JPS6344193B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a device that automatically determines the condition of fruits such as watermelons. Traditionally, when determining the ripeness of fruits such as watermelon,
People used to judge the ripeness of the fruit by looking at it with their eyes and listening to the sounds made by tapping it with their hands. In other words, in the past, in order to estimate the ripeness of fruits such as watermelons using non-destructive testing, it was difficult to estimate the ripeness of a fruit such as a watermelon by examining the characteristics inherent to the object that vary depending on the ripeness (color: apples and tomatoes become redder as they ripen; sound: watermelons become riper). People used to detect changes in the sound (the sound changes as the sound changes) and make judgments based on empirical intuition. Some methods have been put into practical use to determine the ripeness of fruit by measuring its color with a meter, but the surface color of fruits, such as watermelons, hardly changes as the ripeness progresses. Until now, the practical application of a ripeness determination device and automatic sorting using non-destructive testing have not been achieved. For this reason, the judgment results vary greatly depending on the skill level and fatigue level of the operator, the judgment accuracy is low, and the amount of manpower required for the sorting work is large. This invention was made in order to improve the accuracy of judgment and save labor in sorting work, that is, to automate the sorting work. An example in which this invention is applied to the ripeness determination of watermelon (watermelon) will be described below. First, by combining some parameters that represent the characteristics of the sound waves generated by hitting a watermelon (waveform components) and some parameters that represent the shape of the watermelon (external dimensions), the judgment is much better than that based on a single parameter. We have improved the accuracy and established an application method through sufficient data collection and experiments, especially regarding the correlation between sound wave characteristics and ripeness. An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing the configuration of the determination device of the present invention. In this figure, 1 is a watermelon, which is transported by a dish-shaped bucket 6 attached to a bucket tension chain 21 of a sorting conveyor 14, Various values for determining ripeness are measured, the ripeness is determined, and the fruits are sorted into predetermined locations. Reference numeral 2 designates an electric motor that provides a driving force for moving (running) the bucket cart 6, and reference numeral 3 designates a chain gear (sprocket) that transmits the force transmitted from the electric motor 2 to the bucket pulley chain 21. 4 is a sensor (camera) that optically captures the shape of the watermelon 1, and the image is generated using reflected light from the lighting device 5, but in some cases it is placed on the opposite side of the sensor (camera). Transmitted light from a lighting device may also be used. (6a) is the state where the bucket is normal (horizontal state)
In the case shown in (6), the bucket tip falls over and the watermelons are taken out (sorted) from the bucket. 7 is a shape measurement position passage detector for controlling the illumination device 5 that shines light on the watermelon 1 and detecting the reflected light from the watermelon 1 and the bucket 6 when the bucket 6 on which the watermelon 1 is placed is located at the shape measurement position. Captured in synchronization with the passing signal from 8,
This is a shape measurement device that performs various types of image processing (image input, noise removal, shape determination, and output of the results). Reference numeral 15 operates a striking device 12 to generate a striking sound on the watermelon 1 in response to a passing signal from a sound measuring position passage detector 9 which detects that the bucket 6 on which the watermelon 1 is placed is located at the sound measuring position. The receiver contactor 11 is moved so that the pick-up 10 is brought into contact with the watermelon 1, so that the receiver 10 picks up the sound waves, and the pick-up 10 is then moved into contact with the watermelon to help move the watermelon after the inspection. A sorting actuator 18 performs a sorting operation to release the watermelons 1 from the watermelon 1 or to dispense (sort) the watermelons 1 at a predetermined position after inspection.
This is a mechanical drive device that controls the sorting conveyor 14 for conveying the watermelons 1 at a constant speed. A sound wave analyzer 13 receives sound waves from the wave receiver 10, analyzes the sound waves (outputs sound pressure levels for each frequency), and sends the results to the comprehensive judgment processing device 16. 16 receives the shape and sound wave data from the shape measuring device 7 and the sonic wave analyzer 13, and from these data comprehensively determines the ripeness of the watermelon, determines the size and ranks it, and further classifies the watermelon. Referring to the position signal from the position detector 17 that detects the movement position, when the determined watermelon reaches the dispensing port of a predetermined ripeness and size class, the mechanical drive device 1 issues a sorting signal.
