阅读说明：本技术 检测器校准 (Detector calibration ) 是由 S·雷马克斯 于 2018-06-11 设计创作，主要内容包括：一种用于检测器的校准的图案,包括一系列对比的亮段和暗段；这些段被布置为使得能够从图案的窗口部分确定任何点在图案上的绝对位置,图案的窗口部分包括该点；窗口部分由预定的最小数量的段组成,段的预定的最小数量小于该系列中段的总数量。(pattern for calibration of a detector, comprising series of contrasting light and dark segments, the segments being arranged so as to enable determination of the absolute position of any point on the pattern from a windowed portion of the pattern comprising the point, the windowed portion consisting of a predetermined minimum number of segments, the predetermined minimum number of segments being less than the total number of segments in the series.)

patterns for calibration of a detector, the patterns comprising a linear series of contrasting light and dark segments;

the segments being arranged to enable a detector to determine an absolute position of any point on the pattern from a windowed portion of the pattern, wherein the windowed portion comprises the point; the windowed portion includes a predetermined number of segments, the predetermined minimum number of segments being less than the total number of segments in the series.

2. The pattern of claim 1, wherein the segments are rectangular in shape and comprise a length and a width.

3. The pattern of claim 1 or 2, wherein the width of at least segments is different from the width of at least a second segment.

4. The pattern of claim 3, wherein at least segments have a non-uniform width along their length.

The detector calibration system, comprising:

a target having applied thereto a pattern as claimed in any preceding claim.

6. The detector calibration system of claim 5; the system is arranged such that when the detector views the window portion, the absolute position and orientation of the target relative to the detector can be determined.

7. The detector calibration system of claim 6 wherein transitions between at least of the contrasting light and dark segments of the pattern as viewed by the detector comprise gray scale transitions.

8. The detector calibration system of claim 7 wherein the transitions between at least of the contrasting light and dark segments are sinusoidal gray scale transitions.

9. The detector calibration system of claim 6 further arranged to enable the focal point of the detector relative to the target to be determined from the standard deviation of the detected intensities of the light and dark segments of the pattern.

10. The detector calibration system of claim 8, the system further arranged to enable the focus of the detector relative to the target to be determined from a comparison of the difference between the detected intensity of the bright segments and the detected intensity of the dark segments.

11. The detector calibration system of claim 5 wherein the detector comprises a line scan camera.

12. The detector calibration system of claim 11 arranged such that the vertical position and angle of the line scan camera relative to the target can be determined from the angle between the scan line of the line scan camera and the centre line of the pattern.

13, a method of producing a pattern comprising a linear series of contrasting light and dark segments, the method comprising arranging the series of contrasting light and dark segments to correspond with a generated binary output of a Linear Feedback Shift Register (LFSR) such that 1 bit corresponds to a light segment and 0 bits correspond to a dark segment.

14. The method of claim 13, wherein the pattern comprises a linear series of m segments, and wherein the binary output of the LFSR is generated by:

selecting the value n such that m < ═ 2^ n-1+ n-1;

selecting a maximum length LFSR of size n;

selecting an LFSR initial state;

starting from the selected LFSR start state, a binary sequence of length m is created.

15. The method of claim 14, further comprising applying manchester encoding to the binary output of the LFSR.

16. The pattern of claim 1, wherein the linear series of contrasting light and dark segments corresponds to a binary output:

10101010100110100110010110011001101010100110010110011001010110011010101010010101100110010110100101011001010101010101101010010110010110100101010101101001101001011001101001010101010110011010010110010101100110100110011001100101010110010110010101100110011010100101011001011010100110011001101001010110010110011010010101101001011010011001101001101001100101101001010101100110100110101010010110100101011001101001101001010110011010100110101001011010100101011010011010100110011010010101 1010101010011010100110010110101001010110011010010101011010101001101010100110010101101010100101101010011001010101010110101010010110010101011001101001010110011001010101011010101010010110011010101010011010101001011001100110010101010110010110011001011010100110010110100110011001010101010110010110101010101001010101011001011010010110011001101001101001011001010101101001

such that a 1 bit corresponds to a bright segment and a 0 bit corresponds to a dark segment.

