CN114359597A - Oil tank inner cover identification and pose parameter sensing method based on vision - Google Patents

Oil tank inner cover identification and pose parameter sensing method based on vision Download PDF

Info

Publication number
CN114359597A
CN114359597A CN202111425958.1A CN202111425958A CN114359597A CN 114359597 A CN114359597 A CN 114359597A CN 202111425958 A CN202111425958 A CN 202111425958A CN 114359597 A CN114359597 A CN 114359597A
Authority
CN
China
Prior art keywords
inner cover
image
oil tank
area
edge
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.)
Granted
Application number
CN202111425958.1A
Other languages
Chinese (zh)
Other versions
CN114359597B (en
Inventor
徐海波
汪泽玮
沈翁炀
温利涛
李沛轩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN202111425958.1A priority Critical patent/CN114359597B/en
Publication of CN114359597A publication Critical patent/CN114359597A/en
Application granted granted Critical
Publication of CN114359597B publication Critical patent/CN114359597B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • 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
    • G06T2207/10004—Still image; Photographic image
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30248—Vehicle exterior or interior
    • G06T2207/30252—Vehicle exterior; Vicinity of vehicle

Abstract

The invention discloses a visual-based oil tank inner cover identification and pose parameter sensing method, which comprises the following processes: preprocessing an initial image of the inner cover of the oil tank to be identified, and performing edge detection after preprocessing to obtain image edge characteristics; carrying out adaptive template matching on the edge characteristic image in a scaling mode by utilizing the created circular edge template to obtain an inner cover suspected area; carrying out self-adaptive histogram equalization on the initial image; performing target segmentation on the image after histogram equalization and the obtained suspected area by using a graph cutting algorithm, and separating out an inner cover area; reflecting the inner cover pitching degree by utilizing the approximation degree between the fitting outline area of the inner cover area and the ideal area based on the morphological characteristics; and performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray level characteristics of the inner cover, and fitting out a straight line in the handle direction to obtain the rotation angle of the inner cover handle. The invention can obtain the pose parameters of the inner cover with higher precision on the basis of identifying the object of the inner cover.

Description

Oil tank inner cover identification and pose parameter sensing method based on vision
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a method for identifying an inner cover of an oil tank and sensing pose parameters based on vision.
Background
With the rapid development of social economy, continuous improvement of traffic facilities and continuous improvement of the living standard of residents in China, the use demand of gas stations is increased along with the increase of passing vehicles. Under the market trend, the traditional manual refueling mode cannot completely adapt to the development of the society and the living demands of people due to the factors of improved labor cost, lower working efficiency, higher labor intensity and severe partial environment, and the automatic refueling of the unattended gas station realized by the robot becomes a new realization mode.
The visual recognition belongs to an important link of environment perception in the autonomous refueling process of the refueling robot, and the robot needs to complete the recognition of an outer cover of an oil tank and an inner cover of the oil tank aiming at a refueling action flow and an automobile structure. The identification of the inner cover of the oil tank belongs to a key identification link before the oiling action, and because the shape of the inner cover has the characteristics of local difference and easy influence of the characteristic by the illumination environment, the functional products and related researches of the existing oiling robot are mainly focused on the identification link of the outer cover, and the identification link of the inner cover is usually ignored. The difference between the pitch angle and the rotation angle of the inner cover handle exists in the posture of the inner cover of the automobile oil tank, the robot needs to adjust the posture of the robot according to the pitch angle of the inner cover in the oiling process so that the robot can be opposite to the inner cover of the oil tank, and the rotation angle of the inner cover handle is correctly identified so as to finish the follow-up oiling action. Due to the change of illumination conditions and the difference of self poses of the inner cover of the oil tank, the work of visual identification of the inner cover has great difficulty.
Therefore, the design of the oil tank inner cover identification method which is suitable for different types of inner covers and can sense the pose parameters of the inner covers in the actual environment has very important significance for the application of the oiling robot.
