JP2004016945A - Method and apparatus for sorting object - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and an apparatus for sorting objects, by which the various objects such as farm and marine products to be sorted can be sorted according to the individual or common characteristics of the objects to be sorted for the purpose of uniforming processing conditions, sorting according to application, obtaining differentiated goods and improving storage and transportation conditions. <P>SOLUTION: The object to be sorted is sorted precisely according to the actual conditions of a sorting purpose and a change of a probability distribution of the characteristic values by analyzing the probability distribution of the characteristic values obtained for sorting the object to be sorted, deciding a threshold value corresponding to the target value on the basis of the analyzed result under open-loop control and changing parameters of the probability distribution and the calculation range and the weight of each of the characteristic values. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
[Industrial applications]
The present invention is intended to sort or sort various objects, including agricultural and marine products, individually or collectively for the purpose of aligning processing conditions, dividing applications, making differentiated products, and improving storage and transport conditions. It is related to the technology of sorting according to the characteristics of the product, and is different from the sorting that sorts according to a predetermined standard, and is related to a sorting method and a sorting device for a product whose main purpose is to obtain a certain amount as a result of sorting. Things.
[0002]
[Prior art]
Many things, including agricultural and marine products, undergo sorting operations during production and distribution. The selection is performed based on a preset standard. The standard is provided in order to satisfy the necessity of distribution, and is expressed by a characteristic value that allows some measurement (including evaluation judgment by the five senses and the like, hereinafter simply referred to as measurement, etc.). Agricultural and marine products, for example, the characteristic values of fruits, such as weight, size, sugar content, acidity, etc., are selected according to the probability distribution (see FIG. 3... Represented by a histogram in practice, and mathematically represented by a probability density function). The quantity will follow a probability distribution function (integrating the probability density function from -∞, which is practically equivalent to a cumulative frequency (probability) graph), but in many cases this is not considered.
[0003]
Even if a standard is not set, classification based on some characteristic value increases equipment efficiency, aligns processing conditions, divides applications, differentiates products, stores and transports, etc. Benefits arise when conditions are improved. In such a sorting operation, it is often more significant to have a requirement for a quantity or a quantitative ratio than an absolute requirement for a characteristic value. If the probability distribution function of the characteristic value is known, a characteristic value (hereinafter, referred to as a threshold value) serving as a boundary for separating a required quantitative ratio can be calculated. However, in agricultural and marine products, the distribution of characteristic values varies from day to day, depending on the place of production, and may be supplied in a mixed state with different histories, so that the probability distribution is not constant. Shown). Probability distribution functions can be estimated if the characteristic values can be obtained by sampling uniformly from the objects to be sorted.However, at the processing and collection sites for these products, the collection time has a wide range, and the collected items are limited. Since the processing is required at the same time, it is not desirable to sample uniformly from the whole. Also, even if the probability distribution itself is stable, a statistical bias always occurs, and the end of the distribution is only a tenth of the whole (that is, in FIG. 3, a characteristic value 20.5 corresponding to a probability of 0.1). Sorting out can be foreseen to be difficult before trying. Under such circumstances, no effort has been made to sort on the basis of quality evaluation and to operate the threshold for sorting based on the deviation of the quantitative ratio of the sorting result from the target.
[0004]
[Problems to be solved by the invention]
When the supplied sorting objects are divided into a plurality of groups (hereinafter simply referred to as ranks) with the same characteristics, if the goal is to secure a quantitative ratio, the quantitative ratio can be calculated from the probability distribution function. Is obtained as a characteristic value (reference threshold value), and this is sorted as a threshold value. When the quantitative ratio of the rank is insufficient, the range of the characteristic value to be assigned to the rank is expanded, and the amount gradually increases.On the other hand, when the quantitative ratio is excessive, the range of the characteristic value is narrowed. The amount should gradually decrease and approach the target ratio. This is a concept corresponding to feedback control (in other words, closed loop control) in control engineering. However, during the sorting, the probability distribution state of the characteristic value changes, and the threshold value to be used as a reference point for feedback control changes. Further, even if the probability distribution function does not change, it is inevitable that the appearance of the characteristic value is biased because it is a stochastic phenomenon, and it is understood that it is necessary to accurately grasp the state of the probability distribution separately from the closed loop control. The threshold value of this reference is required to follow a change in the probability distribution and not to follow a temporary bias in the probability distribution. In general, in control engineering, a control model is represented by a state equation, which is a function of time. In the case of the present invention, strictly speaking, the order is advanced, and there is no time progress. Instead of the state equation, it is expressed by a probability function as described above. The control target is not a quantity but a stock quantity as a flow. That is, the past control results are included in 100% current control results. Nevertheless, changing the order of supply will change the results. For the sake of convenience, terms commonly used in control engineering are used, but knowledge of control engineering cannot be applied as it is, and control methods must be examined based on probability theory.
