A signal processing system for an article counter.

This invention relates to a signal processing system for an article sensor for sensing and counting articles such as seeds flowing in a chute in a seed planter and which provides a signal representing the amount of radiation unobscured by an article or articles passing through the sensor and reading a radiation detector.

Optical seed sensors in which a seed interrupts a radiation or light beam are known in the art. Such systems are described in US 4,163,507; US 3,537,091; US 3,928,751; US 3,723,989; US 4,166,948; US 3,974,377; and US 4,246,469. For a number of reasons, such seed sensors have been inaccurate. One problem has been the spatial non-uniformity of the light source and/or of the light detectors so that signals generated by the light detectors vary, depending upon what portion of the light beam is interrupted. Means are described below which overcome this problem but the present invention is concerned with a different problem arising in the processing of the sensor signal. In the prior art it has been customary either to compare the sensor signal with a threshold to determine when an article is present (e.g. US 4,166,948) or to derive a pulse from the sensor signal by a differentiating circuit (e.g. US 4,163,507). In either event the system is liable to count a plurality of seeds simultaneously traversing the light beam as a single seed.

The object of the present invention is to provide an improved processing system which can sense and count a plurality of articles in a group of articles which pass simultaneously through the sensor.

The signal processing system according to the invention is characterised as set out in claim 1.

As will be described below, the system can be arranged to compensate automatically for changes in the steady-state output of the detector and to compensate automatically for gradual changes in average article or seed size.

The system may be used to determine the time spacing between the articles for use in testing of planter seed metering devices.

The system can be arranged to ignore articles too small to be the ones monitored.

The sensor itself preferably includes an array of infrared LEDs extending across one side of an article or seed conduit. The array generates a substantially diffuse and uniform radiation beam which is detected by planar photo diodes which extend across the opposite side of the conduit. A pair of oppositely facing mirrors extend between the array and the photo diodes and reflect the LED radiation back into the conduit. Slits between the array and the conduit and between the conduit and the photo diodes narrow the beam which the articles or seeds traverse and prevent extraneous radiation from impinging upon the photo diodes.

With the diffuse, uniform and extended radiation beam produced by the LED array, all articles or seeds passing through the detector have nearly equal effect on the amount of radiation received by the photo diodes, even when multiple articles or seeds are tightly bunched together, even when multiple articles pass simultaneously through the beam, and even when one article is (as viewed perpendicularly to the array) in the partial "shadow" of another article which is between the one article and the array. The signal from the photo diodes has a substantially linear relationship to the total amount of radiation which falls on them, and thus, also has a similar (but inverse) relationship to the quantity of articles or seeds which interrupt the beam inside the detector.

The signal from the photo diodes is processed by an electronic unit which includes a current-to-voltage converter, an A/D converter and a microprocessor. The microprocessor executes an algorithm which accurately counts the articles which pass through the beam by repetitively integrating a value derived from the signal from the photo diodes. The algorithm compensates for changes in the steady-state signal produced by the photo diodes when no articles are in transit through the beam, and determines the number of articles in groups of articles which simultaneously pass through the beam. The algorithm also compensates for gradual changes in average article size.

The invention will now be described, in more detail, by way of example and with reference to the accompanying drawings, in which: <ul id="ul0001" list-style="none"><li>Fig. 1 is a sectional side view of a beam-type article or seed sensor;</li><li>Fig. 2 is a partial sectional view taken along line 2-2 of Fig. 1, with parts removed for clarity;</li><li>Fig. 3 is a view of the photo diode mounting plate looking towards the LED array with the photo diodes removed;</li><li>Fig. 4 is a side view of one of the radiation transmitting windows of the sensor of Fig. 1;</li><li>Fig. 5 is an electrical schematic of a signal processing system embodying the present invention;</li><li>Fig. 6 is a signal timing diagram illustrative of signals which can be produced by transit of seeds through the sensor of Figs. 1 - 4; and</li><li>Figs. 7a - 7e contain a logic flow diagram of the signal processing algorithm executed by the signal processing unit of Fig. 5.</li></ul>

An article or seed sensor 10 includes a conduit 12 which forms an article or seed flow passage 14 and which receives a sensor module 16. Sensor module 16 includes a top 18 and a base 20, each having rectangular openings 22 and 24 which register with the seed flow passage 14.