Disclosure of Invention
Aiming at the inner cover of the oil tank of an automobile with uncertain pose and different local patterns in the actual environment, the invention provides a visual identification method which can adapt to the pattern change of the inner cover of the oil tank and accurately sense the pose parameters of the inner cover, and the visual identification method utilizes the common characteristics of different inner covers to learn and obtains the pose parameters (such as pitching degree, central coordinates and handle rotation angle) of the inner cover with higher precision on the basis of identifying the object of the inner cover.
The invention is realized by adopting the following technical scheme:
a method for recognizing an inner cover of an oil tank and sensing pose parameters based on vision comprises the following steps:
acquiring an initial image containing an inner cover of an automobile oil tank;
carrying out graying processing on the initial image to obtain a grayscale image of the initial image;
carrying out Gaussian filtering on the gray level image of the initial image to obtain an edge characteristic image containing the edge profile of the inner cover of the complete oil tank;
carrying out template matching based on an edge template on the edge characteristic image to obtain a suspected area of an inner cover of the oil tank;
carrying out self-adaptive histogram equalization processing on the initial image to enhance the contrast of the initial image; carrying out graph cutting by using the image with enhanced initial image contrast and the suspected area position of the inner cover of the oil tank to obtain an outline image of the area of the inner cover of the oil tank;
carrying out shape fitting on the outline image of the inner cover region of the oil tank based on morphological characteristics, measuring the pitching degree of the inner cover of the oil tank by using the coincidence degree of the fitting region, and using the center of the fitting region as the center coordinate of the inner cover of the oil tank; and performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray characteristics of the inner cover of the oil tank, fitting out a straight line in the handle direction of the inner cover of the oil tank, and obtaining the rotation angle of the inner cover handle, thereby realizing the identification and pose parameter perception of the inner cover of the oil tank.
Preferably, when the gray scale image of the initial image is subjected to gaussian filtering to obtain an edge characteristic image containing the edge profile of the complete inner cover of the oil tank:
and performing Gaussian filtering on the gray level image of the initial image by adopting a Gaussian convolution kernel, and then performing edge extraction on the gray level image by adopting a canny operator to obtain an edge characteristic image containing the edge profile of the inner cover of the complete oil tank.
Preferably, the process of performing template matching based on an edge template on the edge feature image to obtain the suspected region of the inner cover of the oil tank includes:
creating an initial template image according to the image size of an edge characteristic image containing the edge profile of the complete oil tank inner cover and the size range of the oil tank inner cover;
scaling the initial template image, adapting to the size change of an inner cover of an oil tank in the edge characteristic image within a preset range, sequentially performing template matching on each frame of image to be matched by using the template image obtained under each scaling, and selecting a matching position with the highest similarity degree as a matching result;
carrying out template matching on the edge characteristic image, and carrying out overall traversal type matching degree calculation on the image to be matched by taking the template image as a sliding window according to the coordinate position of the upper left corner of the circular edge template image on the edge characteristic image; and calculating the matching degree of each position, taking the coordinates of the matching points on the template image as pixel coordinates participating in the calculation of the matching degree, and performing pixel-level matching on the circular edge template image and the corresponding area on the edge characteristic image to obtain the suspected area of the inner cover of the oil tank.
Preferably, for the calculation of the matching degree of each position, when performing pixel-level matching between the circular edge template image and the corresponding region on the edge feature image by using the coordinates of the matching point on the template image as the pixel coordinates involved in the calculation of the matching degree:
matching is carried out by adopting a normalized correlation matching method, and a calculation formula of the matching degree is as follows:
Figure BDA0003378454870000031
in the formula, (x, y) is the coordinate of the matching position on the image to be matched; (x ', y') are coordinates of matching points on the template image; col _ t is the size of the template image in the width direction; row _ t is the size of the template image in the height direction; t (x ', y') is the pixel value of the corresponding position on the template image; i (x + x ', y + y') is the pixel value of the corresponding position on the image to be matched; and R (x, y) is the matching degree result on the (x, y) coordinate position.