[0005]
[Means for Solving the Problems]
As a key to harmonizing conflicting objectives, equalizing characteristic values and securing quantities, first, the frequency of appearance of characteristic values obtained by measurement or the like for sorting (in other words, the probability density function) is used to calculate the cumulative frequency (in other words, (Probability distribution function), and the threshold value at that time is calculated. Since no control error is observed, it corresponds to open-loop control instead of closed-loop control. If this alone has little fluctuation in the distribution, or if you sort a lot, you will naturally approach the target. If this is not the case, in order to reduce the error if necessary, this is used as a reference point for the feedback control, and the closed loop control is incorporated as shown by the broken line in FIG. Next, it is required that this reference point does not follow the stochastic temporary bias, but follows the change of the probability distribution function.If the shape of the distribution is known, for example, In the case of a normal distribution, a probability distribution function (corresponding to the cumulative probability graph of the normal distribution in FIG. 3) is determined using the average and the standard deviation as parameters, and a characteristic value corresponding to an arbitrary sorting ratio is obtained. Adopt the threshold of When the change of the probability distribution function is relatively small, (1) when the supplied sorting target is divided into a plurality of groups with the characteristics being aligned while maintaining a quantitative ratio to a required target value, the sorting is performed by the time of sorting. A parameter of the probability distribution is calculated using all of the obtained characteristic values, and a threshold value serving as a sorting reference is determined based on the parameter. If the change in the probability distribution function is relatively large, (2) when the supplied sorting target is divided into a plurality of groups with the characteristics being uniform while the quantitative ratio is set to a required target value, until the sorting point The parameters of the probability distribution are calculated using the data limited to the latest predetermined number among the characteristic values obtained in step (1), and the threshold value serving as a reference for sorting is determined based on the parameters.
[0006]
Next, in combination with these means, (3) the characteristic value is calculated with equal weights in the above-described parameter calculation process. Or (4) calculate a new characteristic value with a large weight. In general, a geometric series is used as a coefficient of the weight, and the sum of the coefficients is 1. Although the case of normal distribution has been described above, if the shape of the distribution is not known (it may be known), (5) the time of sorting is replaced with the calculation of the parameters of the probability distribution function. The reference threshold value corresponding to the target ratio is determined based on the cumulative frequency curve of the characteristic values obtained up to this point. In other words, the obtained characteristic values are rearranged in ascending order to create a cumulative frequency graph (see FIG. 3, actual cumulative graph), and a regression line (it is easy to linearize with an inverse function of the probability distribution function). A characteristic value corresponding to an arbitrary sorting ratio is directly calculated by fitting or simply performing interpolation calculation or extrapolation calculation, and this is used as a reference threshold value. This method can be applied to a case where all the characteristic values obtained up to that point are used, and a case where only the latest fixed number of data among all the characteristic values obtained up to that point are calculated. In general, whether or not a normal distribution is performed is determined using normal probability paper. In fact, even ratios of highly correlated components, such as the N / F index of fresh tea leaves, may be considered a normal distribution. This technique, which does not care about the shape of the distribution, has versatility. If the amounts to be sorted are extremely irregular, or if there is a correlation between the amounts and the characteristic values, (6) the characteristic values are weighted by the amount to be sorted and calculated to accurately grasp the distribution. Although the occurrence frequency is low at both ends of the probability distribution, characteristic values that are abnormally far apart may occur. (7) The characteristic value is limited to a certain range from the previous average value or the like (or a substitute median value). (For example, the data is restricted so as not to exceed about two to three times the standard deviation from the previous average value. In FIG. 2, data of Nos. 11, 24, 57, 64, and 82 are restricted.) calculate. This restriction may be applied to a measured value that is an element of the characteristic value.
[0007]
Further, since there is no reference threshold value at the start of sorting and the calculated parameter is unstable for a while, (8) the parameter of the probability distribution function or the threshold value of the reference is used as an initial value at the start of sorting. And gradually shifts to the reference threshold value obtained from the characteristic values obtained up to the time of sorting. That is, an appropriate initial value is given, and the parameter is gradually shifted to the parameter based on the measurement data as the number of calculation data increases. Alternatively, (9) at the start of sorting, a characteristic value is obtained for a sample of an object in the same manner as for sorting, and a reference threshold value is given. In other words, instead of giving the initial value, the characteristic value of the initial consignment or the sampled sample is determined to determine the reference threshold value, and thereafter, the operation proceeds to the sorting operation. In this case, in a situation where the change of the standard deviation is small, it is sufficient that the average value is known as the parameter, so that 5 to 10 pieces of data can function sufficiently. When the actual data is analyzed, the daily fluctuation of the standard deviation is slight, and the uniformity of the characteristic values is better if the sorting control is performed by regarding the fluctuation as a constant. This is because the probability phenomena of the characteristic values are not only biased with respect to the average value, but also there are phenomena that are extremely concentrated near the average value.
