JPH047678A - Picture recognizing device - Google Patents

Abstract

PURPOSE:To smoothly and stably execute the operation of picture recognition by altering a threshold in accordance with illuminance detected by light quantity sensor. CONSTITUTION:A binarization circuit 27 binarizing a picture signal from an image pickup means based on the threshold so as to obtain picture data and a control circuit 23 altering the threshold in accordance with illuminance are provided. Namely, the picture signal from a video camera 1 is binary-encoded in the binarization circuit 27 and it is accumulated in a frame memory 28. CPU 23 receives an output signal from the light quantity sensor 4 and calculates the threshold Vth for binary-encoding by data of RAM 22, and the threshold is transmitted to the binarization circuit 27 through an I/O port 25. Thus, the operation of picture recognition can smoothly be executed.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is applicable to, for example, the recognition of whether IC elements etc. are correctly mounted on a flat substrate in the manufacturing and assembly process, or whether holes are correctly opened, or the recognition of whether an object is The present invention relates to an image recognition device for performing shape discrimination, industrial measurement of pore diameter, etc.

[Prior Art] In order to perform the above recognition, a television camera images the object to be recognized, obtains a contrast based on density difference, and then binary encodes the image signal using an appropriate threshold (binarization level). hand,
As a result, characteristic regions such as peaks and holes with different colors are extracted with respect to the reference plane, and these characteristic regions are recognized.

By the way, due to changes in the intensity of external light due to changes in the intensity of sunlight, etc., or changes in the amount of light emitted by the light source over time when illuminating with a light source, the amount of light received by the television camera from the object to be recognized may change. When the amount of light received changes, the threshold value must be changed accordingly, but conventionally, the operator manually adjusts the threshold value by applying an appropriate small amount of change, and as a result, the feature region is correctly extracted. I used trial and error to see if I could do it or not, and adjusted the threshold to the correct value.

[Problems to be Solved] Therefore, there were problems in that adjusting the threshold value was time-consuming and image recognition work was time-consuming.

SUMMARY OF THE INVENTION Therefore, it is an object of the present invention to provide an image recognition apparatus in which threshold values can be automatically and easily adjusted, and image recognition operations can therefore be carried out smoothly.

[Means for Solving the Problem] In order to achieve the above object, the image recognition device of the present invention has the following features:
A light amount sensor that detects illuminance based on reflected light from the recognition target object, an imaging means that images the recognition target object and outputs an image signal, and binarizes the image signal from the imaging means based on a threshold value. The image forming apparatus includes a binarization circuit that obtains image data, and a control concave path that changes the threshold value depending on the illuminance detected by the light amount sensor.

Further, a light source for illuminating the recognition target object and a light amount adjusting means for changing the light amount of the light source when the illuminance detected by the light amount sensor exceeds a predetermined allowable upper limit or lower limit may be provided.

Furthermore, an aperture amount adjusting means may be provided that changes the aperture amount of the imaging means when the illuminance detected by the light amount sensor exceeds a predetermined allowable upper limit or lower limit. Both may be provided.

[Example] Hereinafter, an example of the present invention will be described in detail based on the drawings.

First, the overall configuration of the image recognition device will be explained. 1st
In the figure, the video camera 1 detects the recognition target 2 when the recognition target 2 is transported and set at a predetermined inspection position.
It is placed in a position that allows the inspection area above to be viewed. The environment in which the recognition target object 2 is captured by the video camera 1 is in an open environment without any shielding means or the like against external light.
This eliminates the need for members such as covers, allowing the device to be made smaller.

A light source 3 made of a circular fluorescent tube is installed outside the focusing lens 1a of the video camera 1 so that its axis coincides with the light receiving axis of the video camera 1, and uniformly illuminates the surface of the recognition target 2. is possible. The light source 3 is provided with a light amount adjusting means (not shown), and the light amount can be changed.

A light amount sensor 4 is installed near the focusing lens 1a. As shown in FIG. 2, the light amount sensor 4 has a cylindrical light-shielding hood 4b attached to a predetermined length so as to surround a light-receiving surface 4a, so as not to receive direct light from the light source 3. The light receiving axis of the light amount sensor 4 is adjusted so as to intersect with the light receiving axis of the video camera 1 at the reference plane position of the recognition target object 2. It is preferable that the angle θ between the light receiving axis of the light amount sensor 4 and the light receiving axis of the video camera 1 be as small as possible, and the field of view of the video camera 1 and the reflected light source surface inputted by the light amount sensor 4 should be made to coincide. desirable. In order to convert the illuminance within the same field of view as the video camera 1 into voltage, it is preferable to use a photoelectric conversion element such as a photodiode or a photoelectric conversion element such as a Cds (cadmium sulfide cell) as the light amount sensor 4.

