JP2697463B2 - Steel sheet crop cutting position determination device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for determining a crop cutting position of a steel sheet.

[0002]

2. Description of the Related Art When a steel sheet is rolled, irregular portions (hereinafter referred to as "crops") as shown in FIG. 3 are generated at both ends in the longitudinal direction of the steel sheet. 3 (a) to 3 (c) are plan views of a steel plate showing an example of a crop shape. FIG. 3 (a) shows a tongue-shaped crop, FIG. 3 (b) shows a fishtail-shaped crop, and FIG. FIG. 3 (c) shows a crop having a shape combining the tongue shape and the fish tail shape, respectively. In general, many crop shapes have a tongue shape and a fish tail shape. Such crops are cut because they cannot be used as products.

As a conventional apparatus for cutting a crop as described above, an apparatus for automatically determining a cutting position of a crop and automatically cutting the determined position (Japanese Patent Laid-Open No. 60-238223).
Publication). Hereinafter, the apparatus will be described. FIG.
FIG. 1 is a schematic block diagram showing a configuration of the conventional crop automatic cutting device described above.

[0004] In the figure, reference numeral 1 denotes a steel plate. The steel plate 1 is conveyed at a constant speed in the longitudinal direction to a cutting machine 2 by conveying means (not shown) such as conveying rollers and the like. The resulting crop is cut off. A light receiver is provided at a predetermined position on the upstream side of the cutting machine 2 in the conveying path of the steel sheet 1, and by detecting that light to the light receiver is blocked by the tip of the steel sheet 1,
A passage detector 31 that detects the passage of the tip of the steel plate 1, a camera 32 that captures an image of the tip of the steel plate 1,
And a strobe light 33 that emits light toward the tip of the transport path are arranged in this order from the upstream side to the downstream side of the transport path.

[0005] The detection result of the passage detector 31 is given to an imaging control device 4, which controls the operation of the camera 32 and the strobe 33. Further, image data obtained by imaging with the camera 32 is provided to the imaging control device 4. The imaging control device 4 operates the camera 32 and the strobe 33 a predetermined time after the passage detector 31 detects the passage of the leading end of the steel plate 1, whereby the leading end of the steel plate 1 The camera 32 captures a still image of the distal end portion when passing through the imaging region of FIG.

[0006] Still image data captured by the camera 32 is
It is provided to the image storage device 5 via the imaging control device 4 and stored in the image storage device 5. The still image data stored in the image storage device 5 is provided to the image processing device 6. The image processing device 6 performs a process of binarizing the given still image data into an image of the steel plate 1 and an image of the background based on the luminance, and the binarized image data is supplied to the signal processing device 7. It has become.

The signal processing device 7 includes a shape classifying unit 70a for classifying a crop represented by the binary image data into one of a tongue shape and a fish tail shape, and a format classifying unit 70a When classified into a tongue shape, the tongue shape cutting position determining unit 70b that determines the cutting position, and when the crop is classified into a fishtail shape by the shape classifying unit 70a, the cutting position is determined. Determined fishtail shape cutting position determining unit 70c
It is composed of

In the signal processing device 7, first, the shape classification unit 71
In the above, the crop represented in the binarized image data provided from the image processing device 6 is determined whether the shape is closer to the tongue shape or the fish tail shape, and the closer shape is determined. Classify as the shape of The binarized image data is provided to the tongue shape cutting position determining unit 70b and the fishtail shape cutting position determining unit 70c, whichever corresponds to the shape classified by the shape classifying unit 71.

The tongue-shaped cutting position determining unit 70b and the fishtail-shaped cutting position determining unit 70c determine the cutting position of the crop as follows. In the tongue-shaped cutting position determining unit 70b, as shown in FIG. 3 (a), in crop represented by the binary image data, k (0 width W ₀ of the normal portion of the length of the cropping steel plate <k <1 ) A position a which is equal to twice is obtained by searching from the end of the crop and determined as a cutting position, and a distance L from the end of the crop to the position a is obtained.

On the other hand, as shown in FIG. 3 (b), the fishtail shape cutting position determining unit 70c determines, in the crop represented by the binarized image data, from the deepest part of the concave portion of the fishtail shape crop to the longitudinal direction of the steel plate. In addition, a position b separated by a predetermined distance L ₀ is determined as a cutting position,
The distance L from the end of the crop to the position b is obtained.