The sensor module 16 also includes opaque end plates 26 and 28 (see Fig. 2), opaque side plates 30 and 32, mirrors 34 and 36, and glass windows 38 and 40, all held in grooves on the inner surfaces of the top 18 and base 20.

The side plate 30 supports an array 42 (at least 3 and preferably 7) of radiation generators CR1 - CR7. Various known radiation emitting devices could be suitable, but infrared light generators are preferred because of the dust-penetrating ability of infrared radiation. A suitable device is the Siemens No. SFH 407-3 GaAs infrared light emitting diode (LED). Preferably, plate 30 is a PC board with conductive strips forming electrical connections with the LEDs CR1 - CR7 mounted thereon.

As best seen in Fig. 2, this array of LEDs extends substantially across the entire width of the seed flow path 14 and transversely to the direction of seed flow which is downwards, viewing Fig. 1. The radiation beam generated by each LED has a wide angular dispersion approaching that of a point light source mounted on a planar surface. Thus, the beams from adjacent pairs of the LEDs intersect with each other well before they reach the nearest window 38. This assures that all areas in the seed flow path between windows 38 and 40 are illuminated.

The end plate 32 is preferably made of opaque black plastic and has a rectangular recess 44 which receives a pair of flat planar detectors or photo diodes 46 and 48 for generating electrical signals in linear response to radiation received thereby. End plate 32 also includes a longitudinal slot or aperture 45, which has a width which is smaller than a typical dimension of the articles or seeds being sensed, (preferably 1 mm wide). Thus, slot 45 permits only a portion of the radiation from LED array 42 to impinge upon detectors 46 and 48. The slot 45 reduces the amount of ambient radiation (other than from array 42) which impinges upon detectors 46, 48. Slot 39 in window 38 narrows the angular spread of beam B to prevent the beam from reflecting off of articles or seeds which are outside of a small portion of the volume surrounded by mirrors 34 and 36 and windows 38 and 40.

Any detector which is responsive to the radiation generated by array 42 is suitable; however, in the case where infrared LEDs are used, then photo diodes, such as Type No. SP-652S made by Centronic, Inc., or the equivalent, are preferred. The end plate 32 and photo diodes 46 and 48 are positioned parallel to and spaced apart from the array 42 so that seeds traveling through seed passage 14 must pass between the array 42 and the photo diodes 46 and 48, thus varying the amount of radiation received thereby. The photo diodes 46 and 48 thus form a planar radiation detector which extends transversely with respect to the seed flow path across the longer dimension of the rectangular openings 22 and 24.

The radiation reflecting mirrors 34 and 36 are positioned parallel to each other on opposite sides of the seed flow path. Each mirror extends from an edge of side plate 30 to an edge of side plate 32. The mirrors 34 and 36 are preferably silvered or reflectively coated on the sides facing away from the seed flow path so that the reflective coatings will not be damaged due to abrasive contact with seeds.

Viewing Fig. 2, radiation from LEDs CR1 to CR7 which would otherwise be directed out of the path traversed by the seeds is reflected back into the seed path by mirrors 34 and 36. This has an effect similar to having the array 42 extend laterally beyond the plane of mirrors 34 and 36. The array 42 and the mirrors 34 and 36 cooperate to form a substantially diffuse, uniform and essentially extended radiation beam which enables the present detector to, in assence, "look behind" one seed to sense a seed which would otherwise be in the shadow of a seed which is closer to array 42.

The glass window 38 is spaced apart from and parallel with respect to the LED array 42 and is transparent to the infrared radiation emitted thereby and has its inward facing surface in line with an inner wall 50 of the conduit 12. The window 38 extends from mirror 34 to mirror 36. As best seen in Figs. 1 and 4, window 38 has an opaque coating or mask 37 on the side nearest the LED array 42. A longitudinal gap 39 in the mask 37 forms a slit aperture, preferably around 1 mm wide, through which the radiation from array 42 is transmitted. The gap 39 extends the full length of window 38 between mirrors 34 and 36.

