Design of visual positioning system for automatic window cleaning machine

    Abstract: This paper introduces a method for precise positioning using dual vision systems in automatic window cleaning machines. The "cross" line generator circuit in visual positioning is described in detail. At the same time, the precise positioning principle of the system is briefly introduced.

    Keywords: automatic window cleaning machine visual positioning "cross" line generator

In the design of the control system for automatic window cleaning machines in tall buildings, determining the relative position of the window cleaning machine and the visual scrubbing unit is a key technology that must be solved. The process of determining the relative position is divided into two steps. The first step is for the window cleaning machine to find the unit to be wiped. The second step is for the window cleaning machine to find the unit to be scrubbed and accurately determine the relative position between the window cleaning machine and the unit to be scrubbed. position, providing necessary data for automatic cleaning of the window cleaning machine. In the window cleaning machine we designed, these two steps are coordinated through two sets of vision systems. There is an ordinary camera and a CCD camera on the window cleaning machine. The signal of the ordinary camera is connected to a black and white monitor. There is a fixed white or black "cross" line on the black and white monitor. The ground operator can monitor through the black and white The user manually controls the window cleaning machine to find the unit being scrubbed, and can preliminarily determine the relative position of the window cleaning machine and the unit being scrubbed by aligning the "cross" line on the monitor with the edge of the unit being scrubbed. The position error range does not exceed 50mm; then the precise positioning system starts to work, converting the CCD camera signal into a digital signal through the image acquisition card. After computer processing, the precise position of the window cleaning machine and the unit being cleaned can be determined, and the positioning error range does not exceed 3mm. The key technology of the entire visual system is the generation of the "cross" word line and the visual system positioning algorithm. The following will elaborate on the generation of the "cross" line and briefly introduce the precise positioning principle of the visual system.

1 Design of “Ten” word line generator

The "Ten" word line on the visual monitor is realized through the "Ten" word line signal generator. The "cross" word line meets the following conditions: (1) the "cross" word line is located in the center of the monitor; (2) the length reaches two-thirds of the monitor frame, and the line width is the width of one pixel; (3) The "ten" word line is stable and does not flicker or deform. The circuit block diagram of the "ten" word line signal generator is shown in Figure 1. In the scheme of Figure 1, the TV signal from the ordinary camera is synchronously separated to obtain two signals: the field synchronization header (V-sy) and the horizontal synchronization header (C-sy). The field synchronization head is used to start a counter. The counter uses the horizontal synchronization head as the counting pulse. The counting output is connected to a combinational logic circuit. When the counter counts to a certain value, the control circuit outputs a strobe pulse to make it in the prescribed range. The rows (those rows displaying the "cross" word line) can get a strobe (called the full strobe pulse); in the same place, the rows displaying the "cross" word line can get another strobe pulse (called the middle row strobe pulse).

The circuit of the "ten" word line generator is shown in Figure 2. In the figure, LM1881 is a horizontal and vertical synchronization head separation device. It separates the video signal from an ordinary camera into a line sync header and a field sync header. In each field, counter CD4040 counts the row sync header at the end of the field sync header. GAL16V8 decodes the count output to determine whether the word "ten" should be displayed. When reaching the specified row, the SEL end of GAL16V8 will output a full strobe pulse to strobe the middle row pulse generator and point pulse generator. These two pulse generators are implemented through two monostable trigger delay circuits 74HC423 of. Once a row sync header comes, the delay function of 74HC423(1) will be triggered, and its output Q2 will get an intermediate row pulse Wide that lags behind the row sync header T1 and has a width of T2. This pulse is sent to GAL16V8 and triggers 74HC423 (2) at the same time, generating a point pulse Narrow that lags behind the middle row pulse T4 and has a width of T3. The final "cross" word line signal pulse is selected and output by GAL16V8. The "ten" word line generation timing is shown in Figure 3. Take the black-and-white monitor we selected from Japan's Panasonic Company as an example. Its scanning lines are 525 lines, using the NTSC standard and interlaced scanning. Therefore, the entire screen needs to be scanned 262 times. When the scanning behavior is 131, the middle line pulse is output and the horizontal line of the "ten" is displayed; starting from the 40th line and ending at the 220th line, the full strobe pulse is output and the "ten" is displayed. "The vertical line of the word line. The video superposition circuit is shown in Figure 4. Cross IN (the same signal as Cross OUT in Figure 2) is the input terminal of the "ten" word line signal. The input pulse is of negative polarity. When there is no "ten", it is high level. When there is a "ten", a negative pulse is added to the position that needs to be displayed. When Cross IN is high level, D1 turns on P, 1 turns off, and input video signals R4 and R5 are output; when the negative pulse arrives, D1 turns off, P1 is saturated and turned on, and the emitter potential of P1 is higher than the collector potential. This pulse is added in reverse to the television image signal to achieve the purpose of image superposition.

2. Precise positioning of window cleaning machine

The precise positioning algorithm of the window cleaning machine includes image processing algorithms, such as image segmentation, edge extraction, and pose conversion calculations between coordinate systems. Due to space limitations, this article only briefly introduces this. For detailed explanation, see reference [1]. Figure 5 shows the camera model. The coordinate system OXYZ is fixed on the robot. The coordinate system oxyx is the camera coordinate system, xoy is the image plane, w is the object point, and (X, Y, Z) is its coordinate in the OXYZ coordinate system. , c is the image point, (x, y, z) is its coordinate in the image plane, the matrix between wh and ch is expressed as: Ch=Awh; where, wh is the homogeneous coordinate of w, ch is the image point The homogeneous coordinates of c, A is the transformation matrix from the coordinate system OXYZ to the image oxyz. The inverse solution can obtain the relationship between the object point coordinates (X, Y, Z), the image point coordinates (x, y, z) and the object distance z.