17. The pattern of claim 16, wherein the window portion is comprised of 13 bits.

Technical Field

The present invention relates to detector calibration, in particular to calibration of cameras and line-scan (line-scan) cameras in multiple directions.

Background

Calibration of the detector device is typically an time-consuming task, requiring multiple steps and often multiple calibration or alignment targets against which the position and focus of the detector is evaluated.

Sorting machines of different kinds require the use of detectors to generate images which are then analysed, for example, to determine whether the item to be sorted should be rejected or allowed to continue the process flow.

In order to properly direct rejected items based on data from the detector (or other sensor), the exact location on the sorting belt or chute that the detector is looking at needs to be known for each pixel Another reasons for knowing the location of the pixel on the belt are to enable geometric calibration of the detector.

In some belt sorters, a calibration scale is placed on the sortation belt in the scan line of the detector. The scale has an alternating pattern of high intensity segments and low intensity segments. This allows some way of rejection assignment and geometric calibration of the detector to be performed.

However, such systems have many disadvantages, for example, when scan lines are split across multiple detectors, the intermediate detector(s) view the middle portion of the scale, in order to capture the exact position on the belt, the operator typically has to count the th segment within the field of view of each detector and set it as a parameter in the software.

It is therefore desirable to provide more efficient detector calibration that improves upon the above techniques.

Disclosure of Invention

The present invention provides patterns for calibration of a detector, the pattern comprising a linear series of contrasting light and dark segments, the segments being arranged such that the detector can determine the absolute position of any point on the pattern from a window portion of the pattern, wherein the window portion comprises the point, the window portion comprising a predetermined number of segments, the predetermined minimum number of segments being less than the total number of segments in the series.

This is advantageous because it allows the detector to determine absolute position on the pattern by examining a windowed portion of the pattern, the dimensions of the windowed portion are determined to include a sufficient number of pattern segments so that positions within the complete pattern can be determined, this is particularly advantageous for calibration of detectors (such as, for example, cameras) because it eliminates the need for physical markings such as scales required in calibration techniques.

The segments of the pattern may be rectangular in shape and include a length and a width the width of at least segments may be different from the width of at least a second segment.

This is advantageous because the rectangular shaped segments can be easily identified by the series detector type, furthermore, providing segments with different widths enables horizontal positions within the pattern to be identified.

At least segments of the pattern may have non-uniform widths along their lengths this is advantageous because providing segments of non-uniform widths along their lengths enables vertical positions within the pattern to be identified.

The detector calibration system may include a target having a pattern applied thereto. The detector calibration system may be arranged such that the absolute position and orientation of the target relative to the detector is determinable when the detector views the window portion. Such a system has the advantage of allowing rapid calibration and validation of the detector position within the system.

Transitions between at least of the contrasting light and dark segments of the pattern as viewed by the detector can include gray scale transitions the transitions between at least of the contrasting light and dark segments can include sinusoidal gray scale transitions this is advantageous as it allows the sub-pixel accuracy of the detector position relative to the target to be determined.

The detector calibration system may also be arranged such that the focus of the detector relative to the target may be determined from the standard deviation of the detected intensities of the light and dark segments of the pattern. This is advantageous because it allows the detector to obtain an initial or rough focus.

The detector calibration system may further be arranged such that the focus of the detector relative to the target may be determined from a comparison of the difference between the intensity of the detected bright segments and the intensity of the detected dark segments. This is advantageous because it allows the detector to obtain a secondary or fine focus.

The detector calibration system may include a line scan camera. The detector calibration may be arranged such that the vertical position and angle of the line scan camera relative to the target may be determined from the angle between the scan line of the line scan camera and the centre line of the pattern. This is advantageous as it provides alignment and calibration of the camera relative to the target.

The present invention also provides a method of producing a pattern comprising a linear series of contrasting light and dark segments, the method comprising arranging the series of contrasting light and dark segments to correspond to a generated binary output of a Linear Feedback Shift Register (LFSR) such that a 1 bit corresponds to a light segment and a 0 bit corresponds to a dark segment.

In practice, the binary output is transformed into series of light and dark segments that can be viewed by the detector.