Preferably, the adaptive histogram equalization processing is performed on the initial image to enhance the contrast of the initial image; the image with enhanced initial image contrast and the suspected area position of the oil tank inner cover are used for image cutting, and the process of obtaining the area outline of the oil tank inner cover comprises the following steps:
processing the initial image by adopting a self-adaptive histogram equalization method, dividing the initial image into a plurality of local regions, cutting off the height of a histogram of the local regions by using a preset threshold value, and controlling the contrast intensity of the local regions; segmenting an oil tank inner cover object by using an initial image after self-adaptive histogram equalization and an oil tank inner cover suspected area by using a graph cutting algorithm, performing iterative learning on the suspected area and the background area for a preset number of times by using the color and edge characteristics of the oil tank inner cover object, and segmenting a background area and a foreground area in the oil tank inner cover suspected area; obtaining a coding matrix with the same size as the initial image after iterative segmentation, re-coding the coding matrix, representing the background and the possible background by 0, and representing the foreground and the possible foreground by 1; and processing the initial image by using the coding matrix obtained by recoding to finally obtain an image only comprising the segmentation result area of the inner cover of the oil tank, wherein the image is an outline image containing the area of the inner cover of the oil tank.
Preferably, the shape fitting is performed on the contour image of the region of the inner cover of the oil tank based on the morphological characteristics, and the process of measuring the pitching degree of the inner cover of the oil tank by using the coincidence degree of the fitting region comprises the following steps:
graying the outline image of the inner cover area of the oil tank, and performing threshold segmentation on the grayed image by using a fixed threshold value to obtain a binary area image;
closing the binaryzation area image by a preset step, filling discrete black points in an inner cover area of the oil tank, opening the filled image by the preset step, removing an interference area outside the inner cover area of the oil tank, and only keeping the inner cover area of the oil tank;
extracting edge characteristics of the inner cover area of the oil tank, and performing ellipse fitting on the edge to obtain an approximate elliptical side line of the inner cover area of the oil tank and further obtain a circumscribed circle of the elliptical side line;
calculating the area of the difference area between the elliptical side line and the circumscribed circle of the elliptical side line, taking the area proportion as the evaluation index of the pitching degree of the inner cover of the oil tank, wherein the calculation formula of the evaluation index of the pitching degree of the inner cover of the oil tank is as follows:
Figure BDA0003378454870000041
in the formula, result _ position is a pitching degree evaluation index; s (Area _ circle-Area _ oval) is the Area of a phase difference Area between an elliptic side line obtained by fitting and a circumscribed circle of the elliptic side line; s (Area _ circle) is the circumscribed circle Area.
Preferably, based on the oil tank inner cup shape edge characteristic and the shade grey scale characteristic carry out double-deck straight line fitting, the straight line of oil tank inner cup handle direction is gone out in the fitting, and the process that obtains inner cup handle rotation angle includes:
graying the outline image of the inner cover area of the oil tank, extracting the edge of the obtained gray image of the inner cover area of the oil tank by adopting a canny operator with preset upper and lower threshold values, and carrying out Hough line detection with the detection threshold value as a preset value on the edge outline characteristic of the inner cover area of the oil tank to obtain a straight line set of the inner cover area of the oil tank;
performing preset-step closing operation on the linear set of the inner cover area of the oil tank to obtain a suspected handle direction area of the inner cover of the oil tank;
obtaining a dark area of a gray level image of an inner cover area of the oil tank by utilizing self-adaptive threshold segmentation, and obtaining a shadow area formed by focusing pixel points on the edge of a handle;
and reserving a pixel area along the handle direction in the shadow area by using the handle direction suspected area, performing straight line fitting on the pixel area to obtain a handle direction straight line, and obtaining a handle rotation angle according to the handle direction straight line.