The data of the distance L thus obtained is
The information is given to the cutting machine control device 8 as information representing the cutting position command value. The cutting machine control device 8 detects the timing at which the position of the distance L from the end of the steel plate 1 passes through the cutting machine 2 by a predetermined means, operates the cutting machine 2 at that timing, and operates the cutting machine 2 at a distance from the end of the steel plate 1. It cuts at the position of L.

[0012]

However, in the conventional crop cutting apparatus as described above, the crop shapes are classified on the assumption that the crop shape corresponds to either the tongue shape or the fish tail shape, and the crop shape is classified. Since the cutting position was determined, for example, even when a crop having a special shape as shown in FIG.
The crop is classified into either a tongue shape or a fishtail shape.

[0013] For example, the shape of the crop shown in FIG. 3 (c), are classified as fishtail shape, by such a method described above, a distance L ₀ from the recess of the crop in the longitudinal direction of the steel sheet away c is disconnected At this position c, since the width W of the steel sheet on the cutting line is too small compared to the width W ₀ of the normal part of the steel sheet, when the steel sheet is cut at this position c, the steel sheet is finish-rolled. There was an inconvenience that the boarding ability to the machine became poor.

As described above, in the conventional method of classifying a crop shape into either a tongue shape or a fish tail shape and determining a cutting position, an appropriate cutting is performed for a crop having a shape other than those shapes. There was a problem that the position could not be determined.

The present invention has been made in view of the above circumstances, and provides a crop cutting position determining apparatus for a steel sheet that can determine an appropriate cutting position for crops of various types. The purpose is to:

[0016]

According to the present invention, there is provided an apparatus for deciding a position for cutting a steel sheet, which picks up an image of an end of the steel sheet and quantizes the imaged result to obtain image information representing the shape of the crop at the end. A steel sheet crop cutting position determining device for determining a cutting position of the crop in relation to the obtained image information, wherein, based on the image information, the image information for each of crops of a plurality of predetermined shapes. Means for obtaining the shape similarity of the crop expressed by the first neural network, and the cutting position of the crop of the steel sheet based on the shape similarity obtained by the means, the image information and the dimensions of the steel sheet. Means for deciding by a second neural network.

[0017]

[Function] The cropped image of the end of the steel plate is an analog signal.
No. Di that this by quantizing represents the crop shape
Digital image information is obtained. Such digital
By using image information, data to the neural network
Data input becomes possible. Based on this image information input ,
The shape similarity of the crop represented in the image information for each of the crops of a plurality of predetermined shapes is determined by the first neural network, whereby the shape of the crop represented in the image information is obtained. Is
It is represented by a shape similarity to a crop of a predetermined shape (a typical shape), and is not specified as a single type of shape. Then, based on the determined shape similarity, the image information, and the dimensions of the steel sheet, the cutting position of the crop of the steel sheet is determined by the second neural network, and the shape similarity to the crop having a predetermined shape is determined. Since the degree is the information on the shape of the crop represented in the image information, even if the shape of the crop is a shape other than the predetermined shape, the shape of the crop represented by the shape similarity The determined value of the cutting position according to the above is obtained.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be specifically described below with reference to the drawings showing the embodiments. FIG. 1 is a schematic block diagram showing a configuration of a crop cutting device to which a crop cutting position determining device for a steel sheet according to the present invention is applied.

In the figure, reference numeral 1 denotes a steel sheet. The steel sheet 1 is transported at a constant speed in a longitudinal direction to a cutting machine 2 by a transporting means (not shown) such as a transport roller. The resulting crop is cut off. A predetermined position on the upstream side of the cutting machine 2 in the conveying path of the steel plate 1 has a pair of light-emitting device and light-receiving device, and detects that light from the light-emitting device to the light receiving device is blocked by the tip of the steel plate 1. In this order, a passage detector 31 that detects the passage of the tip of the steel sheet 1, a camera 32 that captures an image of the tip of the steel sheet 1, and a strobe 33 that emits light toward the tip of the steel sheet 1 when the camera 32 captures an image. The transport path is arranged close to the downstream side from the upstream side.