The pattern may comprise series of linear m segments, where the binary output of the LFSR is generated by:

selecting the value n such that m < ═ 2^ n-1+ (n-1);

selecting a maximum length LFSR of size n;

selecting an LFSR initial state;

starting from the selected LFSR start state, a binary sequence of length m is created.

The pattern produced by this method can be used for calibration of the detector.

The method may further include applying manchester encoding to the binary output of the LFSR. This is advantageous because it allows no more than two consecutive bits in the pattern to be identical. This provides sufficient edge density, i.e., sufficient transition between the light and dark segments of the contrast, so that the absolute position on the pattern can be determined with improved accuracy compared to the output of non-manchester encoding.

The linear series of contrasting light and dark segments of the pattern may correspond to a binary output:

such that a 1 bit corresponds to a bright segment and a 0 bit corresponds to a dark segment.

The window portion of the pattern may consist of 13 bits. This provides an effective minimum window size to allow the absolute position on the pattern to be determined.

Drawings

FIG. 1 is a representation of a 4-bit sized linear feedback shift register with a mask of 0011;

FIG. 2 is a representation of a maximum length linear feedback shift register of size 4 bits;

FIG. 3 is a representation of a sample sequence generated from a linear feedback shift register with mask 1001;

FIG. 4 is a binary representation of a pattern according to the present invention;

FIG. 5 is a binary representation of a Manchester encoding pattern according to the present invention;

FIG. 6 is a pattern according to the present invention;

FIG. 7 is a representation of a segment of a pattern according to the present invention;

FIG. 8 is a representation of a segment of a pattern according to the present invention;

FIG. 9 is a representation of a pattern of non-uniform segment widths in accordance with the present invention;

FIG. 10 is a representation of sub-pixel fitting;

FIG. 11 is a representation of sub-pixel fitting.

Detailed Description

The present invention will now be described with reference to the accompanying drawings. The pattern for detector calibration comprises a linear series of contrasting light and dark segments; the segments being arranged such that the absolute position of any point on the pattern can be determined from a windowed portion of the pattern, wherein the windowed portion comprises the point; the windowed portion includes a predetermined number of segments, the predetermined minimum number of segments being less than the total number of segments in the series.

Thus, methods of generating a pattern comprising a linear series of contrasting light and dark segments are provided, the method comprising arranging the series of contrasting light and dark segments to correspond with a generated binary output of a Linear Feedback Shift Register (LFSR) such that 1 bit corresponds to a light segment and 0 bits correspond to a dark segment.

Feedback function

The taps may be expressed as a mask, for example, if there are taps on the , third, and fourth bits, the mask is 1101 (note that the th bit is to the right).

The result of the feedback function defined by a given mask for a certain input sequence is either 0 or 1 depending on whether the number of in the input sequence being masked is even or odd.

For example:

input device	0	1	1	0
					Mask code	1	1	0	1
Masked input	0	1		0

count 1 is odd- >1

Input device	1	1	0	0
					Mask code	1	1	0	1
Masked input	1	1		0

has a count of 2 being even- >0

The mask may be represented as a binary number. However, for large masks, it is easier to convert them to decimal or hexadecimal numbers. Therefore, the upper mask is 0b1101 or 13 or OxD.

Linear feedback shift register

Linear Feedback Shift Registers (LFSRs) are defined by the size of their state (in bits) and a linear feedback function that returns 1 bit.

For example:

given value 1101, the feedback function with mask 0011 will return 1, which will be appended to the shifted value.

Therefore, the new value becomes 1011.

FIG. 1 illustrates an example of a state space for a LFSR of size 4 bits, where the feedback function includes an XOR operation on the th bit and the second bit for each value in FIG. 1, the arrow indicates the value that will be the next values using the LFSR with mask 0011.

These types of LFSRs are called fibonacci LFSRs the number of LFSRs of a given bit size n is called maximum length LFSRs such LFSRs have a loop containing 2^ n-1 states (this is all possible states except for states with all zeros). fig. 2 illustrates the number of values in the maximum length LFSR loop with state size 4 bits with mask 1001 or 0x 9. for a length 2 LFSR only masks result in LFSR:0x3, again for a length 3 LFSR may be 0 x6. for a length 4 LFSR there are two maximum srs with masks 0x9 and 0x xC. for a length 5 LFSR there are 6 maximum LFSRs with masks 0x12, 0x 354, 0x 8295, lfx 3870 x 3870, 4830 x 461, 580 x 25.

