CN108364306B - Visual real-time detection method for high-speed periodic motion - Google Patents
Visual real-time detection method for high-speed periodic motion Download PDFInfo
- Publication number
- CN108364306B CN108364306B CN201810114345.8A CN201810114345A CN108364306B CN 108364306 B CN108364306 B CN 108364306B CN 201810114345 A CN201810114345 A CN 201810114345A CN 108364306 B CN108364306 B CN 108364306B
- Authority
- CN
- China
- Prior art keywords
- image
- periodic motion
- motion
- speed
- periodic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000000737 periodic effect Effects 0.000 title claims abstract description 123
- 230000000007 visual effect Effects 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000011897 real-time detection Methods 0.000 title claims abstract description 35
- 230000002159 abnormal effect Effects 0.000 claims abstract description 37
- 238000001514 detection method Methods 0.000 claims abstract description 34
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims description 34
- 238000001914 filtration Methods 0.000 claims description 6
- 239000003550 marker Substances 0.000 abstract description 5
- 239000000284 extract Substances 0.000 abstract description 3
- 230000000875 corresponding effect Effects 0.000 description 25
- 238000004364 calculation method Methods 0.000 description 18
- 230000008569 process Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000011179 visual inspection Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 4
- 230000005856 abnormality Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000011031 large-scale manufacturing process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000009659 non-destructive testing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a visual real-time detection method of high-speed periodic motion, belonging to the technical field of visual detection, and the visual real-time detection method comprises the following steps: extracting the characteristic of the periodic motion to obtain the circumferential angle of the periodic motion; establishing a high frame rate image sample library based on the circular angle of the periodic motion; and performing visual real-time detection and abnormal extraction based on the high frame rate image sample library. The method is a rapid detection algorithm of a monocular camera and a periodic feature marker, can perform nondestructive and visual real-time detection on periodic mechanical motion, automatically extracts abnormal conditions occurring in the periodic mechanical motion, and realizes real-time visual detection of the periodic motion at a high speed.
Description
Technical Field
The invention relates to the technical field of visual detection, in particular to a visual real-time detection method for high-speed periodic motion.
Background
Visual inspection technology is increasingly used in the production and life of human society, and conventional visual technology is to collect a target image by using a camera, store video image data in a computer, and research corresponding image processing and video processing algorithms to complete automatic detection and identification functions based on the visual image. The visual inspection technology can be classified into general visual inspection and high-speed visual inspection according to the real-time processing speed. Common vision inspection systems typically have real-time processing speeds of 25-30 frames/second, and high-speed vision real-time inspection processing speeds, typically above 128 frames/second, for capturing and inspecting objects operating at high speeds.
The high-speed vision system is used for acquiring instantaneous images moving at high speed, rapidly processing target images within a few milliseconds through a high-performance image processing algorithm, outputting key image characteristics and finishing real-time processing of targets. The common vision is not suitable for detecting the high-speed movement because the acquisition speed is low, instantaneous high-definition images cannot be acquired for the high-speed movement phenomenon, and the texture features are fuzzy and difficult to distinguish.
In the industrial production field, high-speed motion presents a periodic and repetitive motion state, and is controlled by a human to complete a set action for large-scale production and processing. Among them, a mechanical system driven by a motor is a typical periodic motion operation system. The single-step execution time of the high-speed mechanical motion is typically completed in a few milliseconds, requiring a 1000fps high-speed camera for detection.
