JPH08133449A - Detection system for specific foreign matter in object to be conveyed - Google Patents

Abstract

PURPOSE: To provide a detection system for specific foreign matter in an object to be conveyed in which particular foreign matter is detected, by integrating some sensor information with large errors. CONSTITUTION: This detection system is provided with various sensors 2-8 for detecting an object in bulky refuse which is conveyed on a conveyance conveyer 1. The weight, length and metal information of the object are obtained by the sensors 2-8. The information is processed by the fuzzy inference part and the neural net inference part of an inference part 15 in parallel. By integrating the two types of inference results, it is possible to detect a specific foreign matter such as a fuel gas cylinder.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for detecting a specific foreign substance in a conveyed product, and is particularly useful for detecting a specific explosive dangerous substance such as a gas cylinder to be removed before crushing processing in a bulky refuse processing facility It is something.

[0002]

2. Description of the Related Art In a bulky waste treatment facility for recycling bulky waste, collected waste is crushed and valuable materials such as iron, aluminum and copper are separated and reused. In the present circumstances,
Explosive hazardous materials such as LPG cylinders may be mixed in the crushed waste, and many processing facilities experience explosion accidents and associated fire accidents. There have also been reports of cases that led to the suspension of the.

For this reason, in many facilities, in a receiving line for the waste collected before the crushing process or on a conveyor line for transporting the crusher to the crusher, these dangerous substances are detected and removed visually by the operator.

[0004]

However, the collected dust has various shapes, compositions, weights, and other forms, and is a steel that causes the most serious disaster in the event of explosion of dangerous substances. Even if it is limited to the fuel gas cylinder made of a product, it is difficult to detect it completely by human eyes because there are many kinds of detection targets.

Particularly, a small-sized steel LPG gas cylinder may be collected while being placed in a corrugated cardboard or a bag, and such a current situation becomes an important problem in maintaining the facility and ensuring the safety of workers. ing. Further, since the work of removing dangerous substances is performed in a noise or dust atmosphere, improving the working environment is also an issue.

From such a background, the advent of a device for automatically detecting and removing dangerous substances such as gas cylinders that are mixed in oversized waste is desired.

However, since the collected dust has various properties, the conventional detection system using the deterministic relationship between the target and the sensor output cannot be put to practical use because of a large error and poor reliability.

In view of the above-mentioned prior art, the present invention integrates several sensor information with large error to detect a specific foreign substance, and by applying this, the explosive dangerous substance is automatically generated in a bulky waste treatment facility. It is an object of the present invention to provide a system for detecting a specific foreign substance in a conveyed product, which can construct a system for removing dust.

[0009]

Means for Solving the Problems The structure of the present invention for achieving the above-mentioned object is a conveyor for conveying a conveyed object such as bulky garbage, and a conveyor arranged along the flow direction of the conveyor. An inference unit that detects specific foreign matter in a conveyed object by integrating various sensors that detect various properties such as weight, length, and material, and processing the output information of the various sensors based on a predetermined inference procedure. And having.

[0010]

According to the present invention having the above-described structure, each sensor detects various properties of the conveyed product. The inference unit integrates the output information of the respective sensors to estimate and determine whether or not the specific foreign matter is present in the conveyed product.

[0011]

Embodiments of the present invention will now be described in detail with reference to the drawings.

This embodiment is a detection system for a fuel gas cylinder which is a specific foreign matter in a bulky refuse which is a conveyed object.
[Fig. 3] is a block diagram showing its overall configuration.

As shown in the figure, the large-sized waste placed on the conveyor 1 passes through the sensors 2, 3, 4, 8, 5, and 6 in order and is a dangerous substance (fuel gas cylinder) in the inference unit 15. It is determined whether or not, and the separator driving device 17 is controlled based on the output signal of the inference unit 15 to drive the separator 18, and when it is determined that it is a dangerous substance, it is determined that it is not the dangerous substance conveyor 20 and otherwise. In the case, it is distributed to the crusher entrance conveyor 19. The dangerous substances conveyed by the dangerous substance conveyor 20 are carried into the dangerous substance storage pit 21. In addition,
Reference numeral 16 in the drawing is a display unit.