The window 40 is positioned parallel to the window 38 on the opposite side of the seed flow passage 14. The transparent glass window 40 has a radiation-blocking opaque mask 41 on the side facing away from seed passage 14. A longitudinal gap 43 in the mask 41 forms a slit aperture, preferably around 2 mm wide, through which the radiation from LED array 42 is transmitted. The gap 43 also extends the full length of window 40 between mirror 34 and 36.

As best seen in Fig. 5, each of LEDs CR1 - CR7 is connected in series with a corresponding resistor Rl - R7 and the resistor/LED pairs are then connected in parallel to a +5 volt power supply. The two detectors 46 and 48 are electrically connected in parallel. As best seen in Fig. 2, the resistors Rl - R7 may be located in the spaces between mirrors 34 and 36 and end plates 28 and 26. The current signal from detectors 46 and 48 is received by a current-to-voltage amplifier 50. Preferably, amplifier 50 includes an operational amplifier (such as an RCA No. CA 3160), a 44 pf feedback capacitor Cl and a 562 kOhm feedback resistor. Amplifier 50 provides an analog voltage to the Vin input of conventional analog-to-digital converter 52 (such as a National Semiconductor ADC 0820). A/D converter 52 provides an 8-bit digital signal (representing the voltage at Vin) to the P0.0 to P0.7 inputs of microprocessor (micro) 54 (such as an Intel 8051). The A/D converter 52 starts an A-to-D conversion in response to a flag signal received at its WR/RDY input.

The micro 54 is supplied with a 12 MHz frequency from crystal oscillator 56. This frequency is divided internally to provide a 1 mHz machine instruction frequency. A timer (not shown), which is internal to the micro 54, counts the machine cycle frequency and generates a flag signal every 100 micro-seconds.

The micro 54 causes a new A/D conversion to be performed by converter 52 and executes an algorithm or instruction set every 100 micro-seconds in response to the occurrence of the flag signal.

The algorithm or program executed by micro 54 is best understood with reference to the signal timing diagrams of Fig. 6 and to the logic flow diagrams of Figs. 7a - 7e.

Turning first to Fig. 6, the upper waveform is typical of an oscilloscope trace of the voltage at the Vin of A/D converted 52 when articles such as ball bearings are passed through the detector 10. The signal pulses at 60, 68 and 70 are representative of the signal produced by a single article passing through the detector 10. The signal pulses at 64, 66, 74 and 76 are representative of 2 articles passing through the detector 10. Pulse 74 is produced when the 2 articles pass sequentially, one immediately following the other. Pulse 64 is produced when the second article enters the radiation beam before the first article leaves it. Pulses 64 and 66 represent situations where 2 articles pass through the detector nearly simultaneously, or in very close proximity to each other, regardless of the orientation of the article grouping. Pulse 62 is produced by 3 articles passing nearly simultaneously through the detector 10. Pulse 72 is produced by 4 articles passing nearly simultaneously through the detector 10. The paranthetical numbers inside the waveform pulses are proportional to the area circumscribed by the pulses in arbitrary units. These waveforms illustrate that the area circumscribed by each is related to the number of articles which produce the waveform.

It should be noted that a differentiating-type counter would probably incorrectly interpret pulses 62, 64, 66 and 72 as being produced by 1, 1, 2 or 3, and 3 articles, respectively, whereas, these pulses are actually caused by groups of 3, 2, 2 and 4 articles, respectively. The following signal-processing algorithm correctly interprets these pulses as being caused by article counts of 3, 2, 2 and 4, respectively.