Researchers have proposed the real-time visual inspection technology research of the motion characteristics of high-speed mechanical systems. The method comprises the steps that a high-speed image sensor is utilized to clearly capture an instantaneous image of high-speed periodic motion, for each pixel point, the gray value change meets the periodic characteristics, so that an original high-frame-rate image is prestored, background pixel points are removed through primary filtering, and periodic characteristic pixel points which have the largest influence on the periodic motion are selected through offline image processing; and calculating a period phase value of the original video by using the pixel points and the original image, and establishing a period phase lookup table, a period vector sample library and a period image sample library for real-time detection. In the online image processing, each frame of high-speed image is converted into a one-dimensional vector according to selected pixel points, similarity matching is carried out from a periodic vector sample library, phase values corresponding to the current frame are searched and extracted according to a periodic phase lookup table, the real-time phase of high-speed periodic motion is calculated through phase calibration of the corresponding image, and the frequency value of the periodic motion is obtained according to long-time real-time calculation. The technical scheme is shown in figure 1. The method comprises the steps of optimizing pixel points of high-speed motion by using a high-speed camera, pre-storing high-speed motion images with normal periods, designing a complex algorithm, extracting a series of key pixel points from global image pixel points through large-calculation operation of off-line image data, and expressing characteristic values of the high-speed periodic motion through vector values of the pixel points. Meanwhile, the pixel points are correlated with the original image to establish a sample library, so that high-speed real-time visual detection is performed. However, the method has the following technical defects: 1. The method needs to store a large amount of normal motion high frame rate image data, utilizes a complex image processing algorithm to calculate and extract key pixel points, has a complex and time-consuming off-line processing process, and usually needs a dozen of minutes of off-line calculation process, so the biggest defects are large data volume, large calculation amount and complex algorithm. 2. The method mainly aims at high-speed periodic motion of an area without obvious circular motion, so that a proper mark point cannot be found. Since no obvious periodic motion characteristic exists, the method has unstable results in the calculation of the periodic characteristic value and generates more noise interference.
Therefore, a rapid detection algorithm of a monocular camera and a periodic feature marker is urgently needed for periodic mechanical motion driven by a motor, so that nondestructive and visual real-time detection can be performed on the periodic mechanical motion, abnormal conditions occurring in the periodic mechanical motion can be automatically extracted, and the real-time visual detection of 1000fps can be realized.
Disclosure of Invention
The invention aims to provide a visual real-time detection method based on high-speed periodic motion of a high-speed camera aiming at the technical problems in the prior art, and solves the following technical problems:
1. for the high-speed periodic motion target, a high-performance image processing algorithm can be designed through a high-speed camera, the high-speed motion is detected in real time, and the abnormal state which cannot be observed by human eyes is automatically extracted.
2. In an acquired image of a high-speed camera, a periodic motion area image and a target detection area image are set and associated, system equipment is simplified, and a high-speed periodic motion characteristic is defined by using the relationship between the two area images.
3. By utilizing the periodic motion mark, the motion characteristic is quickly extracted, a high-frame-rate periodic motion sample library is established, the problem of millisecond-level real-time processing of high-speed vision above 500fps is solved, an abnormal target is automatically detected, and abnormal image data is extracted and recorded.
In order to achieve the purpose, the technical scheme of the invention is as follows: a visual real-time detection method for high-speed periodic motion is characterized by comprising the following steps:
extracting the characteristic of the periodic motion to obtain the circumferential angle of the periodic motion;
establishing a high frame rate image sample library based on the circular angle of the periodic motion;
and performing visual real-time detection and abnormal extraction based on the high frame rate image sample library.
Preferably, the extracting the periodic motion feature and obtaining the circumferential angle of the periodic motion includes:
pre-storing high frame rate image data as primary sample image data, wherein the pre-stored high frame rate image is an image of a target to be detected moving normally;
marking a periodic motion area in the high frame rate image data, wherein the marked periodic motion area belongs to a part of pre-stored high frequency image data;
carrying out local area image processing on the mark periodic motion area, and fusing mark point positions corresponding to the mark periodic motion area of each frame of image to form a circle;
and calculating the circumferential angle of the periodic motion.
Preferably, the establishing a high frame rate image sample library based on the circular angle of the periodic motion includes:
reordering the circumferential angles in the periodic motion;
associating the global images to form a standard high-speed periodic motion image sample library;
establishing a sample image with a mask structure according to the data difference between adjacent sample libraries in the standard high-speed periodic motion image sample library;
and establishing a standard image sample library with mask filtering based on the sample image with the mask structure.
Preferably, the performing, based on the high frame rate image sample library, visual real-time detection and anomaly extraction includes:
calculating the circular angle of the periodic motion of the real-time image;
extracting and processing the corresponding sample library image;
calculating normal and abnormal state values of the real-time high-speed image;
automatic detection and image extraction.
Preferably, the single high-speed camera acquires a complete image area including a circular motion area and a high-speed motion operation area to be detected.
Further preferably, the reordering of the circular angles in the periodic motion specifically comprises: and the high-frequency images are gradually reordered from small to large according to the corresponding angles to form a dense and subdivided periodic motion sample image library.
