CN100557384C - An active real-time 3D positioning system based on binocular vision and laser ranging - Google Patents

Abstract

The invention relates to an active real-time three-dimensional positioning system based on binocular vision and laser ranging, which is composed of three parts: a vision processing subsystem, a control subsystem and a mechanical mechanism. The camera acquires the target image under the control of the image acquisition card, and it is processed by the vision processing computer to obtain the two-dimensional position information in the image coordinates, which is sent to the public storage area. The control computer first obtains the position information from the public storage area through coordinate conversion and Real-time planning, then output to the driver to drive the motor through the D/A board and at the same time control the laser to accurately illuminate the target through the feedback loop of the shaft angle encoder, and finally feed back the depth information obtained by the range finder to the visual computer to obtain the three-dimensional position information of the target . The system has the advantages of high long-distance precision, good initiative, real-time speed, low cost and simple process.

Description

An active real-time 3D positioning system based on binocular vision and laser ranging

Technical field:

The invention is an active real-time three-dimensional position information acquisition system based on binocular vision and laser ranging, which belongs to the field of image processing and automation, and is used for quickly and accurately obtaining three-dimensional position information of objects.

Background technique:

At present, the use of binocular stereo vision is the main means to obtain the three-dimensional position information of the target. The commonly used method is to obtain a stereo image pair synchronously by two cameras, and then calculate the depth information of each point in the public field of view of the image according to the principle of stereo vision, generate a depth map, and then extract the depth information corresponding to the target to determine the target’s depth. 3D position information. But these devices all have the following problems:

1. Low long-distance accuracy: The previously studied binocular and multi-eye stereo vision has relatively high accuracy (above 1 mm) when measuring short-distance targets (within 5 meters), and the accuracy is very low at longer distances (beyond 5 meters). Poor can't even be measured.

2. Poor initiative: When the spatial three-dimensional scene is projected into a two-dimensional image by the camera, the image of the same scene at different points will be very different, and many factors in the scene, such as lighting conditions, scene geometry and physical characteristics etc., are integrated into a single image gray value. Obviously, it is difficult to accurately obtain high-precision depth information from an image containing so many unfavorable factors. In applications, in most cases, it is not necessary to obtain depth information for each point. Therefore, many calculations are unnecessary and the ability to actively select targets is lacking.

3. Large amount of calculation and poor real-time performance: Stereo vision requires the ability to solve each corresponding point in a pair of stereo images. The rich information contained in the image makes the speed of point-by-point correspondence matching very slow, and it is prone to "sick" calculation problems . Although in order to solve this problem, certain constraints are added under the regularization framework to transform the problem into the optimal solution of multiple conditional functions, but the difficulty and amount of calculation are greatly increased, making it difficult to improve the speed of visual processing and unable to meet real-time performance. requirements, limiting the field of application.

In order to improve the initiative of the vision subsystem to target three-dimensional positioning, reduce the calculation amount of stereo vision image processing and improve the real-time performance of the system, and actively and quickly obtain the position information of the target, the present invention proposes a new method for determining the three-dimensional position information of the target method and a set of positioning system implemented by applying this method.

Invention content:

The whole system is composed of three parts: vision processing subsystem, control subsystem and mechanical mechanism.

The visual processing subsystem consists of a visual processing computer (built-in image acquisition card, memo-link main card), camera, and laser range finder. The built-in image acquisition card adopted in the present invention is installed in the PCI slot, and is compatible with the adopted video camera. Memo-link is a device used for high-speed communication between two computers. The device is composed of a main card and a sub-card, interconnected with RS485 interface. In the present invention, memo-link is regarded as the public storage area of the system .

The control subsystem includes control computer (built-in D/A board, memo-link daughter card), driver, motor (integrated shaft angle encoder).

The mechanical mechanism is composed of bearings and fixed brackets for installing motors, cameras and laser rangefinders. It has two-dimensional degrees of freedom and mainly plays the role of installation and support. The specific structure is: two cameras compatible with the image acquisition card are fixed on the rectangular bracket in the mechanical mechanism, and the emission part of the laser rangefinder is the semiconductor laser at the midpoint of the two cameras, which is directly above the laser. The receiving part of the laser range finder is installed, that is, a linear array CCD with a focusing lens with a diameter of 4 cm and a focal length of 8 cm. A 15-watt brushless DC motor is placed horizontally behind the cuboid fixing bracket. The motor has a diameter of 4 cm and a length of 10 cm. A built-in shaft encoder is used to drive the system to rotate up and down (Y-axis direction) to a specified angle. Below it is a barrel-mounted bracket with an outer diameter of 6 cm and an inner diameter of 4 cm. A 15-watt brushless diameter motor is installed inside, which is exactly the same as the horizontal motor, and is used to drive the system to rotate left and right (X axis direction) to the specified angle.