Pattern generation and absolute positioning

To generate a pattern according to the present invention, the result from the feedback function is represented as series contrasted light and dark segments.

Starting from 0001, the result is 1, so the pattern starts with 1.

The next states become 0011, the result is 1, and thus the pattern becomes 11.

The next states become 0111, the result is 1, so the pattern becomes 111.

The next states change to 1111 with a result of 0, so the pattern changes to 1110.

Thus, the pattern 1110 may be represented as 3 light segments for representing three 1 s, and dark segments for representing 0 s.

Since each state of an n-bit LFSR overlaps with its predecessor (which have n-1 bits in common), these states can be linked in a long bit sequence. The sequence is a concatenation of all the generated bits. For the example maximum length LFSR outlined above with mask 1001 and start state 0001, the following sequence is generated: 000111101011001000. sequences of length 2^ n-1+ (n-1) can be created where the last n-1 bits contain a wrap around of the first n-1 bits. Note that the sequence can start at any given state, so 2^ n-1 different sequences can be generated.

The construction of the sequence is such that each 4-bit subsequence in the sequence is only by assigning an index position to each state in the loop, a mapping from subsequence to position can be created table 1 below provides a look-up table for this example sequence 000111101011001000 each 4-bit subsequence (ii) is interpreted as a binary number and the resulting number is used as index (i) in the table, the additional value of the index (iii) in the table is the position of the subsequence in the sequence.

i	ii	iii
				0	0000	-1
1	0001	0
			2	0010	12
3	0011	1
			4	0100	13
5	0101	7
			6	0110	9
7	0111	2
			8	1000	14
9	1001	11
			10	1010	6
11	1011	8
			12	1100	10
13	1101	5
			14	1110	4
15	1111	3

TABLE 1

Thus, for a pattern comprising series of linear m segments, the sequence may be generated by:

selecting the value n such that m < ═ 2^ n-1+ (n-1);

selecting a maximum length LFSR of size n;

selecting the LFSR start state;

starting from the selected LFSR start state, a binary sequence of length m is created.

Horizontal orientation and direction

For example, with respect to FIG. 3, a line across the top of the upper representation indicates that the 6-bit window being read from left to right is 011110. a line across the top of the lower representation indicates that the 6-bit window being read from right to left is also 011110. thus, in this case, a detector (e.g., a camera) may not be able to resolve the direction it is pointing in with respect to the pattern defined by the sequence, as a camera pointing at the pattern from the left may read the sequence 011110 as highlighted in the upper representation of FIG. 3, while a camera pointing at the pattern from the right may read the sequence 011110 as highlighted in the lower representation of FIG. 3.

For the above example, if a window of size 6 is selected, the sequence is found to contain only three positions, while the window contains sub-sequences that also appear as "mirror" sequences (see FIG. 3, where the sequence read from left to right in the above representation corresponds to the sequence read from right to left in the below representation). Increasing the window size helps to reduce the occurrence of "mirrored" bit series. For the example sequence of fig. 3, table 2 indicates the number of instances that a "mirror" sequence will be found relative to a given window size. Thus, a window size of 8 allows determining the direction of line of sight of a detector for which no corresponding mirror sequence exists.

Window size (position)	Examples of "mirror" sequences
		4	15
5	6
		6	3
7	1
		8	0

TABLE 2

Thus, in order to generate a pattern that provides an absolute position found in a sequence of m positions, it must be possible to determine from which direction a given windowed line of sight of the pattern is looking at the pattern.

To generate such a pattern, for a given n, all possible maximum length feedback functions and all possible sequence start states are considered. For each pattern thus generated, the minimum window size required to distinguish the direction of line of sight is determined. The pattern with the smallest window size is then determined.

For a given m, one may choose a larger than th n equation m ≦ for the LFSR<2ⁿBit size where-1 + n-1 will hold. This means that the look-up table for sub-sequence positions in a given sequence will become larger, but a smaller window for determining direction can be created.