[0008]
(10) Object quality measuring means, object quantity measuring means, and characteristic value distribution obtained up to the time of sorting are analyzed to obtain a threshold, and the target is determined based on the threshold. An apparatus for sorting objects according to a probability distribution, comprising sorting control means for sorting and transport means for transporting an object based on the result of the sorting control means, is used (see FIG. 6).
[0009]
Although the processing method in the case where the sorting amounts are extremely irregular in (6) has been described, as a better method, the parameters obtained by dividing the sorting object into a plurality of times and performing measurement or the like should be used. The division may not be strict, but it is desirable that the amount of the target object per measurement or the like is almost the same. Although the sorting operation and the error calculation always correspond to each other, these need not necessarily correspond to the division measurement. Note that the characteristic value can often be expressed as a one-dimensional quantity, but may be captured in a two-dimensional manner as shown in FIG.
[0010]
[Action]
In the present invention, the cumulative frequency (in other words, the probability distribution function) is obtained from the frequency of occurrence (in other words, the probability density function) of the characteristic value obtained by measurement or the like for sorting, and the threshold value at that time is calculated. Since this is used as a reference point for feedback control, even if the probability distribution changes and an error occurs in the quantitative ratio of sorting, first, the reference point for feedback moves to the point where the target amount is obtained, and A feedback function for correcting and recovering an error can be added to the base point (see FIG. 5). Then, by means of (1), the threshold value of the reference is sequentially determined by using the parameters of the probability distribution function calculated using all of the characteristic values obtained up to the time of sorting, so that the most stable reference value is obtained. Is obtained. Alternatively, by means of (2), a reference threshold value is sequentially determined by using a parameter of a probability distribution function calculated from data limited to the latest predetermined number among characteristic values obtained up to the time of sorting, A reference threshold value that can easily follow changes in the form of the probability distribution is obtained. Although it is a so-called moving average, the data before and after cannot be used and only the data before that time is used, so there is an inherent delay, but the tracking is faster than (1) and follows the same direction as the feedback. Since the value is changed, there is an effect of substituting feedback. In particular, if the number of data is reduced, it becomes easier to follow sudden changes in the distribution.
[0011]
Next, in combination with these two means, by means of (3), the characteristic value obtained by each measurement or the like is calculated with an equal specific gravity in the above-described parameter calculation process, so that the shape of the distribution is stabilized. The most accurate reference threshold is obtained. Alternatively, by means of (4), it is possible to adjust the response to the change in the probability distribution by how large a new characteristic value is calculated in the parameter calculation process. The method of calculating the average of the characteristic value of the previous time point and the current characteristic value at a fixed ratio is a simple calculation method and the adjustment of the reaction speed is easy (a modification of the calculation by geometric series) ... see 9: 1 average in Figure 2). By means of (5), the characteristic values obtained by measurement or the like are rearranged from the smaller one to create a cumulative frequency graph, and a regression line is applied to the graph, or an interpolation calculation or an extrapolation calculation is performed. A characteristic value corresponding to the sorting ratio is calculated, and this can be used as a reference threshold value. This method can be applied even when the shape of the probability distribution cannot be specified. The method of drawing a regression line can estimate parameters without being distracted by abnormal data at both ends of the distribution or a large amount of data. By means of (6), even when the amounts to be sorted are not uniform or when there is a relationship between the amounts and the characteristic values, the data to be calculated is weighted by the amount of sorting to calculate the appropriate reference threshold value. Obtainable. By means of (7), abnormal data is limited to a certain range, so that the fluctuation of the average value becomes stable.
[0012]
By means of (8), it is possible to prevent parameters from fluctuating greatly even while the number of data in the initial stage of sorting is small. Further, even when an appropriate initial value cannot be given, the sorting can be started again after obtaining data of the sorting target object without performing the sorting operation by the means (9). It can be carried out. By means of (10), the quality measuring means of the sorting object, the measuring means of the amount of the object, and the distribution of characteristic values obtained up to the time of sorting are analyzed to obtain a threshold value. This can be accurately performed by a facility including sorting control means for sorting the objects based on the threshold value and transport means for transporting the objects based on the result of the sorting control means.