Turning now to Figs. 7a - 7e, the algorithm begins at step 100 by setting a HALF UNIT value equal to 1/2 of a UNIT value which is initially equal to 768 to represent an initial estimate of the typical area circumscribed by the signal pulse produced by passage of a single article through the sensor apparatus. Such a pulse is shown at 60 of Fig. 6. Then, step 102 causes the algorithm to pause until the internal timer generates a flag signal at 100 micro-second intervals. Upon generation of the flag signal, step 104 causes A/D converter 52 to perform a conversion and input a new digital Vin value (INPUT) into the micro 54. Then, in step 106, a SIGNAL value is set equal to OFFSET - INPUT, where OFFSET represents the possibly slowly varying steady-state level of Vin (normally 4 volts) when no seeds are interrupting the beam B. Thus, when a seed is in the beam B, the SIGNAL value will normally be positive and will represent the vertical depth of the Vin signal (see Fig. 6) at each sampling instant relative to the normal or steady-state value of Vin when no seed is in the beam B.

However, the SIGNAL may be negative if no seed is present and if the OFFSET value is lower than the current steady state Vin level. In this case, step 108 directs the algorithm to steps 136 - 144. In step 136, an DNTIME timer is initialized to a value representing a 12 msec interval. Step 138 decrements an UPTIME timer. Step 140 routes the algorithm to step 150 if the UPTIME timer has not counted out; otherwise, in step 142, the OFFSET value is incremented by 1 binary count. Finally, step 144 sets the UPTIME timer to a 3 msec value. Thus, the OFFSET value will be incremented if the SIGNAL value remains negative for more than 3 msec.

If SIGNAL is not negative, then step 108 directs the algorithm to step 110 which determines if SIGNAL = 0. If yes, it means that no seed is present and that the current OFFSET value appears proper and steps 146 and 148 set the UPTIME and DNTIME timers to values representing 3 milliseconds and 12 milliseconds, respectively. If no, then it means it is possible that a seed or seeds are in the beam B.

In step 112, the UPTIME timer is set to a 3 millisecond value. The DNT_IME timer is decremented in step 114. Then, step 116 determines if the DNTIME timer value is greater than zero. If no, it means that SIGNAL has been positive for 12 milliseconds and the OFFSET value is adjusted by 1 digital count in step 118, and the DNTIME timer is again set to a value representing 12 milliseconds. If in step 116 the DNTIME counter is greater than zero (which means that SIGNAL has been positive for less than 12 milliseconds), or after step 120, the algorithm proceeds to step 122.

In step 122, a PULSE value (initially zero), is numerically integrated by adding to its previous value the current SIGNAL value. Thus, the PULSE value represents an area circumscribed by the graphical representation of the Vin signal pulses shown in Fig. 6.

Step 124 determines whether SIGNAL equals a digital count of 2 or 1. If not, it means that SIGNAL must be greater than 2 since steps 108 and 110 have already determined that SIGNAL is non-negative and non-zero. In this case, it means that a seed or seed group has begun or remains in transit through the beam B and the algorithm proceeds to step 126 where a Pl.l flag (initially zero) is set equal to 1. Then, step 128 determines whether the area value PULSE is greater than or equal to the HALF UNIT value (which represents 50% of the typical area of the signal pulse produced by transit of a single seed.) If PULSE has not attained this 50% area value, then the algorithm returns to step 100 for updating of the SIGNAL value in step 106 and further integration of the PULSE value in step 122. However, if PULSE exceeds the 50% area value, then step 130 causes the signal at micro output port Pl.O to toggle to indicate transit of a seed through the sensor. Next, step 132 increments a QUAN value (initially zero) which represents the total number of seeds in seed group which may be passing through the sensor. Then, step 134 sets the area value, PULSE, equal to (PULSE - UNIT) and returns the algorithm to step 100. This makes the PULSE value negative so the condition of step 128 will again be met only upon additional repetitive integration of the PULSE value by step 122 due to transit of a further seed or seeds of a seed group.