Manchester coding

As noted in section above, the resulting pattern may have very long 0 s or 1 s, which can be overcome by applying Manchester encoding to the pattern, although this doubles the number of bits in the pattern, it ensures that no more than two consecutive bits will be the same.

Calibration system

Embodiments of a detector calibration system including a pattern as described above are described. Other arrangements of patterns and detectors are possible in addition to those described below.

The detector calibration system may include a target having a pattern applied thereto as described above. For example, the target may be a scale and the pattern may be applied along the length of the scale. The detector calibration system may be arranged such that when the detector views the window portion of the pattern on the scale, the absolute position and orientation of the target relative to the detector can be determined.

The detector may be used with, for example, sorter . the scale may extend across the conveyor of the sorter in order to be suitable for use with a conventionally sized sorter , the width of the dark and light segments of the pattern cannot be too small.

To accommodate a 2000 mm scan line, the scale comprising the pattern may be made slightly wider on its sides (e.g., 50 mm on each side) making the overall pattern 2100 mm wide.using a segment width of 2.5 mm will result in a 840 segment pattern, which corresponds to a 420 bit LFSR sequence (no manchester encoding).

Thus, a search is made for the minimum window size to determine the absolute position and orientation of the pattern of length 420. Table 3 shows the results for different sizes of LFSR:

TABLE 3

Referring to table 3, column indicates LFSR size n. column two indicates window size w. for use with manchester encoding , the window size should be multiplied by a factor of 2.

A pattern segment width of 2.5 mm was selected. This provides an effective high resolution performance, i.e. a camera implementation allowing the pattern to be seen very close together, and an effective low resolution performance, i.e. allowing the camera to see the pattern from a greater distance.

For example, for the th row of values in table 3, the window width is thus derived from w 2.5 (e.g., 16 x2 x 2.5-80).

It is therefore determined that the minimum window size is 13. select a sequence with 2048 smaller lookup tables this provides 196 patterns with the minimum window it is preferable to ensure that the th and last bits of the selected pattern are 1 this provides symmetric ending at both ends of the sequence and simplifies the processing of the edge view of the scale.

Mask of taps: 0x503(1283) -hexadecimal (decimal)

Initial state: 0x25(37) -hexadecimal (decimal)

The resulting complete pattern (420 bits) is shown in fig. 4. The corresponding manchester encoded pattern (840 bits) is shown in fig. 5.

Figure 6 shows the entire 2100 mm pattern (cut into 17.5 cm pieces) on a 1:1 scale.

Vertical positioning

In addition, at least segments may have non-uniform widths along their lengths.

Referring to fig. 7, there is shown the area below and above the centerline of the pattern of the present invention, where the height of the area is defined by △ y in [ -1,1] the vertical position is defined as being on the centerline for △ y 0, at the top of the area for △ y 1, and at the bottom of the area below the centerline for △ y-1.

Given a light segment of thickness B (i.e., a sequence of bit values corresponding to 1), the thickness may vary linearly in regions from B-B/2 to B + B/2 according to △ y.

Referring to FIG. 8, given a double bright segment of thickness 2B (i.e., a sequence of bit values corresponding to 1 followed by another sequence of bit values of 1), the thickness may vary linearly according to △ y in the region from 2B-B/2 to 2B + B/2.

Depending on its shape, the pattern segments may be referred to as either "odd" or "even". The difference between the "odd" and "even" white segments can be seen, for example, with reference to FIG. 9. the th (odd) white line starts thick at the top of the figure and tapers off at the bottom.the second (even) white line starts thin at the top of the figure and ends thick at the bottom.an "odd" characteristic is also shown in FIGS. 7 and 8. the thickness of the middle part is B in FIG. 7 (or 2B in FIG. 8) plus △ yB/2. for the "even" lines, the "swap" behavior is that the line is thin at the top and thick at the bottom.thus, in the formula for thickness, the sign changes from positive to negative. thus, the thickness of the odd segments (single) is equal to B + △ yB/2, and for the even segments it is equal to B- △ yB/2. this means that the thickness of the dark segments remains unchanged, but their position shifts depending on △ y.

Thus, for a segment of basic width B, if the width of the odd white line is W ═ B + △ yB/2, the height can be calculated to be △ y ═ 2W/B-2 furthermore, the width of the even white line is W ═ B- △ yB/2 and the height can be calculated from △ y ═ 2-2W/B.

