KR20030080288A - Apparatus for leather quality inspection using camera - Google Patents

Abstract

PURPOSE: An apparatus for leather quality inspection using a camera is provided, thereby obtaining the objectified quality grade of leather and rapidly inspecting the leather quality. CONSTITUTION: An apparatus for leather quality inspection using a camera comprises a transporting portion(3) for transporting leather to be inspected; a camera for taking a photograph of the leather; a lighting portion for providing light to the leather; a high frequency inverter(7) for providing electric power to the lighting portion; a light blocking plate for inhibiting leakage of light from the lighting portion; a photograph treating module(6) for treating photograph from the camera; a controlling portion(1) for determining the leather quality based on the photograph treating results; and a displaying portion(4) for displaying the results of the controlling portion(1).

Description

Apparatus for Leather Quality Inspection using Camera}

The present invention relates to an intelligent leather quality sorting apparatus using vision, and more specifically, to obtain an objective quality grade by selecting the quality of leather using a vision system equipped with artificial intelligence, and to shorten the selection time to reduce production efficiency. It is to increase.

Leather is a natural resource with higher added value than any other product, which makes a lot of price difference according to the quality grade. The most important factor in determining the quality of leather products is the visible surface condition. Until now, most leathers have been screened at the leather factory. This method consumes a lot of labor and time to classify the grades, not only is inconsistent, but also cannot detect minute defects or precise dimensions. There was no prepared environment.

In order to solve the problems described above, the present invention obtains an objective quality grade by selecting a quality of leather using a vision system equipped with artificial intelligence, and aims to standardize the grade and the quality of the grade, and to shorten the selection time. An object of the present invention is to supply a device capable of increasing the efficiency.

1 is an overall configuration diagram of an intelligent leather quality sorting apparatus of the present invention.

2 is a block diagram of the vision apparatus of the present invention.

3 is a block diagram of a leather transport apparatus of the present invention.

4 is a flowchart showing a leather quality selection method according to the present invention.

5 is a flowchart illustrating a process of extracting a feature vector from an image acquired for leather quality selection according to the present invention.

6 is a flowchart illustrating a process of performing quality selection using the feature vector received from the vision unit in the control system of the present invention.

7 is a weight distribution for a 10 × 10 panel used for grading in the control system of the present invention.

8 is a flowchart for calculating feature information on a defect used for rating determination in the control system of the present invention.

The present invention for achieving the above object is a control unit (1) for controlling the quality screening device, a vision unit (2) for picking up the screening object, the transporting unit (3) for transporting the screening object and the rating for displaying It is composed of a display unit 4, based on a leather conveying step of conveying leather, an imaging step of imaging the conveyed leather, processing of the captured image, transmitting a result of the processing of the image, and the transmitted result. It consists of evaluating the quality of leather using artificial intelligence techniques.

Hereinafter, the present invention will be described in detail with reference to the drawings.

1 is an overall configuration diagram of an intelligent leather quality sorting apparatus of the present invention. As shown in the figure, the present invention provides a control system unit 1 for controlling a quality sorting apparatus, a vision unit 2 for picking up a screening object, a transporting unit 3 for transporting a screening object, and a grade for displaying the class. It consists of the display part 4.

The control system unit 1 is composed of a control system PC and a control system monitor. When the control system 1 sends a supply signal of the leather that is the screening target 11, the operator supplies the leather to the transport unit 3. do. The conveyer 3 conveys the leather using the conveyor belt, and outputs a trigger signal to the image processing module 6 whenever the leather moves a certain distance by using the proximity sensor 16 installed in the conveyer. The image processing module 6 drives the camera 10 of the vision unit 2 at every trigger signal, receives coordinates from the proximity sensor 16, corrects the position thereof, and captures the leather to be selected at predetermined intervals.

The captured image is transmitted to the image processing module 6 and the image processing module 6 transmits the image processing result to the control system 1. Using the transmitted image processing result, the control system 1 transmits the selection result to the display unit 4. The image processing module 6 may be dedicated image processing hardware or may be a program for software processing in an independent processing unit. The interval at which the leather is imaged can be changed depending on the object to be selected. In addition, the image processing module 6 and the control system 1 may be configured as an integrated system. The display unit 4 may be a separate device or may be configured integrally with the control system 1. The control system (1) can receive the rating determination reference information by the operator and receives the defect information from the leather supply signal output function and the image processing module (6), the classification information and area information (square) calculation and classification results classification system It may have a function to output to (not shown).