[0013]
BEST MODE FOR CARRYING OUT THE INVENTION
An example in which the present invention is applied to a tea factory will be described based on the block diagram of FIG. Fresh tea leaves brought into the tea factory are measured by quality measuring means. The timing of the measurement differs depending on the factory, and may be before or after being put into the receiving device. As a quality measuring means, besides a method of measuring the components of fresh tea leaves, physical properties such as bulk density and hardness, size and color, and generally appearance characteristics by the five senses may be measured. Before and after the measurement, items that are not to be measured, such as varieties, plucking methods, cultivation methods, pests and the like, are determined (these may be treated numerically) and input to the measurement control means. The measurement control means collects the measured values of the quality and the data of the other judgment items and sends them to the sorting control means and the accounting system. In the accounting system, the quality judgment value is determined by integrating them, and is used to determine the purchase price of green tea leaves. On the other hand, the sorting control means serves as data for sorting into groups with uniform characteristics. There are cases where special classification is required only for some of the judgment items.However, in the following description, the case where the characteristic values are combined with measured This will be described as an example.
[0014]
As a quantity for the transaction, the measurement of the accepted weight is generally performed for each vehicle before and after the carry-in, but separately from this, the sorting weight may be integrated as processing data in association with the measurement. The weight data is sent to the sorting control means and the accounting system. The sorting control means stores the characteristic values and the measured weight values as corresponding data via the reception numbers of both data and the like. The sorting control means calculates an average value and a standard deviation of the stored characteristic values, and determines a reference threshold value for each rank based on the average value and the standard deviation. (Even if the calculation of the average value and the standard deviation and the determination of the reference threshold value are performed after the determination of the sorting and applied to the next sorting, there is no practically significant difference. Even if only the data up to is obtained, it can be controlled practically.)
[0015]
The sorting control means determines which rank is to be sorted by comparing the characteristic value with a threshold value corresponding to each rank set according to the processing necessity, and controls the transfer device. The threshold value is a reference. This is obtained by adding a feedback operation amount described later to the threshold value. Here, if it is necessary to limit the fluctuation range of the threshold value and guarantee the fluctuation range of the characteristic value, the limit is set to the range that the threshold value can take. Even in the case where such a need is not particularly required, if the values are divided into three or more ranks, two thresholds may interfere with each other.
[0016]
Further, the sorting control means integrates the weights sorted by rank. Further, the weight of the brought-in fresh tea leaves is integrated, and the required sorting ratio of each rank is multiplied to calculate the required weight for each rank at that time. Then, an error with respect to the required weight of the sorting integrated weight for each rank is calculated. This error is multiplied by a control gain corresponding to the sorting ratio to calculate a threshold feedback operation amount.
[0017]
At the start of the control, an initial threshold value is required. The value is determined based on the past actual value or the forecast of the tea garden for determining the picking start date. If there is characteristic value data that can be referred to, it can be rearranged in ascending order to find a judgment value that provides a required quantitative ratio. The transition of the threshold value from the initial value to the actual value is performed gradually and gradually.
[0018]
The present invention is based on the above-described embodiment.However, when a calculation method in which a change in a reference threshold value is sensitive to a probability distribution is selected, feedback control based on excess or deficiency of weight is performed. Without performing this, simply resetting the reference threshold value automatically according to the transition of the average value and standard deviation, or simply resetting the reference threshold value that can be determined directly from the cumulative frequency, The quantitative balance when divided can be kept within the allowable range of the equipment. The same applies when the probability distribution hardly changes.
[0019]
Further, in the present embodiment, as shown in FIG. 6, the case where the quality measurement, other determination, the weight measurement, and the transport device are one system is described. Further, even if they are remote places and are connected to the sorting control unit by a communication line or the like, they can be executed integrally.