The invention also provides a system for identifying the inner cover of the oil tank and sensing pose parameters based on vision, which comprises the following components:
an image acquisition module: the initial image containing the inner cover of the automobile oil tank is obtained;
a graying processing module: the gray level processing module is used for carrying out gray level processing on the initial image to obtain a gray level image of the initial image;
a noise reduction module: the method comprises the steps of performing Gaussian filtering on a gray scale image of an initial image to obtain an edge characteristic image containing the edge profile of an inner cover of the complete oil tank;
a suspected area obtaining module: the edge characteristic image matching device is used for carrying out template matching based on an edge template on the edge characteristic image to obtain a suspected area of an inner cover of the oil tank;
regional profile image extraction module of oil tank inner cup: the system is used for carrying out adaptive histogram equalization processing on the initial image and enhancing the contrast of the initial image; carrying out graph cutting by using the image with enhanced initial image contrast and the suspected area position of the inner cover of the oil tank to obtain an outline image of the area of the inner cover of the oil tank;
oil tank inner cup position appearance identification module: the shape fitting device is used for carrying out shape fitting on the outline image of the inner cover area of the oil tank based on morphological characteristics, measuring the pitching degree of the inner cover of the oil tank by utilizing the coincidence degree of the fitting area, and taking the center of the fitting area as the center coordinate of the inner cover of the oil tank; and performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray characteristics of the inner cover of the oil tank, fitting out a straight line in the handle direction of the inner cover of the oil tank, and obtaining the rotation angle of the inner cover handle, thereby realizing the identification and pose parameter perception of the inner cover of the oil tank.
The present invention also provides an electronic device, comprising:
one or more processors;
a storage device having one or more programs stored thereon;
when executed by the one or more processors, cause the one or more processors to implement the vision-based tank cap identification and pose parameter sensing method of the present invention as described above.
The invention also provides a storage medium on which a computer program is stored, wherein the computer program is executed by a processor to realize the vision-based tank inner cover identification and pose parameter perception method.
The invention has the following beneficial technical effects:
according to the method for recognizing the inner cover of the oil tank and sensing the pose parameters based on the vision, the suspected contour area is obtained by utilizing the shape commonality of the inner cover in the edge template matching mode, the suspected contour area is used as the input parameter of the graph cutting algorithm, the programmed automatic segmentation effect for the inner cover object of the oil tank is realized, and the manual marking is not needed under the condition of high-efficiency segmentation quality; and performing image cutting on the image subjected to the adaptive histogram equalization, and under the condition of enhancing the image contrast, ensuring a better target segmentation effect by using the color and contour characteristics of the target object when the local illumination environment is poor. Compared with the traditional attitude reverse calculation method in which the characteristic point calculation mode is adopted, the method has better application effect under the condition that the characteristic point grabbing precision cannot be ensured; a suspected area of the handle direction is determined by utilizing a straight line set obtained by fitting the edge characteristics of the inner cover, then pixel points of a handle shadow area in the suspected area are obtained by screening, secondary fitting is carried out on the pixel point set, and the determined straight line has better stability and higher precision aiming at the inner cover images in different environments.
Drawings
FIG. 1 is a flow chart of a method for identifying an inner cover of an oil tank and sensing pose parameters based on vision according to the invention;
FIG. 2(a) is an initial image to be processed in the embodiment of the present invention, and FIG. 2(b) is an edge feature image obtained by using a canny operator in the embodiment of the present invention;
fig. 3(a) is a circular template image created in the embodiment of the present invention, and fig. 3(b) is a matching result image obtained after edge template matching is performed in the embodiment of the present invention;
fig. 4(a) is an image obtained by performing adaptive histogram equalization (CLAHE) on an initial image in the embodiment of the present invention, and fig. 4(b) is a segmentation result image obtained by processing with a graph cut algorithm (GrabCut) in the embodiment of the present invention;
fig. 5(a) is a binary area image obtained by threshold segmentation of the segmented area in the embodiment of the present invention, and fig. 5(b) is a binary area image obtained by processing and only retaining the inner lid area in the embodiment of the present invention;
FIG. 6 is an image of a result of determining the pitch level according to an embodiment of the present invention;
fig. 7(a) is an edge feature image of an inner cover region image in the embodiment of the present invention, and fig. 7(b) is a result image obtained by using hough line detection in the embodiment of the present invention;
fig. 8(a) is a handle edge shadow area image of an inner cover area image in an embodiment of the present invention, and fig. 8(b) is a shadow area image retained by a handle direction suspected area in an embodiment of the present invention;
fig. 9(a) is a result image of straight line fitting performed on the image of the shaded area in the handle direction in the embodiment of the present invention, and fig. 9(b) is a result image of detection of a straight line in the handle rotation direction in the embodiment of the present invention;
fig. 10(a) is a graph of a position accuracy PRA in the embodiment of the present invention, and fig. 10(b) is a graph of a rotation angle recognition error in the embodiment of the present invention.