Output signals from the video camera 1 and the light amount sensor 4 are sent to the main body 6 of the device. The device main body 6 has a built-in control circuit, which will be described later, and is connected to the video camera 1. It controls the light amount sensor 4, sets a threshold value, performs binary encoding of the image signal from the video camera 1, and determines whether the recognition target 2 is good or bad based on the binary encoding. The device body 6 includes a CRT for displaying binary images, etc.
6a is provided. Also, inside the device main body 6, the output voltage S of the light amount sensor 4 is set at a predetermined upper limit Smax.

An alarm sound generator is provided for emitting an alarm sound when the lower limit Sm1n is exceeded (not shown). Furthermore, the apparatus main body 6 is provided with an input device 6b consisting of a teaching keypad for setting monitoring conditions and the like, and an external communication line 6C used for notification of test results and the like.

Next, referring to FIG. 3, the control circuit built into the main body 6 of the apparatus will be explained.

A program for operating a CPU (central control unit) 23 is stored in the ROM 21 .

The RAM 22 stores table data of a gain α, in which the dark level voltage SO of the light amount sensor 4 is set as the first item in the table data, and thereafter every time the output voltage increases by dS. Sl, S
2. S3. ...Si, ..., and the optimal threshold value vthl heno conversion coefficient for each Sl (gay:/)
al (=Vthi/S) data.

Further, the RAM 22 stores an upper limit Smax and a lower limit Sm1n of the output voltage of the light amount sensor 4 as adjustment limit values. Semiconductor light-receiving elements (CCD area sensors, etc.) used in ITVs (industrial cameras), etc. have a narrow dynamic range (measurement tolerance range), so when the output voltage S of the light sensor 4 falls outside of the range between Smax and Sm1n, , adjust the light intensity of the light source 3 to make it measurable illuminance for the video camera 1,
This is to allow the video camera 1 to perform correct measurements.

In addition, the RAM 22 stores pixel signals obtained by capturing an image of a normal recognition target object with the video camera 1 and binary-encoding it.

Tolerance values and the like when determining the quality of the recognition object 2 are stored. Monitoring conditions such as allowable values inputted from the keypad 6b are stored in the RAM 22 via the I/O port 25.

The output voltage S from the light amount sensor 4 is changed to A at every predetermined time interval.
/D conversion 31. It is sent to the CPU 23 via the i10 port 25, and this value S is also stored in the RAM 22.

The image signal from the video camera 1 is binary coded in the binarization circuit 27 and stored in the frame memory 28.

The threshold value Vth for binary encoding is calculated by the CPU 23 based on the data in the RAM 22 upon receiving the output signal from the light amount sensor 4, and is sent to the binarization circuit 2 via the i10 port 25.
This will be communicated to 7.

Note that the binary encoded signal is mixed with the image signal from the video camera 1 in a video mixing 29 so that it can be displayed on the CRTBa.

In addition to calculating the threshold value, the CPU 23 performs calculations for determining the quality of the recognition object 2 based on instructions in the ROM 21 and data in the RAM 22 .

Further, the CPU 23 issues an alarm sound generation signal when the output voltage S of the light amount sensor 4 deviates from the predetermined upper limit Smax and lower limit Sm1n, and the signal is transmitted to the alarm sound generator 33 via the I/O boat 25.

The results of the quality determination of the recognition target object 2 are transmitted to the outside via the communication boat 30 and the communication line 6c.

Next, the operation of this device will be explained using FIG. 5.