Of the above sensors 2-8, sensor 2
Is a color sensor for measuring color, and the sensor 3 is a CCD camera for processing a visible image.

The sensor 4 is a metal sensor, which generates an electromotive force when a magnetic material such as a steel cylinder passes by using a DC coil having a loop surrounding the conveyor 1 as a detection portion.
The sensor 5 is a weight sensor and measures a roller supporting the conveyor 1 and a load applied to the roller. The sensor 6 is an X-ray line sensor in which a large number of X-ray sensors are linearly arranged. The sensor 6 is arranged horizontally below the conveyer conveyor 1 and at a right angle to the traveling direction of the conveyer conveyer 1. When an object passes, the output changes according to the absorption amount of X-rays emitted by the X-ray generator 7. The sensor 8 is a height sensor, which is a photoelectric tube sensor, and has a light source and a light receiving unit that emit light above the transport conveyor 1 at right angles to the traveling direction of the transport conveyor 1 and in the horizontal direction.

The output signals of these sensors 2 to 8 are output to the signal processor 9, the image processor 10, the metal detector 11, the weight measuring device 12, the X-ray image processor 13 and the photoelectric tube amplifier section 14.
Is configured to be supplied to the inference unit 15 after each predetermined signal processing.

The signal processor 9 sends out a signal representing the color information X ₁ of the conveyed product. The image processing unit 10 sends the visible image information X ₂ of the conveyed product. The metal detector 11 sends the metal conveyed object information X ₃ which depends on the size of the magnetic material and the passing speed.
The weight measuring device 12 sends the weight information X ₄ of the conveyed product. X
The line image processing unit 13 obtains a two-dimensional fluoroscopic image by sampling the processed signal of the X-ray line sensor 6 in a time series at regular time intervals, measures the major axis and the minor axis of the fluoroscopic image, and measures them respectively. Major axis information X ₄ and minor axis information X
Send ₅ . The phototube amplifier 14 is a phototube sensor 8
Is processed to measure the length of the conveyed product at a predetermined upper position of the conveyor 1 and the length information X ₇ is sent out.

As described above, the present embodiment is characterized by various sensors 2 to 8 arranged around the conveyer 1 such as the length and height of the target refuse (conveyed object), the property as a magnetic substance, the weight and the shape. Is determined, and the inference unit 15 performs a predetermined process based on these output signals to determine whether or not the dust is a dangerous substance.

As shown in FIG. 2, the inference unit 15 includes a fuzzy inference unit 25 having a large number of inference rules (R _{1 to} R _k ),
It has a neural network inference unit 27 composed of a three-layer neural network.

In the fuzzy inference unit 25, whether the information signals X _{1 to} X ₇ which are input signals are, for example, in the range B of the fuel gas cylinder to be detected or in the range A below it, for each size, Alternatively, the fuzzy set 22 classifies whether or not the range C is more than that.

The inference rule 23 of the fuzzy inference unit 25 uses this set to construct a rule of the form "Z is true (or false) if the condition of X = A is satisfied", and these rules are applied to various dusts. Create based on prior knowledge. For example, from the knowledge that the ratio of the metal transport information X ₃ and the weight information X ₄ is within a certain range if the measured dust is a fuel gas cylinder, the condition of X ₃ = A and X ₄ = C is satisfied. If Z is true, "create an inference rule. The synthesis inference unit 24 synthesizes these rules as a weighted average of the condition part conformance of the rules. Here, if the conclusion part proposition Z is true, a value of 0 is assigned if it is false, and if the result of inference synthesis α _F is close to 1, there is a high probability that the measured dust is a fuel gas cylinder, and close to 0. Therefore, it is inferred that the possibility of a fuel gas cylinder is small. β _F is an index showing the certainty of this inference result, and if α _F is close to 1,
The maximum value of the goodness of fit of the inference rule 23 giving α _F = 1 is β _F
And so on.