Further preferably, the global image is associated, specifically: and combining the reordered images, detecting the area image, performing the characteristic image of the circumference angle, and associating the images with the corresponding angles.
The invention has the following advantages:
(1) the invention can realize automatic detection of the instant abnormal state of the high-speed periodic moving target and the high-speed dynamic background. For the phenomenon of high-speed motion, clear images are difficult to acquire by a common camera, while clear images can be acquired by a high-speed vision system, but the calculation amount of a common algorithm is large, and the real-time processing of 500fps-1000fps is difficult to perform. The invention can perform high-speed real-time image detection aiming at the high-speed periodic motion characteristic.
(2) Aiming at high-speed periodic motion, the invention skillfully utilizes the periodic motion characteristic thereof to quickly extract the motion characteristic value, utilizes the high-speed camera to carry out image detection, and establishes a sample database, thereby realizing high-efficiency and quick visual detection.
(3) The invention only uses one high-speed camera to collect images of the detection area, divides images of the periodic motion area, realizes periodic motion coding by using typical motion characteristics, reduces the establishment time of an image sample library in off-line processing and saves the storage space. The method can realize the algorithm by utilizing software programming, and avoids the realization of hardware programming to a complex algorithm.
(4) The invention carries out real-time visual detection aiming at high-speed periodic motion, particularly complex mechanical motion driven by a motor or a high-speed system with local circular motion, and can also utilize a characteristic marking method of a periodic motion area to carry out rapid characteristic value calculation and periodic motion coding.
(5) The invention provides a non-contact visual nondestructive testing method for high-speed periodic motion, has better flexibility, and can analyze the real condition when abnormal action occurs through the automatic video extraction and detail playback technology.
Drawings
FIG. 1 is a schematic diagram of high-speed motion real-time detection in the prior art;
FIG. 2 is a schematic flow chart of a high-speed visual real-time detection method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a high speed periodic mechanical motion provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of image processing of a periodic motion mark region according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating a calculation of a circular angle of a periodic motion according to an embodiment of the present invention;
FIG. 6 is a mark point center coordinate of real-time image processing;
FIG. 7 is a circular motion angle value for real-time image processing;
FIG. 8 is a diagram illustrating status values of real-time image processing;
FIG. 9 is an image of an abnormal motion element in high-speed periodic motion detected by high-speed vision;
fig. 10 is a binarized extracted image of abnormal motion in high-speed periodic motion detected by high-speed vision.
Detailed Description
The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The embodiment of the invention provides a high-speed camera-based visual real-time detection method for high-speed periodic motion, which comprises three parts, namely periodic motion feature extraction, high-frame-rate image sample library establishment and high-speed visual real-time detection, as shown in figure 2. The respective implementation processes of the three parts are described in detail as follows:
1. periodic motion feature extraction
(1) Pre-stored high frame rate image data
The target to be measured is in high-speed periodic motion, belongs to reciprocating periodic motion driven by a motor, and can perform periodic operation for multiple times in one second. And searching a circular motion area driven by a motor, and fixing the label on a circular motion part. A high frame rate image of an object to be measured in high motion is acquired with a high speed camera at a frame rate of 1000fps, which may be denoted as I (x, y, N) (N is 1, 2, …, N), where N is the number of frames of the acquired high frame rate image, and a 1 second high frame rate video includes 1000 images. The high frame rate image is an 8-bit single-channel gray scale image, and the pixel value ranges from 0 to 255.