Detailed Description
The present invention is further described with reference to the accompanying drawings and the detailed description so that the advantages and features of the present invention will be readily understood by those skilled in the art, and the scope of the present invention will be clearly and clearly defined.
Referring to fig. 9(b) in fig. 1, the method for identifying the inner cover of the oil tank and sensing the pose parameters based on the vision comprises the following steps:
step 1) obtaining an initial image (RGB) containing an inner cover of an oil tank to be identified by using a camera of an oiling robot;
step 2) carrying out graying (RGB-GRAY) processing on the initial image, converting the initial image into a single-channel GRAY image, carrying out Gaussian filtering on the grayed initial image of the inner cover of the oil tank of the automobile, and carrying out low-threshold edge detection on the GRAY image of the initial image of the inner cover of the oil tank so as to obtain an edge characteristic image containing the complete edge profile of the inner cover;
specifically, during Gaussian filtering, Gaussian filtering is performed on the gray-scale image by adopting a Gaussian convolution kernel with the size of k × k, after noise reduction, edge extraction is performed on the image by adopting canny operators with upper and lower thresholds of thre _1 and thre _2, and the edge profile feature of the initial image is obtained. In the present invention, k is 3, thre _1 is 20, and thre _2 is 10.
Step 3) carrying out template matching on the edge characteristic image based on an edge template to obtain an inner cover suspected area; the specific implementation manner of the step 3) is as follows:
step 301) firstly, creating an initial template image according to the size of an image to be matched and the size range of an inner cover of an oil tank, wherein the size of the image to be matched is 640 multiplied by 480, and a circular template image with the size of 240 multiplied by 240, the circle center positioned at the center of the image and the radius pixel value of 110 is created;
step 302) scaling the template image to adapt to the size change of the inner cover of the oil tank in the image to be matched within a certain range, wherein the scaling adopted in the method is 1.1 to 1.5, template matching is carried out on each frame of image to be matched by sequentially utilizing the template images obtained under each scaling, and the matching position with the highest similarity degree is selected as a matching result;
step 303) performing template matching on the image to be matched, taking (x, y) as the coordinate position of the upper left corner of the template image on the image to be matched, and taking the template image as a sliding window to perform global traversal type matching degree calculation on the image to be matched, wherein x is 0,1,2, …, col _ i-col _ t, and y is 0,1,2, …, and row _ i-row _ t; for the matching degree calculation of each position, pixel-level matching of the corresponding region on the template image and the image to be matched is performed with (x ', y') as a pixel coordinate participating in the matching degree calculation, where x 'is 0,1,2, …, col _ t-1, y' is 0,1,2, …, row _ t-1; the matching process adopts a normalized correlation matching method for matching, and a matching degree calculation formula is as follows:
Figure BDA0003378454870000081
in the formula (1), (x, y) is the coordinate of the matching position on the image to be matched; (x ', y') are coordinates of matching points on the template image; col _ t is the size of the template image in the width direction; row _ t is the size of the template image in the height direction; t (x ', y') is the pixel value of the corresponding position on the template image; i (x + x ', y + y') is the pixel value of the corresponding position on the image to be matched; and R (x, y) is the matching degree result on the (x, y) coordinate position.
Step 4) carrying out adaptive histogram equalization (CLAHE) processing on the color initial image (RGB) to enhance the image contrast; specifically, an adaptive histogram equalization algorithm (CLAHE) is adopted to process an initial image, the initial image is divided into a plurality of local regions of M multiplied by M, histogram equalization is carried out to obtain an accumulative distribution function of each region, the maximum slope of the accumulative distribution is limited by limiting the maximum height of the histogram, the height of the histogram is cut off by a threshold value T, the cut-off part is uniformly distributed in the whole gray scale range, and then the contrast intensity of the local regions is controlled; the pixel variation range of the image is expanded through gray level remapping, so that the image contrast is obviously improved. In the present invention, M is 8 and T is 2.