First, a normal recognition target is set within the field of view of the video camera 1, the light source 3 is adjusted to change the illuminance from dark to bright, and the output voltage S of the light amount sensor 4 is set from the dark level SO as a starting point.
is changed every dS, and the signal is sent to the i/o boat 25
It is then stored in the RAM 22. And CRT6
A is used to project a binarized image at each Sl, and an operator gives the optimum threshold value v thi at each Si to create a conversion table in the RAM 22. As a result, the conversion coefficient αi −Vthl /
Sl is found as table data (102, 104)
. Next, video camera 1 has a dynamic range.
If the amount of light received by the video camera 1 is larger or smaller than a predetermined value, the image of the recognition object 2 will not be expressed as a difference in contrast, so the output voltage S will fall within the range in which the image of the recognition object 2 can be expressed as a difference in contrast. Upper limit Sm
ax and the lower limit Sm1n are determined (103). Then table data αi, Smax, and Sm1n are stored in the RAM22.
memorize it in

Next, when the recognition object 2 to be inspected is transported within the field of view of the video camera 1 (this time is assumed to be 1-tp),
The amount of light from the light source 3 is adjusted so as to obtain an appropriate illuminance that allows the video camera 1 to take an image. Then, as shown in FIG. 4(b), the light amount sensor 4 detects at this time (t = tp)
, receives light from the reflected light source surface and outputs a voltage Sp, and the signal is sent to the CPU 2 via the I/O boat 25.
Sent to 3. When this value is an intermediate value between the values of S in the table data, the CPU 23 selects a value smaller than Sp and
The value closest to is regarded as the Sp value obtained by the light amount sensor 4, αp at this Sp value is read out from the RAM, Vthp = ap − Sp is calculated, and this value is sent to the I/O port 25. The signal is then sent to the binarization circuit 27, where a threshold value V thp is set.

Further, this value Sp is stored in the RAM 22.

Then, as shown in FIG. 4(a), the optimal threshold value V
thp is converted into binary code by the binarization circuit 27, and the second
The value encoding signal is stored in the frame memory 28 and the CPU
23 compares each pixel with the data stored in the RAM 22 as a normal 21-inch coded image, calculates the number of pixels with different codes (difference pixel number), and if the value is smaller than a predetermined tolerance value, If it is good or large, it is judged as bad, and the result is sent to the communication line 6c via the communication port 30.

Next, as shown in FIG. 4(b), after a predetermined period of time has elapsed, t=
tq, the output voltage Sq of the light amount sensor 4 is
This will be communicated to 23. The CPU 23 calculates l5q-Sp I using Sp stored in the RAM 22, and when this value is smaller than dS (107), the 2ifi conversion circuit 2
Therefore, the binarization circuit 27 performs binarization processing at V thp at t-tp (110).

When l Sq -6p l >dS, Sm1n≦S≦
To determine whether Smax or not, 10g), Sm1n≦
When S≦5max, 1-(Sq-5p)/dS
seek. Note that if a number below the decimal point occurs in i, the number below the decimal point is rounded down, and the value of Sq is set as S in the calculated i, that is, Sl. Find α1 at that i from the RAM 22, and from that αl, V thq−α1
- Calculate Sq and send this value to the binarization circuit 27.

Further, S at this time is corrected and stored in the RAM 22 as sp (109). As a result, the binarization circuit 27
Binarization is performed with q, and at this time, as shown in FIG. 4(a), the level of the image signal of the video camera 1 is also changing, but the threshold value v In order to be coded, it is binarized using an optimal threshold value, and the 21 encoded signals are stored in the frame memory 28, and the quality of the recognition object 2 is judged in the same way as above (110).
.

Next, when S<Sm1n or Smax<S, a predetermined alarm sound or the like is emitted. Then, the operator calculates the required voltage of the light source 3 from the output voltage S and adjusts the light source 3 to obtain Sm1.
Make sure that n≦S5Smax (111. 11
2). Then, i is determined using this S, and V thl is calculated (109), and binarization processing is performed using this threshold value.

Thereafter, similarly to the above, the output voltage S is sent from the light amount sensor 4 at predetermined time intervals, the threshold value is controlled, and the binarization process is performed based on the threshold value, and the recognition target object 2 is A pass/fail judgment is made.

Through the above operations, an optimal threshold value is automatically set, and image recognition and quality determination of the recognition target object 2 are performed based on this threshold value.

In the above embodiment, when entering the inspection process, the operator first adjusts and applies the light intensity of the light source 3, after which the light intensity sensor 4 automatically measures the light intensity, and the CPU 23 calculates the optimal threshold value. However, the operator first gives an optimal threshold value, extracts the output voltage of the light amount sensor 4 at that time from the table data α, and stores the value as Sp in the RAM 22 using the keypad 6b. Thereafter, the device may automatically control the threshold value.