5 to operate the sorting actuator 18 to overturn the bucket 6 and dispense the watermelons 1. At the same time, if necessary, other peripheral devices, such as a display device 19 for displaying the sorting results or for printing the sorting results, are controlled. This is a comprehensive judgment processing device that controls the printer 20 and the like. In addition, electric motor 2,
The chain gear 3, the bucket 6, the sorting conveyor 14, and the bucket pull chain 21 constitute a conveyance device that sequentially moves and conveys the fruits. The function and operation of one embodiment of the present invention (when the judgment target is a watermelon) will be described below. First, the bucket 6 attached to the sorting conveyor 14
Place the watermelon 1, which is the object to be judged, on top of the watermelon. This way of placing the sample in a specific direction will improve the judgment accuracy because it will align the measurement conditions for the measuring instrument that will be used later. Generally, it is best to position the watermelon in a direction where the vertical axis of the watermelon (the straight line connecting the stem and the navel) is perpendicular to the visual axis of the camera, that is, in a position where it can be seen from the camera as shown in Figure 2A. In addition, in terms of acoustic characteristics, the direction of contact between the striking direction and the receiver is approximately perpendicular to the vertical axis of the watermelon, and when the striking rod or receiver is brought into contact with the watermelon, the contact surface of the striking rod and the receiver is The direction of close contact has the least measurement error. Since the sorting conveyor always moves at a constant speed, the watermelons placed on it pass through the shape measurement position (shape detector = in front of the camera) together with the bucket. At this time, a passing signal is output from the shape measurement position passage detector 8, and this In response to the command from the shape measuring device 7, the camera 4 sends buckets 6 and watermelons 1.
The image is taken into the shape measuring device 7, and the shape dimensions (shape parameters) of each part of the watermelon 1 are measured by measurement circuit processing. In the case of watermelon, it was found through data analysis that the height 25 and equatorial diameter 24 shown in Figure 2 A have a strong correlation with sugar content (Table 1), so they can be used as shape parameters when determining the sugar content of watermelon. measures the height H25 and the equatorial diameter D24. When the bucket 6 moves again and the sound measurement position passing detector 9 is activated, it indicates that the watermelon 1 has come to the sound measurement position, so the mechanical drive device 15 receives the passing signal and operates the striking device 12 to remove the watermelon. A hitting sound is generated at the watermelon 1, and the receiver contactor 11 is operated immediately before the hitting sound is generated to bring the receiver 10 into close contact with the watermelon 1. At the same time as the hitting sound is generated, this sound wave is transmitted to the sound wave analyzer 13. Incorporate into. The sound wave analyzer 13 detects the sound pressure level for each frequency, and uses the waveform discrimination circuit included in the comprehensive judgment processing device 16 to determine the sound pressure level for each frequency.
The respective frequencies and sound pressure levels of the first resonance point 26, second resonance point 27, and first anti-resonance point 28 in Figure B, that is, Hz ₁ , dB ₁ , Hz ₂ , dB ₂ , Hz ₁ - ₂ , dB ₁ ― ₂
Measure. (The above sound-related coordinates are called sound parameters.) Here, talking about the relationship between the method of measuring sound parameters and ripeness, each individual watermelon has its own natural frequency, and the qualities that make up each watermelon do not change. As long as the natural frequency remains the same.