[0020]
As mentioned above, sorting of collected tea leaves has been described.In addition, processing conditions are aligned, applications are separated, differentiated products are improved, storage and transportation conditions are improved, and such sorting technology is also used. There are many fields in which it can be applied, and not only agricultural and marine products but also those that follow a probability distribution can be similarly sorted. Here, an example of agricultural and marine products is described as one of the examples. As a recent trend, producers have seen an increase in direct sales to consumers, but in this case, the idea that direct sales customers have the highest priority and that they want to secure the shipment volume from high-quality products works. When it comes to deciding what percentage of the harvest will go to direct sales, you need to think about "sorting" instead of traditional sorting. Conversely, not only major distributors, but also food service and ready-to-eat meal businesses are procuring to bypass market functions. In order to obtain what they need practically and economically when dealing directly with producers, it is reasonable to set up transaction conditions that are in line with the actual conditions of the products. Can accelerate the realization. It can be pointed out that most products are undersupply under such specialized movements. Under these circumstances, the production area is very sensitive to market price trends, and information has been promoted by shipping organizations. Adjustments are inevitable. For example, in the case of fruits, since there is a presence or absence of suitability for storage separately from the standard of shipping, sorting in which the ratio of shipping and storage is determined can be used together with the fruit selection step. In addition, when it is possible to select between raw food and processing such as juice and canned food, it is possible to adjust the amount by sorting and controlling boundaries within and outside the standard for raw food. Further, taking canning as an example, it is possible to sort and adjust the ratio of cut and whole products according to the production plan within a range that is in proportion to the size of the can. Sorting objects to which the sorting control described above can be applied are wide-ranging, but examples of common products include tomato, asparagus, potato, onion, melon, citrus, apple, grape, peach, Plums, ume, kuri, bonito, saury, eel, salmon, mackerel, sardine, or fish fillets, shellfish, eggs, rice, brown rice, milled rice, and the like.
[0021]
【The invention's effect】
The present invention has the following effects by the above configuration. Calculate the parameters of the probability distribution from the characteristic value data measured up to that point for sorting, and adjust the range and weight of the data to be calculated when resetting the threshold as a reference for sorting based on the parameters. Thus, it is possible to accurately follow the change in the probability distribution of the characteristic value. Even when the form of the probability distribution function cannot be specified, the reference threshold value can be similarly reset from the cumulative frequency curve, and the sorting can be started smoothly by giving an appropriate initial value at the start of the sorting.
[Brief description of the drawings]
FIG. 1 is a conceptual diagram showing a method of sorting objects according to a probability distribution according to the present invention with a ratio as a target, in correspondence with general control engineering.
FIG. 2 is a diagram illustrating an example of a change in a characteristic value of a supplied material and a change in a standard deviation value calculated by movement, and also illustrates an average value recalculated each time sorting is performed.
FIG. 3 is a histogram showing the distribution of characteristic values in the example of FIG. 2, superimposed on a normal distribution drawn based on a calculated average value and a standard deviation, and a probability distribution obtained by integrating a density function from −∞. The figure which also showed the function (cumulative probability). (However, the actual cumulative probability graph that overlaps this curve is a cumulative graph weighted by the sort weight.)
FIG. 4 is a conceptual diagram for sorting a two-dimensional distribution having correlation.
FIG. 5 is a conceptual diagram of a relationship between a sorting error and an amount by which a threshold value is shifted from a reference threshold value, that is, a control gain.
FIG. 6 shows a block diagram of a control system.

Claims (10)

Calculate the parameters of the probability distribution using all of the characteristic values obtained up to the point of sorting when dividing the supplied sorting target into multiple groups while keeping the quantitative ratio to the required target value. And determining a threshold value as a reference for sorting based on the parameters. When sorting the supplied sorting target objects into a plurality of groups while keeping the quantitative ratio to the required target value, the data limited to the latest predetermined number of the characteristic values obtained up to the time of sorting A method of calculating a parameter of a probability distribution using the parameter, and determining a threshold value serving as a reference for sorting based on the parameter. 3. The method according to claim 1, wherein the characteristic values are calculated with equal weights. 3. The object sorting control according to claim 1, wherein a new characteristic value is calculated with a large weight. 3. The method according to claim 1, wherein a reference threshold value corresponding to the target ratio is determined based on a cumulative frequency curve of characteristic values obtained up to the time of sorting, instead of calculating a parameter of the probability distribution function. 3. The method for sorting objects according to 3 or 4. 6. The method according to claim 1, wherein the characteristic values are weighted by an amount to be sorted. 7. The method according to claim 1, wherein the characteristic value is calculated by limiting the characteristic value to a certain range from a previous average value or the like. At the start of sorting, a parameter of a probability distribution function or a threshold value of a reference is given as an initial value, and a gradual transition to a reference threshold value obtained from characteristic values obtained up to the time of sorting is performed. Item 1, 2, 3, 4, 5, 6, and 7. 4. The method according to claim 1, wherein at the start of the sorting, a characteristic value is obtained for the sample of the object in the same manner as for the sorting and a reference threshold value is given. 7. The method for sorting objects according to 7. Object quality measurement means, object quantity measurement means, measurement control means, and parameters of probability distribution are calculated using characteristic values obtained up to the time of sorting, and the parameters serve as a reference for sorting. An object sorting apparatus comprising: sorting control means for determining a threshold value; and transport means for transporting an object based on the result of the sorting control means.

Images