And the pre-stored high frame rate image is an image of the normal movement of the target to be detected and is recorded as primary sample image data. Wherein the acquired image area needs to comprise a complete mechanical operation. As shown in fig. 3, the high-speed periodic motion example is that the motor drives the slider-crank mechanism, the right square frame area is the motor driving device, the left square frame area is the mechanical operating device of the slider and the track, and the right circular line is the motor circular motion, which represents the motion track of the mark point center. The whole mechanical system can perform high-speed reciprocating motion of 0-50 Hz. According to the method provided by the invention, the complete image area acquired by a single high-speed camera needs to comprise a circular motion area and a high-speed motion operation area to be detected. The acquired complete image I (x, y, n) and the circular motion area image Im(x, y, n), image of region to be detected Id(x, y, n), the following can be classified: the complete image is divided into: circular motion area image Im(x, y, n), image of region to be detected Id(x, y, n) and other background areas are separated from each other without intersection, and then:
the complete image is divided into: circular motion area image Im(x, y, n) and the image I of the region to be detectedd(x, y, n), separating from each other without intersection, then:
Im(x，y，n)+Id(x，y，n)＝I(x，y，n)
the complete image is the image I of the area to be detectedd(x, y, n), wherein,circular motion area image Im(x, y, n) is a local area, namely, the local area participates in the calculation of the circumferential characteristics, and the abnormal state detection is carried out, then:
Id(x，y，n)＝I(x，y，n)，
image I of area to be detectedd(x, y, n) is a portion of the complete image I (x, y, n), the circular motion region image Im(x, y, n) is a part of the image of the region to be detected, namely, the part participates in the calculation of the circumferential characteristics and carries out the detection of the abnormal state, then:
for the above situations, it is necessary to ensure that the circular motion region image is used for calculating the characteristic of high-speed motion, and for the image region to be detected, the segmentation can be performed according to actual requirements. The following method will be described by taking the case (i) as an example, and as shown in fig. 3, the detection method in other cases is the same as the case (i), except that the areas of the regions to be detected are different.
(2) Marking regions of periodic motion
The high-speed camera is fixed in position, the acquired image area is unchanged, and the motion area of the motor-driven execution device is selected and marked according to the characteristics of the periodic motion to be detected, as shown in the right square area of fig. 3, the motion area comprises circular motion mark points and does not comprise other areas to be detected. The marked periodic motion area ensures that the background and the marked point have a larger contrast effect. The mark periodic motion region belongs to a part of prestored high frame rate image data and is defined as Im(x，y，n)(n＝1，2，…，N)。
(3) Local area image processing
For periodic motion region I of markm(x, y, N) (N is 1, 2, …, N), performing threshold segmentation, and reassigning the circular mark and the background according to 0 and 255 to form a binary image:
wherein E ism(x, y, N) (N ═ 1, 2, …, N) is a binary image containing only black marker points, TmThe white marked points can be distinguished from the black background for threshold. In the local area image, the circle is marked black and the background is white, as shown in fig. 4. In the pre-stored high frame rate image, the circular motion mark area of each frame of image has the corresponding mark point position, and all the positions are fused to form a circle, as shown in fig. 4.
(4) Periodic motion circular angle calculation
Binarizing image E of circular motion marking areamIn (x, y, N) (N is 1, 2, …, N), black circular pixel regions are fused, as shown in fig. 5(a), all circular markers form a circular black ring with a pixel value of 0, and the center 0(x, y) of the black pixels is calculated, which is the center coordinate of the circular motion. Using a binarized image Em(x, y, N) (N is 1, 2, …, N), calculating the central point X, Y coordinate of the black circle in the images, and calculating the included angle of the mark point center relative to the horizontal axis of the track center 0(x, y) by combining a basic trigonometric function formula, wherein the included angle is marked as thetan(N is 1, 2, …, N), as shown in fig. 5(b), and the prestored image I (x, y, N) and the corresponding marker point angle θ are obtainednRelationship I (x, y, θ)n) (N-1, 2, …, N), i.e. the periodic motion is encoded with an angular value.
2. Establishment of high frame rate image sample library
(1) Circumferential angle reordering in periodic motion
The pre-stored images I (x, y, N) (N ═ 1, 2, …, N) capture high frame rate images, a 1 second video comprising 1000 images, including mechanical manipulations greater than one cycle, i.e. the corresponding circular motion markers have multiple cycles. These high frame rate images I (x, y, theta)n) The new high is obtained by re-ordering according to the corresponding angle theta (N) (N is 1, 2, …, N) from small to large in an increasing wayFrame rate sample image set R (x, y, phi)n) (N-1, 2, …, N) a more densely subdivided library of periodic motion sample images will be formed. The corresponding to-be-detected region image and the circular motion region image are respectively expressed as: i isd(x，y，θn) And Im(x，y，θn)，(n＝1，2，…，N)。
(2) Global image correlation
Set of images R (x, y, phi) to be reorderedn) (N is 1, 2, …, N), the detection region image R is performedd(x，y，φn) Circular angle characteristic image Rm(x，y，φn) Phi of corresponding anglen(N-1, 2, …, N). Satisfying any nth frame image:
and forming a standard high-speed periodic motion image sample library as a correct reference for subsequent high-speed visual real-time detection after the associated image set is finished. In the real-time detection, the instantaneous high frame rate image at any position can find the standard image through the corresponding marking angle, and the normality and the abnormality of the real-time state are judged through the comparison of image data.