Referring back to step 124, if the SIGNAL value has a digital value of 2 or 1, it is interpreted to mean that the passage of a seed or a seed group through the beam B has just begun or has just been completed (or that noise or negative drift of the bias level has occurred) and the algorithm proceeds to step 150 and further integration of the PULSE value is prevented. Step 150 determines if a Pl.1 flag value (initially zero) is equal to 1. If Pl.l does not equal 1, then it means either that step 126 has not yet been executed because there is no convincing evidence (i.e., SIGNAL &gt;2) that a seed is in transit and that P1.1 was previously cleared to zero at step 151 when the last seed transit was finished. In this case, the algorithm is directed to steps 208 - 212 wherein the PULSE and QUAN values are cleared and the algorithm is returned to step 100. If, on the other hand, the P1.1 value equals 1 in step 150, then it means a seed transit is just ending and the algorithm is directed to step 151 where Pl.1 is cleared.

In step 152, the area value PULSE is compared to the HALF UNIT area value. If PULSE is less than HALF UNIT, then the algorithm proceeds to step 160. However, if PULSE is not less than HALF UNIT, then step 154 causes the micro output port Pl.O to toggle (as at step 130) to indicate transit of a seed through the sensor. Then, the total seed number value, QUAN, is incremented in step 156, and the PULSE value is reset to a negative value in step 158 (as in step 134).

At this point, it is helpful to understand how the value, PULSE, varies as a single seed passes through the beam B. Initially, the PULSE value will be zero. Then, as a seed transit produces a waveform, such as 60 of Fig. 6, the PULSE value will be repetitively integrated by the addition of the increasing SIGNAL values in step 122 until PULSE equals the HALF UNIT value, at which time, the Vin level reaches a minimum and the SIGNAL value reaches a maximum. Then, step 128 operates to direct the algorithm through steps 130 - 134, wherein step 134 resets the PULSE value to a negative value, typically equal to -(HALF UNIT), if the UNIT value accurately represents the area circumscribed by the waveform pulse being processed. Then, during the remainder or second half of waveform 60, step 122 integrates the PULSE value back up so that when Vin returns to its steady state value and when SIGNAL reaches zero, the PULSE value will return to zero, again assuming that the UNIT value was an accurate estimate of the total area of pulse waveform 60.

Now, if, in fact, the estimated area value UNIT, was too large, then at the end of a seed transit, the PULSE value in step 122 will be slightly negative. Thus, as described later in detail, this slightly negative PULSE value will be utilized in algorithm portion 180 to slightly reduce the TOTAL value. Since the TOTAL value is stored as a 3-byte value (each byte consisting of 8 bits) and since, by definition, the UNIT value is that which is stored in the 2 most significant bytes of TOTAL, therefore, a reduction in the TOTAL value also reduces the UNIT value, thus making the UNIT value more closely approximate the typical or average signal pulse area produced by a single seed transit. Similarly, if the estimated area value, UNIT, was too small, then the PULSE value in step 122 (at the end of pulse area integration) will be slightly positive. This will cause the algorithm portion 180 to slightly increase the TOTAL value, and will cause a corresponding increase in the UNIT value for use during the next seed transit. Thus, by adjusting the TOTAL and UNIT values, the algorithm automatically compensates for changes in the average size of seeds passing through the sensor.

Steps 160 - 210 will now be described. To summarize, steps 160 - 210 operate to make major adjustments (if ever needed) in the estimated signal pulse area value, UNIT, so that the correct values of UNIT and HALF UNIT will be utilized in steps 100, 128, 134, 152 and 158.

Steps 160 - 166 determine whether the QUAN value (initially zero or set in steps 132 or 156) equals 0, 1, 2, 3 or more (representing signal pulses caused by the transit of something less than a seed (QUAN = 0) or by the transit of seed groups consisting of 1, 2, 3 or more seeds, respectively).