Also, when a watermelon is shocked, it vibrates (sounds) at its own natural frequency. It is known that this can be used to empirically determine the internal state of a watermelon, and people have been using this to judge ripeness for a long time, but this has been further elucidated and applied technically. There have been no examples yet. Next, I will explain the relationship between sound and ripeness. The watermelon's unique vibration frequency mentioned earlier can be measured by the sound produced by hitting a watermelon, but when measured using this method, the waveform varies depending on the shape (and size) and sugar content. Figure 3 shows a typical example, with the horizontal axis representing frequency and the vertical axis representing sound pressure dB. Figure 3 (A) is an experimental example showing that even watermelons of the same size have different waveforms when the sugar content is different, and Figure (B) is an experiment that shows that even if the sugar content is almost the same, the waveforms of watermelons of different sizes are different. This is an example. In addition, in Figure 3 A, the solid line indicates sugar content α=
11.0, size is class M (medium), weight 5.27Kg, equatorial diameter D = 218mm, height H = 208mm, the dashed line indicates sugar content α = 10.2, size is class M (medium), weight 5.30
Kg, equatorial diameter D = 213 mm, and height H = 220 mm. In addition, in Figure 3 B, the solid line indicates sugar content α = 9.4, size class L (large), weight 6.33 kg, equatorial diameter D = 230 mm, height H = 233 mm, and the broken line indicates sugar content α = 9.2, Size is class SS (extremely small), weight
The one shown is 3.86Kg, equatorial diameter D = 200mm, and height H = 192mm. As you can see from the diagram, in order to understand the characteristics of such a sound waveform, if you focus on the positions of the peaks (first and second resonance points) and valleys (first anti-resonance point) of the waveform,
Very convenient. The inventors discovered such resonance points 26, 27,
We also determined the coordinates of the anti-resonance point 28 and used them as sound parameters. These parameters alone, ripeness, and correlation are shown in Table 1, and the significance test for each parameter and the presence or absence of nests (cavities) is shown in Table 2
Shown in the table.

【table】

[Table] Tables 1 and 2 show the results of judgments made using each parameter independently to estimate certain elements (sugar content and presence or absence of nests). The method of determination in this case will be explained with reference to FIG. In FIG. 5, the abscissa represents one parameter of the watermelon, and the ordinate represents the number of individuals (=frequency) that takes a certain value for that parameter. Two types of frequency distribution diagrams for samples representative of a certain watermelon population are shown as examples in Figure 5 (a) and (b). In this figure, there are two estimation targets (types) a,
(b) Two cases are shown: (a) where the estimation targets are sufficiently far apart from each other, and (b) where the estimation targets cannot be separated. In the case shown in Figure 5 A, it is impossible for different types to take the same parameter value, that is, the frequency distributions 31 and 32 for parameter A of the two types a and b do not intersect with each other. If you separate the a and b types by a dividing line (dividing line) like α ₀ in the figure, you can 100% distinguish between the two types.
On the other hand, when the two types of frequency distribution diagrams intersect as shown in (b), no matter how the dividing line α is set, one dividing line cannot 100% correctly discriminate between types a and b. Since a statistical method for determining the optimal position of α ₀ has been established, even if the position of α that minimizes misclassification is set using this method, in the case of Figure B, the range 33 where type a is mistaken for type b is There is a range 34 where type b is mistaken for type a. If there is no error in the classified items, but if you draw a dividing line that suspends the classification of items that may be misclassified (i.e., makes them undeterminable), then it will be at the positions α ₁ and α ₂ in B of the same figure. By drawing a dividing line, it can be determined that the area to the left of α ₁ in the diagram is the 100% correct species a, and the area to the right of α ₂ is the 100% correct species b. However, with this method of drawing two dividing lines, the range 33 where the two peaks (frequency distribution map) intersect with each other is sufficiently far apart.