(3) Neighborhood image fusion
The high frame rate image acquisition process is a discretization sampling process of high-speed periodic motion to be detected, so that a certain time interval, namely a periodic action interval exists between adjacent sample images in a standard periodic image sample library. In high-speed real-time detection, a high-frame-rate image of instantaneous motion may be between angles of adjacent sample images corresponding to an angle value of a mark point at the moment, so that in image difference comparison, two matched images are not identical, a false edge or a ghost image occurs, and certain noise interference is generated after difference calculation, thereby affecting detection accuracy.
The invention establishes a sample image with a mask structure through the data difference of the sample images corresponding to two adjacent angles, and makes the difference between the adjacent images into a mask image for filtering the noise of a differential image in real-time detection, wherein the noise generated by the differential calculation of the normal state image does not need to participate in the final calculation:
after the difference of adjacent images is defined, the pixel value is in a value of 64-256, and the pixel value is obvious false edge. For gray values less than 64, the noise interference considered to be negligible can be filtered out by threshold segmentation in real-time detection. The sample image of the area to be detected and the mask image pass through phi of the corresponding anglen(N-1, 2, …, N), a final periodic image sample library may be established:
Rf(x，y，φn)＝{Rd(x，y，φn)，Ry(x，y，φn)}
(4) establishing a final image sample library
Based on the off-line image processing method, a final standard image sample library with mask filtering can be established and is marked as Rf(x，y，φn) (N-1, 2, …, N). Indicating that the mark has a circumferential angle of phi when moving at high speednWhen the position is determined, the sample data corresponding to the image of the detection area at the moment is Rf(x，y，φn)。
3. High speed visual real time detection and anomaly extraction
(1) Circumferential angle calculation of real-time image periodic characteristics
After a final image sample library is established by offline image processing, high-speed real-time video detection is carried out on high-speed periodic motion. And defining the image acquired in real time by high-speed vision as G (x, y, t), wherein t represents the time corresponding to high-speed real-time detection. The method in the periodic motion characteristic extraction is also utilized to mark out a circumference mark region image Gd(x, y, t) and local target detection region image Gm(x，y，t)。
For markingRegion of periodic motion Gm(x, y, n) carrying out threshold segmentation, and reassigning the circular mark and the background according to 0 and 255 to form a binary image:
wherein, GEm(x, y, T) is a binary image containing only black marker points, TmThe white mark point can be distinguished from the black background for the threshold value, the detected system environment is unchanged, the value is the same as the parameters of local area image processing, the center 0(x, y) of the circular mark motion is utilized, the angle between the mark center point and the center 0(x, y) in the real-time binary image is calculated and is marked as thetatThe method is the same as that shown in FIG. 5 (b). In the high-speed real-time visual detection, the coordinate values of the center point x and y of the circular motion mark can be calculated, and as shown in fig. 6, the central point x and y conform to the positive line function waveform. Meanwhile, the corresponding circumferential angle is calculated by utilizing a trigonometric function formula and is shown in fig. 7, the change range of the circumferential angle is 0-360 degrees, a periodic change rule is presented, in real-time detection, a circumferential angle output value with the interval of 0.001 second can be obtained, and the high-speed visual real-time calculation of 1000fps is realized.