The detection result of the passage detector 31 is given to an imaging control device 4, which controls the operations of the camera 32 and the strobe 33. Further, image data obtained by imaging with the camera 32 is provided to the imaging control device 4. The imaging control device 4 operates the camera 32 and the strobe 33 a predetermined time after the passage detector 31 detects the passage of the leading end of the steel plate 1, whereby the leading end of the steel plate 1 When passing through the imaging region, the camera 32 captures a still image of the distal end.

Still image data captured by the camera 32 is
It is provided to the image storage device 5 via the imaging control device 4 and stored in the image storage device 5. The still image data stored in the image storage device 5 is provided to the image processing device 6. The image processing device 6 performs a process of binarizing the given still image data into an image of the steel plate 1 and an image of the background based on the luminance, and the binarized image data is supplied to the signal processing device 7. It has become.

The signal processing device 7 includes a shape determining neural network 71 for determining the shape of the crop represented in the binary image data, and a shape determining neural network 71.
And a cutting position determining neural net unit 72 that determines the cutting position of the crop represented in the binary image data based on the binary image data and the width and thickness data of the steel plate 1. It is configured.

Next, the configuration of the neural network in the shape determining neural network 71 and the cutting position determining neural network 72 will be described. FIG. 2 is a block diagram showing a configuration of a neural network in the shape determining neural network 71 and the cutting position determining neural network 72. The neural network in the shape determining neural network 71 and the cutting position determining neural network 72. Is basically the same as that of the first embodiment, and will be generally described with reference to FIG.

The neural networks in the shape determining neural network 71 and the cutting position determining neural network 72 include an input layer I for receiving an input signal, an output layer O for providing an output signal to the outside, an input layer I, and an output layer. The input layer I has a hierarchical structure consisting of three layers with an intermediate layer M interposed between O, and the input layer I has K neurons I ₁ , I _{2 ,.}
And the hidden layer M is composed of L neurons M ₁ , M ₂ ,.
Consists M _L, the output layer O of the N neurons O _1,
O _2, is made up of ... O _N.

Each neuron I _i of the input layer I (where i =
1,2, ..., K) are all neurons M _j of the hidden layer M (however,
j = 1, 2,..., L) and different coupling coefficients w _ij .

The neurons M ₁ , M _{2 ,.}
Determines its output value based on the total value of inputs from input layer I. The determination of the output value is performed by calculation using a sigmoid function. Also, each neuron M _j of the intermediate layer M
(Where j = 1, 2,..., L) is a coupling coefficient v _ij different from all neurons O _j (where j = 1, 2,..., N) of the output layer O.
Are joined by The neurons O ₁ , O ₂ ,.
O _N determines its output values y ₁ , y ₂ ,... Y _N based on the total value of inputs from the middle layer M. The determination of the output value is also performed by calculation using a sigmoid function.

The coupling coefficients v _ij and w _ij are changed to optimal values by learning such as learning based on the back propagation rule. Next, a learning method based on the back propagation rule will be described. The learning of the back propagation rule is divided into forward calculation and backward calculation.

In the forward operation, first, input data f ₁ ,
f _2, ... neurons I ₁ of each of the input layer I the f _K,
I _2, ... is input from I _K, the output layer neurons O ₁ of O each, O _2, ... output data y ₁ of O _N, y _2, determining the ... y _N. In the backward calculation, the teacher data t
_1, t _2, ... t _N and the output layer the output data y ₁ of O,
y _2, ... as sequentially reduces an error between y _N, the coupling coefficient v _ij between the coupling coefficient w _ij and the output layer O and the intermediate layer M between the intermediate layer M and the input layer I Make changes.

In the neural network for shape determination 71 and the neural network for cutting position determination 72 using such a neural network, learning for shape determination and cutting position determination is performed in advance, and then actual shape determination and cutting are performed. The position is determined.

The neural network of the shape determining neural network 71 is provided by the image processing device 6.
The valued image data is input data f ₁ , f ₂ ,... F _K of the neural network, and the similarity to each crop is output for a plurality of crops (N in this case) of a predetermined typical shape. Data y ₁ , y ₂ ,... Y
_N is a neural network. The neural network receives binary image data of a crop of a known shape as input, outputs the similarity, and sequentially calculates an error between the output data and teacher data (teacher data given by a human). Coupling coefficient w to decrease
Learning to change _ij and the coupling coefficient _vij is performed in advance.