Each of the two in the neural network inference unit 27
Ulong's input weighting factor U₁₁~ U _7n, V₁₁~ V_n2Is
Actually known dust is thrown away, and this is the teacher signal α_NO, Β_NOWhen
It is determined in advance by the learning algorithm 28. Book
In the embodiment, the waste is a fuel gas cylinder as an output.
Probability α_NAnd the inference confidence β_NIs inferred,
Teacher signal α at the learning stage_NO, Β_NOAs the waste is fuel
Α for a bomb_NO= 1, and this is clearly judged
Β if possible_NO= 1, can be judged clearly enough to be mixed with other garbage
If not, β_NOA human gives a numerical value such as = 0.5.

The final decision is made by the decision unit 26, that is, the decision unit 26 weights the inference results of the fuzzy inference unit 25 and the neural network inference unit 27 with certainty factors β _F and β _N , and combines the results. If α is larger than a predetermined value, it is determined that the dust is a dangerous substance (fuel gas cylinder), and a signal indicating this is sent.

In this embodiment, by keeping the speed of the conveyer conveyor 1 that conveys oversized dust constant, a relationship between a specific dust and the outputs of the sensors 2 to 8 can be obtained. By increasing the speed of the conveyor 1, it is possible to separate and convey the overlapped dust, and it is possible to reduce the noise of the sensors 2 to 8.

For example, if the gas is a fuel gas cylinder such as LPG made of dangerous steel, if the waste is a fuel gas cylinder, the outputs of the sensors 2 to 8 fall within a limited range. For example, if the sensor 5 detects the weight, most fuel gas cylinders are in the range of 1.5 to 10 kg, and
If it is below, it can be determined that the fuel gas cylinder is not the target one, and if it is 10 kg or more, the possibility is low.

On the other hand, in the case of the sensor 8 which is a photoelectric tube sensor, if the detected height is set to the minimum diameter of the fuel gas cylinder to be detected, if the dust does not block this optical axis, it must not be the target fuel gas cylinder. I understand.

Further, since the fuel gas cylinder falls within a certain length range, if the dust is the fuel gas cylinder, the sensor 8
The duration of the output signal of is also within a certain range.

As described above, each of the sensors 2 to 8 provides information that can deny that the dust is a target dangerous substance or information that indicates the degree of "possibility" of the dangerous substance.
Here, the possibility is large if the dust is within the output range of the fuel gas cylinder, and is small if it exceeds the range. For example, even if the weight exceeds the detection target range, there is a possibility that the fuel gas cylinder and other waste may be conveyed in an overlapping manner. Therefore, the possibility that the waste is not the fuel gas cylinder is low, but it cannot be denied.

Thus, when the individual outputs of the respective sensors 2 to 8 are intended to detect the fuel gas cylinder in the waste,
Although it serves as an indicator of the characteristics of the dust, it contains many errors and is therefore problematic in practice as a fuel gas cylinder detection sensor by itself.

The inference unit 15 of the present embodiment integrates information including many errors in consideration of the characteristics of the sensors 2 to 8 as described above, so that the dust can be collected in a fuel gas cylinder with practical accuracy. It has a function of determining whether or not there is.

In the fuzzy inference unit 25, the sensors 2 to 8 are used.
Inference regulation having the following form with single output or combination of multiple inputs, eg, Z is true (or false) if the conditions of X ₇ = A and X ₈ = B are satisfied. Then, by integrating these inference results, the certainty of whether or not it is a fuel gas cylinder is calculated. Here, X ₇ and X ₈ are the outputs of the sensors 7 and 8,
A and B represent fuzzy sets such as "the output is in the range of the cylinder", "greater than or equal to the upper limit", "less than or equal to the lower limit", and Z represents the proposition "the garbage is a cylinder".

On the other hand, the neural network inference unit 27
Is the degree to which the conditional part of the fuzzy inference rule 23 holds,
That is, when the conformity is small over all the rules, it has a function of making a determination and calculates the probability that the dust is a fuel gas cylinder.

In the judging section 26, the fuzzy reasoning section 25 weights the conclusions of the neural network reasoning section 27 by the certainty factor of each reasoning and makes a final judgment.