Its main technical solutions are:

After the image of the target in front is acquired by the camera, it is transmitted to the image acquisition card, and the image acquisition card converts the image of the target acquired by the camera into a digital signal, and obtains the target position information, and stores it in the memo- link in the public storage area of the main card. The control computer obtains the two-dimensional position information from the public storage area by the memo-link sub-card and converts the position information of the target in the visual coordinate system into the position information in the coordinate system with the laser transmitter as the origin, and calculates the distance of the target offset from the origin. Control quantity, and then output the control quantity to the driver of the motor through the 12-bit D/A converter to drive the motor to rotate, and at the same time control the motor to rotate accurately to the required angle through the feedback loop formed by the shaft angle encoder built in the motor ----The target is located at the receiving end of the laser rangefinder. The focusing lens gathers the reflected signal of the target to the linear array CCD, and calculates the distance information (that is, the depth information) of the target by triangulation according to the output of the linear array CCD. , and finally the detected depth information and two-dimensional position information are fed back to the visual computer to display the three-dimensional position information of the target.

The beneficial effects of the present invention are:

1. High precision. The laser rangefinder has high precision. In addition to being able to achieve the accuracy of stereo vision at a short distance (within 5 meters), especially when measuring a target at a long distance (> 5 meters), there is incomparable stereo vision precision and stability.

2. Be proactive. In the application, the distance information of the points of interest can be obtained according to the requirements of the application-that is, the depth, instead of solving the entire image point by point, reducing unnecessary calculations.

3. Low cost. The present invention uses a dual-CPU structure, and the CPU can also be replaced by DSP or MCU, which is lower in price than similar products.

4. Real-time fast. The depth information of the target in the present invention is measured by laser ranging, which can reach a test frequency above 100 Hz. Moreover, there is no need to perform corresponding point matching on stereoscopic images, so the speed is fast and the real-time performance is very good.

5. The process is simple. The mechanical part is only composed of a motor with two degrees of freedom, a fixed camera, and a mounting bracket for the laser rangefinder, which is easy to process and install.

Moreover, the present invention has the characteristics of small size and light weight, and can be easily integrated into other systems as a functional unit.

Description of drawings:

Below in conjunction with accompanying drawing and embodiment the use is further described.

Fig. 1 is a schematic diagram of the system of the present invention

Figure 2 is a front view of the mechanical mechanism

Figure 3 is a top view of the mechanical mechanism

Figure 4 is a workflow diagram of the visual processing subsystem

Figure 5 is a flowchart of the work of the control subsystem

Figure 6 is a block diagram of the control subsystem

In Fig. 1, 2, and 3: 1. Target, 2. Camera, 3. The transmitting part of the laser rangefinder is the semiconductor laser, 4. The receiving part of the laser rangefinder is the linear array CCD, 5. The Y-axis motor ( Realize up and down movement), 6. X-axis motor (realize left and right movement), 7. Cuboid support, 8. Visual processing computer (built-in image acquisition card), 9. Public storage area, 10. Control computer (built-in D/A board, driver).

Detailed ways:

The specific steps of the system to achieve three-dimensional positioning are as follows:

The first step: the binocular camera 2 sends the image of the target 1 in the visual field to the image acquisition card in the vision processing computer 8

Step 2: The image acquisition card converts the image signal into a digital signal, which is processed by the visual processing computer 8 ( FIG. 3 ), calculates the two-dimensional position of the target on the visual plane and sends the information to the public storage area 9 .

The third step: the control computer 10 reads the position data in the public storage area 9, and carries out coordinate conversion and real-time motion planning according to the installation positions of the camera 2 and the semiconductor laser emitter 3 by the control program (Fig. 4).

Step 4: The control program sends the driving signal to the built-in D/A board of the control computer 10 . After the D/A board is converted into an analog signal, it drives the corresponding X and Y axis motors 5 and 6 to the driver, so that the semiconductor laser 3 is accurately irradiated on the target 1 (Fig. 5).

Step 5: According to the reflected light obtained from the target 1, the depth information of the target 1 can be obtained through the receiving part of the laser rangefinder, that is, the linear array CCD4, and the depth information of the target 1 can be obtained by triangulation and sent to the visual processing computer 8 for comprehensive processing to obtain the target 1 three-dimensional information.

Claims (1)

1. An active real-time three-dimensional positioning system based on binocular vision and laser ranging, including three parts: a vision processing subsystem, a control subsystem and a mechanical mechanism, wherein the vision processing subsystem consists of a vision processing computer, a camera, and a laser ranging The visual processing computer has a built-in image acquisition card and a public storage area, and the control subsystem includes a control computer with a built-in D/A board, a motor and a driver integrated with a shaft angle encoder, and is characterized in that: the camera acquires the front After the image of the target is captured, the image acquisition card converts it into a digital signal, obtains the target position information and stores it in the public storage area, and the control computer obtains the two-dimensional position information from the public storage area, and converts the two-dimensional position information from the visual coordinate system Convert it to the coordinate system with the laser transmitter as the origin, calculate the control amount of the target offset origin, and then output the control amount to the motor driver through the D/A converter to drive the motor to rotate, and at the same time through the built-in motor The feedback loop formed by the shaft angle encoder controls the motor to accurately rotate and align with the target to be measured. The receiving end of the laser range finder gathers the reflected signal of the target to the linear array CCD, and calculates the distance information of the target according to the triangulation method. That is, the depth information. Finally, the depth information and two-dimensional position information are detected and fed back to the visual computer to synthesize the three-dimensional position information of the target.

Images