(2) Extraction and processing of corresponding sample library images
For the angle theta corresponding to the position of the mark point in real-time high-speed motiontIn the sample image library Rf(x，y，φn) In the middle, the two most similar angles phi are quickly searchedGPhi and phiG+1And extracting the corresponding image Rf(x，y，φG) The method comprises the following steps of: rf(x，y，φG)＝{Rd(x，y，φG)，Ry(x，y，φG)}. Calculating the difference between the real-time image and the sample image by using the image difference:
D(x，y，t)＝|G(x，y，t)-Rd(x，y，φG)|
setting an empirical value T of an image binarization threshold value according to experimental data of the same environmentGMeter for measuringCalculating a binary image of the difference image:
the pixel value of the background image is 0, and the pixel value of the foreground image is 255. When the high-speed motion is in a normal state, abnormal objects do not exist in the background image after the image difference, and the whole image pixel value is 0; when an abnormal state occurs in high-speed motion, a part of pixel regions with the pixel value of 255 appear in the binary image after image differentiation, which indicates that an abnormality occurs. Then, noise filtering is performed through the mask image:
FG(x，y，t)＝DG(x，y，t)·Ry(x，y，φG)
wherein, when the corresponding 0 value in the mask image is the mask structure, the binary image D is multiplied by the valueGAfter (x, y, t), the corresponding mask region is zeroed, and for the 1 value in the mask image, no processing is performed and the original value is kept.
(3) Real-time high-speed image normal and abnormal state value calculation
Calculating a binarized image FGIn (x, y, t), the number of pixels with a gray value of 255, that is, the number of the pixels is counted, and the pixel area occupied by the abnormal object in the real-time detection is represented as follows:
for the high-speed real-time detection, the real-time status value Sum (t) obtained by image processing at time t ideally indicates that no abnormal value is detected when the device is in normal motion, where Sum (t) is 0. However, in an actual environment, due to the existence of a system error or external interference of a sensor to acquire an image in real time, the image has a certain noise, so that the calculated real-time state value Sum (t) is not all 0 and fluctuates in a low value region. When there is an abnormal condition, large value fluctuation occurs in Sum (t), and it can be determined that a true abnormality has occurred, as shown in fig. 8 (top).
For this situation in actual detection, the present invention utilizes a threshold value TSMasking the noise value in real-time detection:
this results in a waveform diagram as shown in fig. 8 (bottom) where there is a continuous square wave signal for a certain time when an anomaly occurs, and at other normal states, there is a substantially 0 value, and there is only occasional step signals at individual times. The invention calculates the sum H (t) of s (t) values of continuous K frames of the high-speed video stream:
while using a threshold value THTo determine if an abnormal condition occurs, when H (T) > THAnd when the abnormal state is detected, the abnormal values of the continuous multi-frame images in the detected real-time video images are shown, the abnormal action is really caused in the detected high-speed motion, at the moment, a signal is sent to an acquisition system, the moment and the subsequent high-speed images are automatically stored in the memory, the image data are stored in the hard disk by utilizing multithreading, and the judgment and the recording of the abnormal state are completed. Because the current image needs to be judged in real time, continuous K frames of images are needed, and the images also comprise a part of abnormal states, a memory area is opened up in a program, the previous K frames of images are pre-stored, and finally the images and the subsequently stored images are stored together to form a complete video image. When the abnormal state is over, H (T) is less than or equal to THWhen it is, the high-speed image storage is ended. Fig. 9 shows the process of detecting abnormal motion in high-speed periodic motion by the method of the present invention, wherein the abnormal object flies through the high-speed periodic motion mechanical system. FIG. 10 is a binary image F of an abnormal motion detected by a high-speed vision system and obtained by high-speed periodic motionG(x, y, t) processing the binary images processed in real timeThe row exception pixel counts and Sum (t) is obtained and is noted as the (upper) exception waveform of fig. 8.
4. Principle of operation
A high speed periodic motion system to be tested is selected, which typically has a circular motion component or is motor driven. A high-speed visual system driven by a calculator is built by utilizing a high-speed camera, the high-speed camera is fixed, and the shot scene is ensured to comprise: a high-speed periodic motion detection area and a high-speed circular motion area. On the surface of the circular motion assembly or the motor driving device, parts similar to circular motion are found, and mark points are fixed and distinguished from background colors. Turn on the light source and keep all devices unchanged.
The high-speed vision detection system is operated to enable the high-speed movement equipment to be detected to be in a normal state and operate at a high speed, the high-frame-rate image is collected by the high-speed camera and recorded as image data stored in advance, the high-speed movement can be temporarily stopped, and sample data is established. In the off-line image processing, a region image to be detected moving at a high speed is set, and a region image with circular motion characteristics is set. And the algorithm designed by the invention is utilized to perform off-line image processing to generate a final image sample library.
