94results about How to "Accurate Defect Detection" patented technology

A defect detecting device 1 includes a pixel value correcting section 16, a block-division processing section 17, and a defective/non-defective determining section 19. The pixel value correcting section 16 corrects a pixel value of an inspection-target image, on which detection of a defective area is to be carried out, in such a manner that the defective area of the inspection-target image is emphasized with respect to the other areas of the inspection-target image. The block-division processing section 17 divides, into plural blocks, the inspection-target image with pixel values having been corrected, and obtains a block addition value, which is a value obtained by adding pixel values of pixels present in a block, or a block mean value, which is a value obtained by dividing the block addition value by the number of pixels present in the block. The defective/non-defective determining section 19 determines whether the defective area is present or not, by carrying out statistical processing of a statistic processing section 18 to determine whether an outlier of the block addition value or an outlier of the block mean value is present or not. The foregoing realizes a defect detecting device and a defect detecting method, by which defective areas of image-sensor images are emphasized to enable accurate detection of the defective areas.

It is possible to inspect scratches and staining on a wafer surface on the basis of an LPD map obtained from a particle counter 11, by providing a means 21 for detecting aggregation of clustered point defects (LPD) from two-dimensional distribution information 30 for such fine LPD on the surface of a silicon wafer, and an improvement in the inspection efficiency and the precision of judgements of “defective” status can be achieved. Furthermore, the system is devised so that the trend of generation of scratches and staining in a specified process can easily be detected by accumulating wafer surface information such as scratch information, staining information and the like for the wafer surface detected by a wafer surface inspection device 11 (especially as image information or numerical information), and superposing sets of information thus accumulated. Plans for improving processes can be made by both the wafer supplier and wafer consumer by sharing such information with both parties.

The invention provides a detector for efficiently and automatically identifying faults of a flexible printed circuit board (FPCB). According to the invention, a high-precision marble platform is connected with a base through bolts, a table top of the marble platform is used for bearing an X-axis portal frame structure and two moving platforms in a Y-axis direction, the parts of two working platforms on the marble platform in the Y-axis direction are symmetrical and consistent and are free of mutual interference during working, and high-precision grating bars are respectively arranged on the moving platforms in the X-axis direction and the Y-axis direction so as to ensure that the closed-loop high-precision position motion control of the three platforms is realized, thus, the purpose of efficiently and automatically identifying the faults of the FPCB is achieved. The detector provided by the invention has the beneficial effect of being capable of detecting the faults of the FPCB stably and accurately at a high speed.

The invention relates to the technical field of mechanical part detection, in particular to a mechanical casting defect detection method and system based on artificial intelligence. The method comprises the following steps: acquiring at least two mechanical casting initial images; performing image segmentation on the image to obtain a suspected crack defect area image; obtaining principal component directions of the suspected crack defect area images, and determining different convolution kernel sizes and sliding step lengths of the suspected crack defect area images based on different principal component directions; inputting each suspected crack defect area image into a convolutional neural network for image feature extraction, and finally outputting the probability that each pixel point in the image belongs to the crack defect; and detecting the defects of the mechanical casting based on the trained convolutional neural network. The method and device greatly improve the speed of network training, improve the details of feature comparison in the network training process, improve the detection accuracy, and improvs the efficiency of defect recognition.

A system and method for detecting a defect in a solar cell and repairing and characterizing a solar cell includes applying a test signal to the solar cell, monitoring the response of solar cell, detecting a defect associated with its location during the monitoring step, removing or isolating the defect from a solar cell and characterizing solar cell performance. The defect may be a short between the emitter and the base of solar cell. The system and method also detect a precise location of the defect based on the use of light valve panel (LVP), which can control the input beam to or output beam from the solar cell in terms of size, position, gray level, and wavelength of the transmitted light. The LVP may be realized in any one of a variety of ways. For example, the active matrix liquid crystal display (AMLCD) such as Thin Film Transistor driven LCD (TFT-LCD) may be used as the LVP.

A fruit surface defect detection method based on image marking includes the following steps that firstly, a surface picture of a to-be-detected fruit is taken and saved, and an original image is obtained; secondly; the original picture is uploaded to a server to be analyzed and processed; processing of the server includes the steps that a, the obtained original image is converted into a space where the visual system of human beings is applied, and an H component and an I component are extracted; b, dynamic threshold segmentation is performed on the H component; c, gray histogram statistics is performed on the I component, segmentation is performed through a fixed threshold method, and a threshold is selected between two wave peaks; d, the H value segmentation result and the I value segmentation result are operated, and a binary image with defect areas is obtained; denoising is performed on the obtained binary image; f, the binary image is enhanced, hole noise may exist in the defect areas, and filling is performed on the noise; g, the obtained binary image is marked, and the number and the area of defects are calculated; a detection result is output; labor intensity of workers is reduced, and production efficiency is improved.