If 34 is small, there is no problem, but if it is large, most of the things to be discriminated are caught between α ₁ and α ₂ and cannot be judged, which means that the amount of correct discrimination is too small to be practical. The above example describes a case where there are two estimation targets for one parameter, but if there are three or more estimation targets, the number of dividing lines may be increased as the number of estimation targets increases. Generally, there are few cases in which it is possible to determine which group a watermelon belongs to using one parameter. Currently, a judgment device that uses a single parameter to judge the ripeness of fruits and vegetables has been put into practical use, but this method can only be applied in very limited cases under extremely favorable conditions, making it extremely difficult to popularize it. Running late. The inventors researched how to combine two or more parameters for successful discrimination and developed a specific method. The method of using three or more parameters is an extension of the idea of the method of combining two parameters, but since the explanation is very complicated, the case of two parameters will be explained with reference to FIG. 6. In FIG. 6, A and B are parameters, and N is a frequency. If A and B are respectively set on the horizontal horizontal axis and horizontal vertical axis, and N is set on the vertical axis, the frequency distribution of two general types a and b is illustrated as shown in FIG. 6A. A plan view of this figure (a view viewed from the direction of the coordinate axis N) is shown in FIG. When this A is projected onto the vertical plane AN, it becomes like R, which is the same as that shown in FIG. In the same way, the figure obtained by projecting A onto the BN plane is Low, and this is Figure 5 when A is replaced with B, so we will explain either one (Low in this case). The same can be said for the other case, so we will explain R and R here. The discrimination based on only one parameter corresponds to the case of dividing by α ₀ -α ₀ ' in the rho diagram, but in this case, the portions that are erroneously determined are the ranges indicated by diagonal lines 33 and 34 in the diagram. Similarly, when the row diagram is divided by β ₀ -β ₀ ', the parts that are erroneously determined are the shaded parts 47 and 48. On the other hand, if we consider two parameters and draw the optimal dividing line using statistical discriminant analysis, we get γ ₀ -
γ ₀ ′. In this case, the erroneous judgment area is the area indicated by the horizontal hatched line (parallel to the OA axis) in the diagram, and as is clear from the diagram, the range (33+34) >
The range is (49+50). In other words, it can be said that the γ ₀ - γ ₀ ′ division has a narrower range than the α ₀ -α ₀ ′ (or β ₀ -β ₀ ′) division, that is, there are fewer misjudgments. As explained above, two variables (parameters)
It has been found that by combining the parameters, it is possible to perform discrimination (division) with a higher correct judgment rate than discrimination based on each parameter individually. The same thing can be said even if the number of variables is increased to three or more, and the correct judgment rate further increases, but if the number of parameters is increased to a certain extent, the increase in the correct judgment rate slows down, so it is easier to control the process and improve cost performance. In practice, the number of variables is limited to an appropriate number, taking into consideration the following.
Further, when the number of estimation targets increases to three or more, the same effect can be achieved by applying the discriminant analysis method stepwise. Judgment of sugar content based on a single parameter (whether it is above or below the standard value) gives the correlation coefficients shown in Table 1, and even in the case of D, which has the highest correlation coefficient, the correct judgment rate is about 75%. This is controlled by many parameters (e.g. D, Hz ₁ ,
dB ₁ , Hz ₂ , dB ₂ ), the results are shown in Table 3.

[Table] By the way, the parameters dB ₁ - ₂ and Hz ₁ - ₂ are not used in this case, but using them will certainly increase the correct judgment rate, but since it is only about 1 to 2%, we do not force them and simplify the judgment formula. The focus was on making it possible. Similarly, for the determination of a single parameter in Table 2, if several parameters are used for determination, the fourth
Good results as shown in the table.