Thus, if we assume that in the detector calibration system, the detector (e.g., camera) should be focused on the centerline, then the calculation △ y or centerline may indicate misalignment with the center, and thus recalibrate the detector to the centerline accordingly.

In operation, the camera may be set to view the pattern, but need not necessarily be viewed parallel to the centerline of the pattern.

The more white segments, the more accurate the fit will be, so this is another advantage for using Manchester coding for the pattern.

Sub-pixel fitting

In the patterns described in the preceding sections, the positions on the pattern are determined by the edge positions of the light and dark segments, and in fact, position data is encoded in the edges between the dark and light segments. This allows position detection to be reduced to the pixel level when viewed through a camera or other digital optical device. However, pixels in dark areas and pixels in light areas of the pattern are not themselves used to determine locations within the pattern. To detect segment edges with a finer resolution than the pixel level, the gray value of the pixels (2 pixels) at the edge can be checked. However, this method is generally prone to noise generation. An alternative is to define a grey pattern around the edges. This enables a plurality of pixels to be used in order to define the position of the local edge with sub-pixel accuracy.

In the following, a sinusoidal based approach is described, referring to fig. 10, the edges between high and low are replaced by sinusoidal shapes that change from light to dark (or vice versa) in the thickness of segment (B), half of B precedes the edge of segment and half of B follows the edge of segment.

referring to FIG. 10, the vertical and horizontal straight lines represent the gray values of the original pattern image the curved lines are the gray values of the new image Note that the middle portion of the 2-bit wide section is either completely light or completely dark (see straight lines in FIG. 10).

The threshold of the image may be in the middle of the light and dark values (see dashed lines) and the edge of the binarized image will be at the same position as the original image.

For high resolution (e.g., where the detector is positioned closer to the pattern), there are fewer edges to average over, i.e., when the camera is viewing fewer pattern segments.

For vertical positioning, the thickness or width of the pattern segments varies along their height. This means that the edge transition between the light and dark segments shifts with the width. When the edge is shifted, the threshold line and the sinusoidal shape should still intersect at the edge. Thus, the sinusoidal curve is also linearly stretched depending on the height. This is shown in fig. 11. The middle portion (denoted "0" in the vertical direction) is at the centerline (no shift), the top (denoted "1" in the vertical direction) and bottom (denoted "-1") portions are at the extremes of the vertically aligned stripes (maximum shift). The light segments are denoted by 1 and 3 and the dark segments are denoted by 2 and 4.

The vertical lines indicate linear shifts of the edges. For example, the first two vertical lines indicate narrowing of the bright section (1), and the second and third vertical lines indicate shifting of the edge position of the dark section (2). Thus, it can be seen that the center of the bright segment remains at the same location, but the segment narrows or widens, while the dark segment remains at the same width, but the center changes position.

Optimizing detector focus

The focus of the detector in the detector calibration system may be optimized by first providing a coarse focus indication and subsequently providing a finer focus indication.

The coarse indication of the focus of the detector on the pattern of segments can be achieved by analyzing the standard deviation of the gray values in the image or frame grabber obtained from the detector.

An th embodiment of finer focus indication makes use of the presence of thin and thick white segments in the pattern, for thin segments (using a sinusoidal transition as described above) only pixels will be at the highest intensity value.

A second embodiment of finer focus indication uses checking segment edges without sinusoidal transitions. For example, each ninth segment edge may be a high contrast edge. Assuming that a poor focus will blur such transitions over multiple pixels, the maximum difference between two adjacent pixels is measured around the segment edge without a sinusoidal transition. The higher the difference, the better the focus, i.e. indicating a strong transition from white to black segments on two adjacent pixels. To obtain a stable indication, the fine focus measurements for all edges may be averaged. This provides the following advantages: due to the high edge density (i.e., multiple segments), the focus can be measured and optimized over the entire scan line, rather than just at a particular point.

As used herein with reference to the present invention, the terms "comprises," "comprising," and the terms "having/including" are used to specify the presence of stated features, integers, steps, or components, but do not preclude the presence or addition of or more other features, integers, steps, components, or groups thereof.

It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable subcombination.