A system and method for detecting a defect in a transistor array includes applying a test signal to the array, monitoring pixel voltages along a gate line of the array, and detecting a defect associated with the gate line based on a variation in the pixel voltages along the gate line during the monitoring step. The system and method can also detect a precise location of the defect based on a rate of change in the variation of the pixel voltages along the gate line.

A method for performing circuit defect analysis and process problem identification includes applying a test signal to a circuit, obtaining a signal generated in response to the test signal, comparing the response signal to reference information, classifying a defect in the circuit based on a result of the comparing step, and identifying a problem in a manufacturing process which caused the defect based on the classification. The reference information may include one or more signal profiles corresponding to predefined types of defects that can occur during the manufacturing process. Defect classification is preferably performed by determining whether the response signal falls within one or more of the signal profiles. If the response signal falls within two or more signal profiles, then probabilities may be determined for each profile. The defect may then be classified as corresponding to the defect type whose signal profile has the highest probability. A processing system performs defect classification and process problem identification using a similar approach.

In a method of detecting a defect on an object, a preliminary reference image can be obtained from a plurality of comparison regions defined on the object. The preliminary reference image is divided into reference zones by a similar brightness. Each of the reference zones is provided with substantially the same gray level, respectively, to obtain a reference image. Whether a defect exists in an inspection region in the comparison regions is determined using the reference image. Thus, defects in the inspection regions having different brightnesses can be detected using the properly obtained reference image.

An apparatus for defect detection in a printing system, comprising a sensor device for sensing a surface of a sheet and for generating data representing a height map of the sheet, and a processor device for processing the data from the sensor device, wherein the processor device is configured to analyse the height map to detect a presence of a deformation in the sheet. The processor device is further configured to, when a deformation is detected, to determine at least one property of the deformation, to classify the deformation with respect to a plurality of deformation classes by means of the at least one property of the deformation, and to determine a suitability of the sheet for printing based on the deformation class wherein the deformation has been classified.

The invention relates to a real-time detecting method of metal arc plane flaws based on machine vision. The method comprises the following steps: step 1. obtaining N first sample images and M second sample images under the off-line condition, and carrying out data combination; step 2. establishing an image histogram after the combination so as to obtain a linear relationship equation between an image background and a corresponding gray value; step 3. calculating a refection component of a first sample image and a second sample image selected; step 4. establishing a corresponding relationship equation between a qualified working refection component standard difference and a corresponding gray value; step 5. on-line obtaining a detection image of a polished metal arc plane workpiece at real time, and calculating a first standard difference and a second standard difference; step 6. carrying out threshold partitioning so as to obtain a corresponding binary image; and step 7. comparing the connected region area of the binary image with a preset judgment threshold so as to judge the flaws of the polished metal arc plane. According to the method, the operation is convenient, the detection precision is high, the detection adaptability is good, and the stability and the reliability are realized.

The invention discloses a forging defect detection method based on deep learning, which comprises the following steps: firstly, establishing a YOLOv4 algorithm model for detecting forging defects, then training the YOLOv4 algorithm model, and then carrying out defect detection on a picture of a to-be-detected forging by adopting the trained YOLOv4 algorithm model. The method is characterized in that after residual errors in a feature extraction network of a YOLOv4 algorithm model are connected, a CBAM attention module is inserted, and features are screened. The method has the advantages of being good in robustness, capable of efficiently and accurately conducting defect detection and the like.

The invention relates to a combination electromagnetism explore injure equipment, it includes sensor, its feature is that it also includes bearing installed on the pipe, bar straightener, supporting axle, supporting roller armed with envelope curve rolling curve and sensor supporting, of which supporting axle installed on the bearing, supporting roller installed on the supporting axle, bearing fixed on the dead plate, arc slot set on the bearing and dead plate relative position and dent into bolt fastening, one end of sensor supporting through the through hole of bearing center and connect with the screw-thread hole of the dead plate center, the other end of sensor supporting connect with the sensor which installed between the both sides supporting roller, supporting sleeve covered sensor supporting and dead connect with bearing, transverse hole installed on the supporting sleeve bottom, one end of supporting axle cantilever reeve through the hole. The invention characteristics is that it could improve measurement precision and accuracy, which integrates a entire with explore injure function and straightening function, the invention could be used on the pipe or bar or sheet or section steel straightening machine production line.