The neural network of the cutting position determining neural network 72 includes the similarity data provided from the shape determining neural network 71 and the width and thickness of the steel sheet 1 provided from a predetermined means (not shown). Data and the binarized image data provided from the image processing device 6 are used as input data f ₁ , f ₂ ,
... and f _K, outputs a cutting position with respect to the crop data y _1, y _2, a neural network that ... y _N. This neural network receives as input binary image data of a crop having a known shape, data on the similarity of the crop, and data on the width and thickness of the steel plate 1 and outputs data on the cutting position with respect to the crop. Let
Learning is performed in advance to change the coupling coefficient w _ij and the coupling coefficient v _ij so as to sequentially reduce the error between the output data and the teacher data (teacher data given by a person).

In the signal processing device 7 configured as described above, first, the shape determination neural network 71
For the crop represented in the binarized image data, the similarity to each of a plurality of crops having a predetermined typical shape such as a tongue shape and a fishtail shape is obtained.
As a result, the shape of the crop represented in the binarized image data is represented by the similarity to the crop having a typical shape, and is not specified as a single type of shape.

The neural network 72 for determining the cutting position first determines the cutting position for each of the typical shapes based on the binarized image data and the data on the width and thickness of the steel sheet. As an example of a method of obtaining the cutting position, in the tongue shape, as shown in FIG. 3A, in the crop represented by the binarized image data, the width of the crop is k (k) of the width W ₀ of the normal portion of the steel plate. A position a that is equal to 0 <k <1) times is obtained by searching from the end of the crop and is set as a cutting position.
As shown in (b), the crop represented by the binary image data, in the longitudinal direction of the steel sheet from the deepest portion of the concave portion of the crop, the position b at a distance L ₀ to a predetermined and cutting position.

Next, a plurality of data of the determined cutting position is
The data is weighted with the similarity data and combined to determine a final cutting position. That is, the cutting position determining neural network unit 72 obtains a cutting position in each shape when it is regarded as a typical shape, and interpolates the data of this cutting position between the respective shapes. And the final cutting position is determined. The determined cutting position is represented as a distance from the end of the crop to that position, and is given to the cutting machine control device 8 as information indicating a cutting position command value.

[0035] The cutting machine control device 8
The position of the distance detects the timing of passing through the cutting machine 2, the cutting machine 2 is operated at that timing, and the position of the distance is cut from the end of the steel plate 1.

[0036]

As described above, according to the present invention,
Based on the image information, the first neural network determines the shape similarity of the crop represented by the image information with respect to each of the crops of a plurality of predetermined shapes, so that the shape of the crop is simply It is not specified for one type of shape. Then, based on the obtained shape similarity, the image information, and the dimensions of the steel sheet, the cutting position of the crop of the steel sheet is determined by the second neural network, and the shape similarity to the crop having a predetermined shape is determined. Since the information of the shape of the crop represented in the image information is included, even if the shape of the crop is a shape other than the predetermined shape, the shape of the crop is determined according to the shape of the crop represented by the shape similarity. The present invention has an excellent effect such that a cutting position can be obtained.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram showing a configuration of a crop cutting device to which a crop cutting position determining device for a steel sheet according to the present invention is applied.

FIG. 2 is a block diagram showing a configuration of a neural network in a neural network for shape determination and a neural network for cutting position determination.

FIG. 3 is a plan view of a steel plate showing an example of a crop shape.

FIG. 4 is a schematic block diagram showing a configuration of a conventional crop automatic cutting device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Steel plate 2 Cutting machine 6 Image processing device 7 Signal processing device 32 Camera 71 Neural net part for shape judgment 72 Neural net part for cutting position determination

Claims (1)

(57) [Claims] An image of an end of a steel plate is imaged, the imaged result is quantized to obtain image information representing a shape of a crop at the end, and a cutting position of the crop is determined in relation to the obtained image information. In the apparatus for determining a position for cutting a steel plate crop, based on the image information, a shape similarity of a crop represented by the image information with respect to each of crops having a plurality of predetermined shapes is determined by a first neural network. Means for determining a cut position of a crop of the steel sheet by a second neural network based on the shape similarity obtained by the means, the image information and the dimensions of the steel sheet. For determining the position of crop cutting of a steel plate.