Under normal conditions, the signal pulse which is produced most often will be that which is caused by the transit of a single article or seed through the beam B, thus QUAN will most often be equal to 1 (assuming a reasonably accurate UNIT value). In this case, step 162 will route the algorithm to a portion of the algorithm represented by 180 which has the effect of deriving an updated TOTAL value equal to the sum of the current TOTAL and residual PULSE values. Since, as previously described, the TOTAL value is related to the UNIT value, this, in effect, repetitively adjusts the UNIT value so that it continues to represent the signal pulse area caused by transit of a single seed. Then, step 182 decrements a ONES counter (initially 256 or reset to 256, at step 200). If the ONES counter is decremented to zero, then step 184 recognizes this overflow condition and routes the algorithm to steps 198 - 210 which reset the ZEROES, ONES, TWOS, THREES and FOURS counters to 256 and which clear to zero the PULSE and QUAN values so that they can be redetermined by steps 100 - 158. If the ONES counter has not overflowed, then the algorithm is directed by step 184 directly to steps 208 and 210. Thus, if the UNIT value accurately represents the estimated single seed pulse area, the algorithm will most often incrementally adjust the UNIT value (via adjustment of the TOTAL value in 180), and will continuously reset the ZEROS, TWOS, THREES and FOURS counters in steps 198, 202, 204 and 206 so that the algorithm will never execute step 178 or step 172 which either divides TOTAL by 2 or multiplies TOTAL by 2.

However, if the UNIT value is too large, then the QUAN value will most often be zero because steps 128 and 152 would prevent incrementing of the QUAN value in steps 132 or 156. In this case, the algorithm will most often be directed by step 160 to step 174 which decrements the ZEROES counter. If this situation persists, then step 174 will eventually decrement the ZEROS counter to zero, whereupon step 176 will recognize this overflow condition and will route the algorithm to step 178. Step 178 reduces the TOTAL value by 50% (for example) and thus, causes a corresponding reduction in the UNIT value. Eventually, this process will reduce the UNIT value to a level whereby single seed transits will produce QUAN values equal to 1.

If the estimated pulse area value, UNIT, is too low, then the most often occurring single seed transits can result in QUAN values of 2, 3 or more. In this case, steps 164 and 166 will route the algorithm to steps 186, 192 or 194 where TWOS, THREES and FOURS counters (initially 256 or reset to 256 in steps 202 - 206) are decremented. When any of these counters reaches zero, then steps 188, 194 or 170 will recognize the overflow condition and will route the algorithm to step 172. Step 172 multiplies the TOTAL value by 2, thus causing an increase in the estimated area value, UNIT. Otherwise, steps 188, 192 and 170 will route the algorithm directly to steps 208 and 210 and thence, back to step 100.

It has been found that it is adequate merely to double the TOTAL value (such as in step 172) regardless of which of the TWOS, THREES or FOURS counters overflows first. However, it would be possible to change the TOTAL value by different amounts, depending upon which counter overflowed first by adding separate TOTAL recalculating steps after each of steps 188, 194 and 170.

Another alternative would be to route the "NO" branch from step 162 directly to step 186 (eliminating steps 164 - 170, and steps 192 - 194) and to make the initial and reset value of the ONES counter smaller than that of TWOS counter so that under normal circumstances, the ONES counter will continue to overflow before the TWOS counter (which, in this case, would be decremented upon the transit of any seed group producing a QUAN value of 2 or more.)

At the end of this "Detailed Description" are object and source code listings of the computer program which is illustrated by the logic flow chart of Figs. 7a - 7e. The source code listing includes labels such as READ: and ADDPULSE:, which corresponds to similar labels in the flow chart. There also follows a cross-reference symbol table listing which includes various acronyms used in the flow chart and program listing.

The signal-processing algorithm described herein could be used in conjunction with another type of article or seed sensor as long as the sensor can generate a signal which varies substantially linearly with the number of articles or seeds within it. <tables id="tabl0001" num="0001"><img file="EP0174818A2_D0001.tif" /></tables><tables id="tabl0002" num="0002"><img file="EP0174818A2_D0002.tif" /></tables><tables id="tabl0003" num="0003"><img file="EP0174818A2_D0003.tif" /></tables><tables id="tabl0004" num="0004"><img file="EP0174818A2_D0004.tif" /></tables><tables id="tabl0005" num="0005"><img file="EP0174818A2_D0005.tif" /></tables><tables id="tabl0006" num="0006"><img file="EP0174818A2_D0006.tif" /></tables><tables id="tabl0007" num="0007"><img file="EP0174818A2_D0007.tif" /></tables>