2 is an enlarged part of the vision unit of the present invention. As can be seen in FIGS. 1 and 2, the vision unit is largely a camera 10, an illumination unit 9, a light shielding plate 8, an image processing module 6, and a high frequency inverter. It consists of (7). Camera 10 is supported by a support is installed on the carrying unit 3, the number of cameras can be configured differently depending on the object to be selected and the arrangement can also be arranged in various forms depending on the line or situation. A lighting unit 9 is provided at a constant distance from the camera, and a light shielding plate 8 is provided around the camera to prevent leakage of light emitted from the lighting unit. The structure of the lighting unit can be changed and configured according to the situation. The image processing module 6 is configured to enable high speed image processing, and the high frequency inverter 7 serves to supply power to the lighting unit 9.

Figure 3 shows the configuration of the carrying section of the present invention. As shown in the figure, the transport unit is composed of a drive module 12, a controller module 13, a mechanism unit 14, a vacuum device 15, a sensor module 16, a conveyor belt 17. The drive module 12 includes a motor, a gear, a power supply, and the like for rotating the conveyor belt, and the controller module 13 includes various electronic devices and electric devices for adjusting the speed of the conveyor belt. In addition, the vacuum apparatus 15 provides a suction force for adsorbing the leather to be screened to be fixed to the conveyor belt, the conveyor belt 17 is a vacuum suction is distributed in the passage that can pass the air to a certain size on the surface This is a possible structure. The sensor module 16 automatically detects the movement of the leather to be selected and outputs a trigger signal to the vision unit 2 at every predetermined length of movement.

4 is a flowchart illustrating a quality selection method of leather according to the present invention. When the selected leather is placed on the conveyor belt, the conveyor belt is driven to move the leather to the part where the camera is installed. When the leather moves, the detection sensor attached to the conveyor detects this, and from this time, the encoder (not shown) installed at a predetermined place of the moving unit proceeds a certain distance necessary for sorting a pulse of a certain shape. It is transmitted to the vision unit as a trigger signal (18). Then, the vision unit recognizes the signal and photographs the leather to be selected using the attached camera (19). The captured image is acquired as a still image by the image acquisition module of the vision unit 20. The acquired image is transmitted to the DSP (Digital Signal Processor) board of the image processing module (21), and undergoes a preprocessing process to be converted into an image for selection. The preprocessed image is again stored in memory (22). The CPU, which is a central processing unit of the image processing module, extracts a bend feature for the preprocessed image from the DSP board of the image processing module (23). After extracting the bend feature, the image processing module converts it into a feature vector of a form suitable for learning. The feature vector of a form suitable for learning can be changed according to the quality sorting type of leather (24). The extracted feature vector is transmitted to the control system via the LAN (25).

5 is a flowchart illustrating a process of extracting a feature vector from an image acquired for leather quality selection according to the present invention. In other words, as shown in FIG. 4, the CPU, which is the central processing unit of the image processing module, extracts the feature of the bend from the preprocessed image on the DSP board of the image processing module.

The acquired image is first subjected to a line buffer acquisition process to obtain a variance value and an average value. Average video (Mean video) Is obtained using the following formula.

Where m is the mean value, M is the horizontal size of the image, N is the vertical size of the image, and f (r, c) is the brightness value at (r, c). The variance image is configured after acquiring the variance value through the following process.

Where E stands for expectation, W hal f stands for 1/2 of the window size used to find the variance, and f (r, c) is the brightness at (r, c).

Through this process, the distributed image and the mean image (Mean image) are obtained and transmitted to the control system unit for quality determination.

6 shows a process in which the control system performs quality determination using the feature vector image received from the vision unit. The control system consists of a neural network learning process for quality determination and a quality determination process using the learned neural network. In the present invention, prior to performing online leather quality determination, the neural network to be used for leather quality determination in the control system is trained using the learning data configured by the user. As the learning progresses, the neural network outputs a difference from the target value specified by the user as an error. If the error is smaller than the error allowed by the user, the neural network is regarded as learning completed and stores the weight created when the learning is completed.