[Table] When selecting parameters, choose one parameter that is effective for determination, or one that is effective when combined with two or more parameters, and select the minimum number so that the determination process does not become too complicated. Note that it is convenient to measure sound parameters as follows. In the case of a watermelon, it is good to have the first and second resonance points and the first anti-resonance point, whereas the normal hitting waveform of a normal watermelon has only the first and second resonance points and the first anti-resonance point (the fourth (See Figure A) However, in rare cases, the waveform may be distorted and become waveforms such as those shown in Figure 4 B and C. In this case, there are 3rd, 4th, etc. resonance points and 2nd, 3rd, etc. anti-resonance points, and their positions are considerably shifted compared to the normal case in Figure 4A. In order to measure these correctly, if you set a frequency range and find the maximum value (the largest of several local maximum values) or minimum value within this range, in most cases you can find the sound parameters as in case A. I found out that I can handle it. Figure 4 shows an example of an actual watermelon.The first resonance point is the frequency and sound pressure that take the maximum sound pressure of 51 between 20Hz and 50Hz, and are Hz ₁ and dB ₂ , respectively. The frequency at which the maximum sound pressure occurs and the sound pressure at that time are respectively Hz ₂ and dB ₂ , and the frequency at which the minimum sound pressure occurs within the range of 53 between 50Hz and 200Hz and the sound pressure at that time are Hz ₁ - _2. ，
If it is set to dB ₁ - ₂ , it can be set as a universal parameter for almost all watermelons. This is true for the first and second waveforms in Figure 4 B and C.
It can be seen that the peaks and valleys of the resonance point and the first anti-resonance point can be considered to be cracked (separated). In the above example, two measurement devices are used: one for shape measurement (shape detector (camera) + shape measurement device) and the other for sound wave measurement (wave receiver + sound wave analyzer), but this configuration is applicable to plants other than watermelon. It can also be applied to other fruits such as melons. In addition, in the above embodiment, the sorting conveyor is always driven at a constant speed, but if the conveyor is stopped at a certain point when the object arrives at the inspection position, the sorting speed will be reduced, but it will be faster. Reliable judgment and selection becomes possible. Further, in the above description, data from the measuring instrument is input to the comprehensive judgment processing device, but it may be input from another input device. As described above, according to the present invention, the geometric dimensions of the fruit (for example, high This also has a large correlation coefficient with the ripeness of the fruit, and is based on the sound pressure and frequency of the resonance or anti-resonance point of the sound generated by hitting the fruit. It is remarkable that the ripeness of a large amount of fruit can be determined in a short time and with high accuracy by drawing an optimal dividing line that narrows the range where frequency distributions overlap, that is, by using a statistical method to find the optimal criteria. It has a great effect. Moreover, there is a remarkable effect that the ripeness of the fruits can be automatically and continuously determined.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the determination device according to the present invention, Fig. 2 is a diagram showing shape parameters and sound parameters among the parameters used to judge the ripeness of watermelon, and Fig. 3 is a diagram showing the ripeness of watermelon. Figure 4 shows how the sound waves that determine sugar content, which is one of the sugar levels, changes depending on the sugar content and shape (size). Figure 5 is a waveform diagram for explaining the measurement method used, Figure 5 is a diagram for explaining the case where two types of estimation targets are divided by a single parameter, and Figure 6 is a diagram for explaining the case where two parameters are combined for determination. 1 is a watermelon,
4 is a camera, 6 is a bucket, 7 is a shape measuring device,
8 is a shape measurement position passing detector, 9 is a sound measurement position passing detector, 10 is a wave receiver, 11 is a wave receiver contactor,
12 is a striking device, 13 is a sonic analyzer, 14 is a sorting conveyor, 15 is a mechanical drive device, 16 is a comprehensive judgment processing device, 17 is a position detector, and 18 is a sorting actuator.

Claims (1)

[Claims] 1. A conveyance device that sequentially moves and conveys fruits; A shape measurement position passing detector that detects that the fruit is located at a shape measurement position; A detector that detects that the fruit is located at a sound measurement position. a sound measurement position passing detector; a shape measuring device for measuring the geometric dimensions of the fruit based on the passing signal from the shape measuring position passing detector; a striking device; a receiver that captures the sound generated by striking the fruit with the striking device; It is characterized by comprising a sound wave analyzer that measures frequency, and a comprehensive judgment processing device that judges the ripeness of the fruit by a statistical method based on the geometric dimensions, the sound pressure, and the frequency. Judgment device.