Furthermore, in the above embodiment, the appropriate threshold value Vth was determined from the output voltage S of the light amount sensor 4 using the table data α.
Appropriate amount of change △V in the threshold value with respect to the change in the amount of light received by the light amount sensor 4, that is, the change △S in the output voltage S of the light amount sensor 4
th is experimentally determined, a conversion coefficient β between them is calculated, and using this β, the CPU 23 calculates the threshold change amount ΔV.
th-β・ΔS is calculated, and ΔVth is the original threshold value Vt
h may be added to modify the threshold value.

This makes it possible to eliminate the influence of noise. Note that if the output voltage of the light sensor 4 and the appropriate threshold value change proportionally, β will have only one value, and this value will be stored in the RAM.
22.

Furthermore, in the above embodiment, the lighting 3 is adjusted when 3<Smin or Smax S, but the video camera 1
It is also possible to provide an aperture amount adjusting means to change the aperture amount so that the amount of light received by the video camera 1 falls within a predetermined range within the dynamic range. Note that in this case, since the relationship between the amount of light received by the light sensor 4 and the amount of light received by the video camera 1 changes, table data α for various aperture amounts is obtained, and when the aperture amount is changed, the table data α is The threshold value Vth is determined based on the table data α regarding the aperture amount. Note that both the light amount adjustment means and the aperture ffi adjustment means may be provided and both of these may be operated. The image recognition device may automatically perform the adjustments by the light amount adjusting means and the aperture amount adjusting means.

Furthermore, in the above embodiment, the light receiving axis of the light sensor 4 and the light receiving axis of the video camera 1 were different, but a beam splitter was attached to the lens holder in the video camera 1 to split the light amount for the sensor, and the light receiving axis of the optical finder was By attaching a light sensor in place of the eyepiece, the light receiving axis of the light sensor 4 and the light receiving axis of the video camera 1 can be made the same.
The amount of light received by the video camera 1 and the amount of light received by the light amount sensor 4 can be made completely proportional. In this case, when the aperture amount is changed, the amount of light received by the light amount sensor also changes at the same time, so it is not necessary to obtain the table data α for the various aperture amounts.

[Effect] According to the present invention, since the threshold value is changed according to the illuminance detected by the light sensor, even if the amount of light received by the imaging means fluctuates, the threshold value can be tracked and corrected in real time with high accuracy. Therefore, it is possible to perform image recognition work smoothly and stably.

Further, a light source for illuminating the recognition target object and a light source light amount adjustment means are provided, and when the output of the light amount sensor exceeds a predetermined allowable upper limit or lower limit, the light source light amount is changed, and/or
Alternatively, by adjusting the aperture amount of the imaging device using the aperture amount adjustment device, it is possible to change the amount of light received by the imaging device to an appropriate value even when the amount of light received by the imaging device changes significantly.

[Brief explanation of the drawing]

The drawings show one embodiment of the present invention, in which Fig. 1 is an external perspective view showing how an image recognition device is capturing an image of a recognition target, Fig. 2 is a perspective view of a light amount sensor, and Fig. 3 is a perspective view of a recognition target object. System block diagram, Fig. 4(a) is a graph showing the time change of camera scanning line output on one scanning line, Fig. 4(b) is a graph showing the time change of camera scanning line output on one scanning line,
A graph showing changes in sensor output in case a), and FIG. 5 is a system flow diagram. 1...φ video camera, 2...Recognition target, 3...Light source, 4...φ light amount sensor, 23.φCPU. 27...Binarization circuit. that's all

Claims (3)

[Claims] (1) A light amount sensor that detects illuminance based on the reflected light from the recognition target object, an imaging means that images the recognition target object and outputs an image signal, and a threshold value for the image signal from the imaging means. an image recognition device comprising: a binarization circuit that binarizes the image data based on the image data; and a control circuit that changes the threshold value in accordance with the illuminance detected by the light amount sensor. . (2) A light source for illuminating the recognition target, and a light amount adjusting means for changing the light amount of the light source when the illuminance detected by the light amount sensor exceeds a predetermined allowable upper limit or lower limit. The image recognition device according to claim 1, characterized in that: (3) An aperture amount adjusting means for changing the aperture amount of the imaging means when the illuminance detected by the light amount sensor exceeds a predetermined permissible upper limit or lower limit is provided. image recognition device.