On the other hand, when learning is completed, an online leather quality judgment is made. The following describes the online leather quality judgment process. The control system first loads the weights stored at the completion of learning when the on-line quality judgment process starts, not learning. After the weight is loaded, the feature vector transmitted from the vision unit is input as an input vector of the neural network. The input feature vector performs weight and vector product, and the result of the vector product is output through the transfer function of the hidden layer of the neural network, and this output value is again a vector product with the weights connected between the output layer and the hidden layer of the neural network. The result of is passed the final output through the transfer functions of the nodes of the output layer. The control system outputs the quality determination result corresponding to the node having the largest value among the output nodes of the output layer, and this output value is transmitted to the grade display unit attached to the conveyor system which is the leather conveying device and stored in the database of the control system.

7 is a weight distribution for a 10 × 10 panel used for grading in a control system. It is assumed that the leather image photographed in the apparatus is photographed over a 5x9 panel. First find the width and width of the whole leather, then calculate the center of gravity. Next, calculate the distance D of each panel from the center of gravity using the following equation. The calculation results are shown in Table 2.

----(One)

Table 1. Parameters required for calculating the center point, distance D of leather

X Y node 3 4 Wx, Wy 1 / (Right-Left) ⅹ1.5 1 / Bottom Px, Py 2 2 Factor 2 2 Left One Bottom 8 Top 0

Table 2. Distance from center point to each panel

From the distances calculated in Table 2, calculate the intermediate parameter, P, using the following equation: The calculated results are shown in Table 3.

----(2)

However, when calculating the parameter in the center of gravity, P, if the above equation is applied as it is, it has a sharp shape, so the value multiplied by 0.7 is used.

Table 3. Parameter P for each panel

Next, the strain density value is calculated from the parameter, P and the texture density values in each panel of Table 3 using the following equation.

---- (3)

Table 4 shows the newly calculated density values considering the location of the leather. That is, the total density value of the leather, the density value of each panel should be more sensitive to the total density value as the center of the leather. The new density value was used as the equation (3) above to change exponentially with the change of density value in the center of leather.

Table 4. Tissue Densities for Each Panel

Table 5. New Density Calculations

Calculate the sum of the new density values calculated in Table 5, and then divide the sum of the parameters, Psum, by the sum Psum and use it as the reference value for leather grade.

8 is a flowchart for calculating feature information on a defect used for rating determination in the control system of the present invention. The defect information of the leather used in the present invention is three kinds of the position of the scratch, the number of pixels, the type of defect (hole / scratch). In the present invention, the scratches are divided into two types, that is, holes or surface scratches, and the number of scratches is calculated. The number of holes is calculated by dividing the actual area (number of pixels) by the unit area. Since the scratches such as wrinkles or scratches exist in most of the lengthwise directions, the diagonal length is calculated from the position information of the scratches. The sum of the diagonal lengths of the detected scratches is then divided by the unit diagonal length to calculate the number of wrinkles or scratches. Scratches smaller than the unit diagonal length are excluded from the grading here.

The present invention configured as described above can improve the inaccuracy of the quality selection resulting from the fatigue of the worker, can provide an objective leather quality standard, it can significantly improve the productivity and work process due to automation.

Claims (2)

A carrying part 3 for carrying the sorting object, A camera 10 for photographing the transported sorting object, Illumination 9 for irradiating light necessary for imaging; A high frequency inverter 7 for supplying power to the lighting unit; Light blocking plate 8 for preventing the leakage of light irradiated from the lighting unit, An image processing module 6 for processing an image picked up by a camera, The control system unit (1) for selecting the quality of the leather by receiving the result of the image processing It consists of a display unit (4) for displaying the screening results of the control system and the control system Intelligent leather quality sorting device characterized by using the defect information of leather for grading The method of claim 1, wherein the defect information for grading of the control system is characterized in that the position of the scratch, the number of pixels, the type of the defect is used as holes, scratches, wrinkles, and the relationship between the types and positions of the defects An intelligent leather quality sorting device, comprising calculating using a weight table and using the calculated result as a parameter for grading the leather.