difference is called parallax. According to the triangle similarity principle, with the help of the distance between the two image sensors, that is, the baseline length, the depth of the point in space can be obtained, and combined with the camera Calibration techniques can obtain the three-dimensional information of this point. The working principle of tracking a target in the space is as follows: firstly, it is necessary to find the target in the field of view of the two sensors or artificially specify a target, and then use the image tracking algorithm to continuously search in the current field of view until a suitable target is found. In various tasks in various application situations, the required baseline lengths between the two image sensors are different. When the baseline length is large, the common visual range of the two cameras is small, and the visual range of the two cameras is not the same. Larger sets are beneficial to target tracking, but at this time only the 3D information of distant targets can be calculated; when the baseline length is small, the common visual range of the two cameras is large, and the 3D information of closer targets can be calculated, but The union of the visual ranges of the two cameras is relatively small, which is not conducive to the search and tracking of objects. At the same time, through the formula $z=\frac{\mathrm{fb}}{d}$ (f is the focal length, b is the baseline distance, and d is the disparity obtained by matching the left and right images) It can be seen that when the baseline length is large, the accuracy of the calculated depth information of the object is relatively high, but when the baseline length is large, it gives Object matching increases the difficulty.

The beneficial effect of the utility model is that relative to the binocular camera with a fixed baseline length, the utility model adds a degree of freedom for adjusting the baseline length, which is similar to the convergence function of the human eye, so that the utility model can be freely transformed during use The length of the baseline makes the utility model work in the best state of adapting to each task, so as to meet the requirements of different tasks such as matching or tracking, and simultaneously the variation of this baseline length does not bring great difficulty to the algorithm process (from above It can be seen from the formula that only one of the values needs to be changed), the utility model can adapt to more complex visual tasks and has certain flexibility.

Description of drawings

Fig. 1 is the basic composition block diagram of the utility model

Fig. 2 is the structural representation of the utility model

Fig. 3 is the stereo vision model diagram of the present utility model

Fig. 4 is a functional structure block diagram of the utility model

In the figure: 1. Stereo vision information processing part, 2. Stereo imaging part, 3. Motor transmission part, 4. Bracket, 5. CCD image sensor, 6. Connecting block, 7. Motor, 8. Driven wheel, 9. Belt , 10. Image plane, 11. Baseline, 12. Common visual range, 13. Host computer, 14. Motion control card.

Detailed ways

Below in conjunction with accompanying drawing 1～4, the utility model is further described:

This embodiment includes a stereo vision information processing part 1 , a stereo imaging part 2 and a motor transmission part 3 . As shown in Figure 1, the stereo vision information processing part 1 receives the synchronous image collected by the stereo imaging part 2, and calculates the depth map of the image in real time. The length adjustment command is sent to the motor transmission part 3, and the motor transmission part 3 acts on the stereoscopic imaging part 1 to realize the adjustment of the length of the baseline 11 between the two CCD image sensors 5. As shown in FIG. 3 , the image planes 10 of the two CCD image sensors 5 are coplanar, and when the baseline length 11 is adjusted, the common visual range 12 of the two image sensors 5 will change.

As shown in Figure 2, the stereoscopic imaging part 2 includes two CCD image sensors 5, which are installed on the bracket 4 with two parallel slide bars, wherein one CCD image sensor 5 is fixed on the bracket 4, and the other can be driven by the motor transmission part 3 The guide slides freely on the bracket 4. The motor transmission part 3 includes a DC servo motor 7 , a motion control card 14 and a belt 9 . The motor 7 drives the driven wheel 8 to rotate through the belt 9, and the belt 9 is fixedly connected to a CCD image sensor 5 through the connecting block 6, so that the motor 7 drives a CCD sensor 5 to slide along the slide rod through this connection.

The CCD image sensor 5 of the stereoscopic imaging part is in a progressive scanning mode, which is more suitable for image acquisition in a dynamic environment, and can be equipped with various types of standard C/CS interface lenses.

Stereoscopic vision information processing part 1 is an embedded image processor, and what this embodiment selects is a DSP processor, is furnished with codec chip, RS232 serial interface and IEEE1394 serial bus interface above, and this processor both can pass RS232 The interface is connected with the motion control card 14 to realize the control of the motor 7, and can also be connected with the upper-level host computer 13 through the 1394 interface to realize the real-time transmission of image information.

The motor 7 of the motor transmission part is a DC servo motor, which is provided with a motor encoder, and is connected to one of the CCD image sensors 5 through a belt 9, so that the position of the image sensor can be precisely positioned. The RS232 interface on the motion control card 14 can be connected with the DSP processor of the stereo vision information processing part 1 and can also be connected with the host computer 13 to meet various application environments.

The synchronous image information collected by the CCD image sensor 5 is transmitted to the stereo vision information processing part 1, and the matching work of the two images is completed in the stereo vision information processing part 1, and the length of the baseline 11 is determined according to the quality of the matching effect, and then the baseline is passed through the serial port. 11 The length adjustment command is sent to the motion control card 14, the motion control card 14 drives the motor 7, and the motor 7 drives one of the image sensors 5 to a designated position through the belt 9. At the same time, the stereo vision processing part 1 can also send the images collected by the image sensor 5 or processed images to the host computer 13 through the 1394 interface, and the host computer 13 can do some high-level work, such as tracking targets, robot navigation, etc. The machine 13 can adjust the length of the baseline 11 according to the needs of the task, and then send instructions to the motion control card 14 through the serial port, and the motion control card 14 drives the motor 7 to realize the movement of the image sensor 5 .

Claims (4)

1, a kind of binocular stereo camera of baseline length automatic adjustment, comprise stereoscopic imaging part (2), described stereoscopic imaging part (2) comprises two CCD image sensors (5); It is characterized in that: also comprises Stereo vision information processing part (1) and motor drive part (3); Wherein: described stereo vision information processing part (1) is an embedded image processor, links to each other with two CCD image sensors (5), receives from The image information of two CCD image sensors (5) is processed, and the two CCD image sensors (5) are installed on two parallel sliding rods on the support (4), and one of the CCD image sensors (5) is connected with the motor transmission part (3) connected, guided by it, can slide along the sliding rod on the bracket (4), and another CCD image sensor (5) is fixedly connected with the sliding rod; the motor transmission part (3) is connected with the embedded image processor; 2. A binocular stereo camera with automatic adjustment of baseline length according to claim 1, characterized in that: said motor drive part (3) includes a motor (7), and a motion control device for controlling the motion of the motor (7). The card (14) and the belt (9), the belt (9) is fixedly connected to one of the CCD image sensors (5) through the connection block (6), and the motion control card (14) is connected to the motor (7) to drive the motor (7) motion, the motion control card (14) is connected with the embedded image processor of the stereo vision information processing part (1). 3. A binocular stereo camera with automatic baseline length adjustment according to claim 2, characterized in that: it also includes a host computer (13), and the host computer (13) communicates with the stereo vision information processing part (1) and the stereo vision information processing part (1) respectively. The motion control card (14) of the motor transmission part (3) is connected. 4. A binocular stereo camera with automatic baseline length adjustment according to claim 2, characterized in that: said motor (7) is a DC servo motor with a motor encoder.

Images