And in the high-speed visual real-time detection process, the position of the high-speed camera is kept unchanged, the motion state of the target to be detected is unchanged, and the external light source equipment is unchanged. The high-speed motion equipment is enabled to run for a long time, and the abnormal state value of the image is calculated by utilizing the image sample library and the real-time detection algorithm. And when the high-speed vision system judges the abnormal state, automatically storing the high-speed video image data with abnormal motion for state playback. Meanwhile, the judgment and automatic storage of the next abnormal action are prepared.
The embodiment of the invention provides a high-speed visual detection method of a circumferential characteristic mark in high-speed periodic motion, which can be used for detecting a target with circumferential motion characteristics in the high-speed periodic motion, can also be used for approximating the circumferential motion, and can perform coded periodic motion by utilizing angles.
The embodiment of the invention utilizes a high-speed camera to carry out high-speed visual detection, and can also utilize the same sample library establishment method to detect periodic motion for a visual detection system with low-speed motion or common speed.
The invention designs a camera for high-speed image acquisition, circular marking is carried out on one local circular motion area, marked motion is calculated, and the characteristic value calculation or coding of the whole motion is realized. The invention can also use two or more marks to calculate or encode the characteristic value of periodic motion by calculating the relative relationship between the mark points, and the invention also protects similar methods.
Although the invention has been described in detail above with reference to a general description and specific examples, it will be apparent to one skilled in the art that modifications or improvements may be made thereto based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.
Claims (3)
1. A visual real-time detection method for high-speed periodic motion is characterized by comprising the following steps:
extracting the characteristic of the periodic motion to obtain the circumferential angle of the periodic motion; the extracting of the periodic motion characteristic to obtain the circular angle of the periodic motion comprises the following steps: pre-storing high frame rate image data as primary sample image data, wherein the pre-stored high frame rate image is an image of a target to be detected moving normally; marking a periodic motion area in the high frame rate image data, wherein the marked periodic motion area belongs to a part of pre-stored high frequency image data; carrying out local area image processing on the mark periodic motion area, and fusing mark point positions corresponding to the mark periodic motion area of each frame of image to form a circle; calculating a circumferential angle of the periodic motion;
establishing a high frame rate image sample library based on the circular angle of the periodic motion; establishing a high frame rate image sample library based on the circular angle of the periodic motion, including: reordering the circumferential angles in the periodic motion; associating the global images to form a standard high-speed periodic motion image sample library; establishing a sample image with a mask structure according to the data difference between adjacent sample libraries in the standard high-speed periodic motion image sample library; establishing a standard image sample library with mask filtering based on the sample image with the mask structure;
performing visual real-time detection and abnormal extraction based on the high frame rate image sample library; the visual real-time detection and abnormal extraction based on the high frame rate image sample library comprises the following steps: calculating the circular angle of the periodic motion of the real-time image; extracting and processing the corresponding sample library image; calculating normal and abnormal state values of the real-time high-speed image; automatic detection and image extraction;
a complete image area acquired by a single high-speed camera comprises a circular motion area and a high-speed motion operation area to be detected;
the periodic motion is circular motion.
2. The visual real-time detection method of claim 1, wherein the reordering of the circumferential angles in the periodic motion comprises: and the high-frequency images are gradually reordered from small to large according to the corresponding angles to form a dense and subdivided periodic motion sample image library.