In the test results using the actual collected trash according to the above-mentioned embodiment, the neural network inference unit 27 which has learned about 50 trash in advance for several hundred trash.
When operating only the fuel gas cylinder detection rate is 100
When set to%, the misjudgment is 14%, and the fuzzy inference unit 2
When inferred in parallel in 5 and judged, the false judgment is 3%,
Practicality has been confirmed. Of the fuzzy and neural network inference units 25 and 27, the fuzzy inference unit 25 has a knowledge-based inference function,
Shares the inference function for complex input patterns that cannot be obtained with fuzzy reasoning.

With these inference configurations, if new judgment knowledge is obtained, it can be easily added as an inference rule, and the rule can be easily modified. Further, since it is easy to incorporate a new sensor into the system, there is an advantage that it is highly expandable.

Further, according to the above-mentioned embodiment, it is possible to detect the fuel gas cylinder which could not be discriminated by the human eye because it was mixed in the bag or the corrugated cardboard, so that it is possible to provide a detection system of extremely high practicality. To do.

The order of detection by the sensors 2, 3, 4, 8, 5, 6 need not necessarily be the order of this embodiment, and the sensor configuration can be changed depending on the object to be detected.

Further, although the above-mentioned embodiment is a system for detecting a fuel gas cylinder in oversized waste in a waste treatment facility, it is not limited to this, of course. INDUSTRIAL APPLICABILITY The present invention has a versatile system as a detection system for foreign substances mixed in products in a factory production line or agriculture / fishery industry, for example, as a detection system for removing a foreign substance in a sorting field and determining a fish species in fishery processing. obtain.

[0039]

As described above in detail with reference to the embodiments, according to the present invention, there is a risk that the judgment of a specific foreign matter such as a dangerous object which is conventionally judged by a human cannot be recognized by the human eye. Since even objects can be detected automatically, by combining this detection signal with a removal device, for example, explosion accidents in oversized waste treatment equipment can be significantly reduced, and safety and economy that prevent human and equipment loss. This has the effect of obtaining surface advantages.

That is, in general, the present invention, which is a technique for detecting a specific object from various miscellaneous substances, has a wide range of applications such as extraction of substandard agricultural products at a sorting plant and discrimination of fish species in a fish processing plant. Various applications can be expected in the field.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a block diagram showing in detail the inference unit of FIG. 1 and showing this portion in detail.

[Explanation of symbols]

 1 Conveyor 2-8 Sensor 15 Reasoning part 25 Fuzzy reasoning part 26 Judgment part 27 Neural net reasoning part

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G06F 9/44 554 L 7737-5B C2 (72) Inventor Tomito Yonezawa Nishiki, Naka-ku, Yokohama-shi, Kanagawa 12 Mitsubishi Heavy Industries, Ltd. Yokohama Works (72) Inventor Norio Yoshimitsu 12 Nishikicho, Naka-ku, Yokohama-shi, Kanagawa Mitsubishi Heavy Industries Yokohama Works, Ltd.

Claims (2)

[Claims] 1. A conveyer conveyor for conveying a conveyed object such as bulky refuse, and various kinds of articles arranged along the flow direction of the conveyer and for detecting various properties such as weight, length and material of the conveyed object. A system for detecting a specific foreign matter in a conveyed product, comprising: a sensor; and an inference unit that integrates output information of various sensors and processes it based on a predetermined inference procedure to detect a specific foreign substance in the conveyed product. . 2. The inference unit has a fuzzy inference unit, a neural network inference unit, and a determination unit, and the fuzzy inference unit infers output information of each sensor based on a predetermined inference rule, and the inference result at this time. In addition to determining the certainty whether or not the conveyed object is a specific foreign matter, the neural network inference unit uses the multilayer neural network model with the output information of each sensor as the input layer to determine the inference rule of the fuzzy inference unit. Judge the degree to be established,
The determination unit is a final determination unit that weights the conclusions of the fuzzy inference unit and the neural network inference unit by the certainty of each inference, and determines the specific foreign matter in the conveyed product according to claim 1. Detection system.