Application Research and Analysis of Machine Vision Sherlock Software in Detecting Parts Positioning

Overview

In the machine vision inspection process, a fixture is required to position each inspected part to ensure that the light source illuminates the part, so that the camera can capture the illuminated part, so that the position sensor can be triggered at the same position to prevent the part from shaking or rotating . When each part to be inspected reaches the inspection area, the vision system will inspect it. However, in many cases, even if the fixture is special, there is no guarantee that the position will not change, and the part cannot always be accurately positioned at the desired position, that is, the part cannot be positioned with respect to the camera repeatedly, as shown in Figure 1.

Application Research and Analysis of Machine Vision Sherlock Software in Detecting Parts Positioning

In many cases, this misalignment can be compensated by using related measurement methods. For example, the length of the part can be obtained in Sherlock software by using the line area of interest (peek) to find the end point to be measured, and then using the distance formula to calculate the distance between the two end points. In this way, the measured value will remain constant, even if the part moves a little bit, as long as it does not move very far, one of its endpoints does not move out of the corresponding line area.

However, in some cases, alignment is required. If an ideal template is used, or a small part is to be measured, regardless of normal changes in the position of the part, excessively expanding the area of interest, it is unlikely to ensure robustness. In these cases, it is easy to solve the problem that parts cannot be aligned by using the landmark of the machine vision software sherlock.

Parts alignment process

When the detected part is not in the initial training position, Sherlock's part alignment function automatically finds the position to be detected depending on the positioning point. The readings of all points can be marked as anchor points in the reading edit dialog box. Return the position of the point marked by the anchor point through instructions, and then automatically locate the region of interest to be processed according to the position of the anchor point.

● Alignment when parts only move

When the detected part only moves in the X direction or the Y direction, only one positioning point is needed for positioning to achieve automatic alignment. The operation process is as follows.

(1) First select a point on the captured image of the qualified part as the positioning point. For example, use the SmartSearch analysis algorithm of the rectangular area of interest to train a certain characteristic part of the part, and use the best point as the best point. location point.

(2) Extend the range of the rectangular region of interest, so that the feature part corresponding to the trained feature on the detected part is always within the range of the region of interest.

(3) Train the areas that need to be detected and add corresponding programs.

(4) Run the inspection program. When the part position moves, the area of interest of the part to be inspected will automatically find the part to be inspected for inspection based on the positioning point.

In Figure 2, the cross is used as the training template, the center of the cross is used as the positioning point, and the position to be detected is the comb tooth. When the part moves, the region of interest B can automatically find the position to be detected.

● Alignment when parts move and rotate

When the inspected part moves in the X and Y directions relative to the training template, and rotates, an anchor point cannot find the position to be inspected. Automatic alignment can be achieved using two anchor points in Sherlock. Among them, one anchor point calculates the movement information, and the other calculates the rotation information. The operation process is as follows.

1) Select two feature points that can be reliably located on the captured images of qualified parts as positioning points for training.

2) After the training is completed, the regions of interest of the two feature points are respectively enlarged so that each feature point is within its own region of interest.

3) Train the areas that need to be detected and add corresponding programs.

4) Run the inspection program, when the position of the part moves and rotates, the area of interest of the part to be inspected will automatically find the part to be inspected for inspection based on the positioning point.

When the part position is rotated, three positioning points can also be used to achieve alignment. Assuming that the point is a fixed point relative to the part, the following two points are only used to calculate the rotation information. Record the original angle between the points, find the new angle at runtime, and calculate the rotation distance through the comparison between the new angle and the original angle, so as to realize the alignment of the parts.

● Use the results of Smart Search360 to achieve part alignment

The analysis algorithm Smart Search360 in Sherlock software is a pattern recognition analysis algorithm, which can identify rotating objects, and can also achieve component alignment. There is a point and an angle in the returned result of Smart Search360. Use the point conversion function Formula-Align to convert the point and the angle into two positioning points, so as to achieve part alignment.