3. The visual real-time detection method according to claim 1, wherein the associated global image is specifically: and combining the reordered images, carrying out detection area image and circumference angle characteristic image, and associating with the corresponding angle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810114345.8A CN108364306B (en) | 2018-02-05 | 2018-02-05 | Visual real-time detection method for high-speed periodic motion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810114345.8A CN108364306B (en) | 2018-02-05 | 2018-02-05 | Visual real-time detection method for high-speed periodic motion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108364306A CN108364306A (en) | 2018-08-03 |
| CN108364306B true CN108364306B (en) | 2020-11-24 |
Family
ID=63004465
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810114345.8A Active CN108364306B (en) | 2018-02-05 | 2018-02-05 | Visual real-time detection method for high-speed periodic motion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108364306B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110006630B (en) * | 2019-04-10 | 2020-01-14 | 广东工业大学 | On-site detection method based on multi-period differential sampling and high-speed point motion equipment |
| CN110414322B (en) * | 2019-06-14 | 2024-05-28 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for extracting picture |
| CN110987383B (en) * | 2019-12-19 | 2021-07-13 | 浙江理工大学 | Image-based measuring method for high-frequency two-position reciprocating motion state |
| CN111369469B (en) * | 2020-03-10 | 2024-01-12 | 北京爱笔科技有限公司 | Image processing method and device and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100245527A1 (en) * | 2009-03-25 | 2010-09-30 | Fuji Xerox Co., Ltd. | Exposure device, image forming apparatus, and computer-readable medium |
| CN104915957A (en) * | 2015-05-29 | 2015-09-16 | 何再兴 | Matching rectification method for improving three dimensional visual sense identification precision of industrial robot |
-
2018
- 2018-02-05 CN CN201810114345.8A patent/CN108364306B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100245527A1 (en) * | 2009-03-25 | 2010-09-30 | Fuji Xerox Co., Ltd. | Exposure device, image forming apparatus, and computer-readable medium |
| CN104915957A (en) * | 2015-05-29 | 2015-09-16 | 何再兴 | Matching rectification method for improving three dimensional visual sense identification precision of industrial robot |
Non-Patent Citations (2)
| Title |
|---|
| "一种高速视觉实时定位与跟踪***的研制";郑西点等;《上海电气技术》;20150331;第8卷;论文第2.2节 * |
| "高速立体视觉与机械运动异常动作三维检测";王耀东等;《仪器仪表学报》;20131031;第34卷(第10期);论文第3.1-3.2节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108364306A (en) | 2018-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108364306B (en) | Visual real-time detection method for high-speed periodic motion | |
| CN107194559B (en) | Workflow identification method based on three-dimensional convolutional neural network | |
| CN106407974B (en) | The method of target positioning and Attitude estimation for interesting target | |
| Schmugge et al. | Detection of cracks in nuclear power plant using spatial-temporal grouping of local patches | |
| CN109900711A (en) | Workpiece, defect detection method based on machine vision | |
| CN107657244B (en) | Human body falling behavior detection system based on multiple cameras and detection method thereof | |
| CN103454285A (en) | Transmission chain quality detection system based on machine vision | |
| JP2008046903A (en) | Apparatus and method for detecting number of objects | |
| CN107240112B (en) | Individual X corner extraction method in complex scene | |
| CN111199556A (en) | Indoor pedestrian detection and tracking method based on camera | |
| CN102831378A (en) | Method and system for detecting and tracking human object | |
| CN105844282B (en) | A method of atomizer O-Ring defect is detected with line scan camera | |
| CN113822810A (en) | Method for positioning workpiece in three-dimensional space based on machine vision | |
| CN114463296A (en) | Light-weight part defect detection method based on single sample learning | |
| Singh et al. | A new algorithm designing for detection of moving objects in video | |
| CN110853077B (en) | Self-adaptive infrared dynamic frame feature extraction method based on morphological change estimation | |
| CN114613006A (en) | Remote gesture recognition method and device | |
| Zhang et al. | An optical flow based moving objects detection algorithm for the UAV | |
| KR101214858B1 (en) | Moving object detecting apparatus and method using clustering | |
| CN111435429A (en) | Gesture recognition method and system based on binocular stereo data dynamic cognition | |
| JP4918615B2 (en) | Object number detection device and object number detection method | |
| CN111582076A (en) | Picture freezing detection method based on pixel motion intelligent perception | |
| CN116310263A (en) | Pointer type aviation horizon instrument indication automatic reading implementation method | |
| CN103473753A (en) | Target detection method based on multi-scale wavelet threshold denoising | |
| CN110866917A (en) | Tablet type and arrangement mode identification method based on machine vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240116 Address after: Room 1518B, Unit 2, 12th Floor, Huizhi Building, No. 9 Xueqing Road, Haidian District, Beijing, 100085 Patentee after: Beijing contention Technology Co.,Ltd. Address before: Beijing Jianzhu University, No.1 zhanzhanguan Road, Xicheng District, Beijing 100044 Patentee before: Beijing University of Civil Engineering and Architecture |
|
| TR01 | Transfer of patent right |