Step 5) carrying out graph cutting by utilizing the equalized image and the suspected area position of the inner cover, and segmenting to obtain a relatively complete oil tank inner cover area with a basic outline; specifically, a graph cut algorithm (GrabCT) is adopted, segmentation of an inner cover object is carried out by utilizing an initial image after self-adaptive histogram equalization and an inner cover suspected region (rectangular frame region), iterative learning is carried out on the suspected region and a non-suspected region (background) for d times by utilizing the color and edge characteristics of the target object, and a background region and a foreground region (target object) are segmented in the suspected region; obtaining a coding matrix with the same size as the initial image after iterative segmentation, re-coding the coding matrix, representing the background and the possible background by 0, and representing the foreground and the possible foreground by 1; and processing the initial image by using the coding matrix obtained by recoding to finally obtain the image only containing the segmentation result area. In the present invention, d is preferably 5.
Step 6) carrying out shape fitting on the separated outline of the inner cover area based on morphological characteristics, measuring the pitching degree of the inner cover by using the coincidence degree of the fitting area, and using the center of the fitting area as the center coordinate of the inner cover; specifically, the specific implementation method of the pitch degree perception based on the morphological feature fitting is as follows:
step 601) graying the area image of the inner cover of the oil tank obtained by segmentation, and performing threshold segmentation on the grayed image by a fixed threshold value thre _3 to obtain a binaryzation area image;
step 602) closing operation of k _1 order is firstly carried out on the image of the binarization area to fill discrete black points in the inner cover area, opening operation of k _2 order is then carried out on the filled image to remove the interference area outside the inner cover area, and only the inner cover area of the oil tank is reserved;
step 603), extracting the edge characteristics of the inner cover region obtained after the treatment, and performing ellipse fitting on the edge to obtain an approximate inner cover region ellipse side line and further obtain a circumscribed circle of the ellipse side line;
step 604) calculating the area of a difference area between the elliptic contour and the circumscribed circle contour obtained by fitting, taking the area proportion as an evaluation index of the pitching degree of the inner cover, wherein the closer the index is to 0, the smaller the pitching angle of the inner cover is, and the calculation formula of the index is as follows:
Figure BDA0003378454870000101
in the formula (2), result _ position is a pitching degree evaluation index; s (Area _ circle-Area _ oval) is the Area of a phase difference Area of the elliptic Area and the external circle Area obtained by fitting; s (Area _ circle) is the circumscribed circle Area.
In the present invention, more preferably, thre _3 is 5, k _1 is 3, and k _2 is 21.
And 7) performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray characteristics of the inner cover, and fitting out a straight line in the handle direction to obtain the rotation angle of the inner cover handle. Specifically, the specific implementation method of handle rotation angle sensing based on the inner cover shape edge feature and the shadow gray scale feature is as follows:
step 701) graying the image of the inner cover area of the oil tank obtained by segmentation, extracting the edge of the grayed image by adopting canny operators with the upper and lower thresholds thre _4 and thre _5, and detecting the Hough line with the threshold thre _6 for the edge contour characteristics of the inner cover area;
step 702) performing k _ 3-order closing operation on the detected straight line to obtain a suspected area of the handle direction of the inner cover of the oil tank;
step 703) obtaining a dark area of the gray image of the inner cover area by utilizing self-adaptive threshold segmentation, and obtaining a shadow area formed by focusing pixel points on the edge of the handle;
step 704) retaining a pixel area along the handle direction in the shadow area by using the suspected area in the handle direction, and performing straight line fitting on the pixel area to obtain a straight line in the handle direction, so as to obtain a handle rotation angle.
In the present invention, more preferably, thre _4 is 120, thre _5 is 80, thre _6 is 75, and k _3 is 15.
Examples
Referring to fig. 1, the method for identifying the inner cover of the oil tank and sensing the pose parameters based on the vision provided by the embodiment includes the following steps:
step 1) an oil tank inner cover initial image to be identified is obtained by an oil filling robot camera; step 2) carrying out Gaussian filtering on the initial image of the inner cover of the automobile oil tank subjected to the graying treatment, and carrying out low-threshold edge detection on the gray image to obtain an edge characteristic image containing the edge profile of the complete inner cover; step 3) carrying out template matching on the edge characteristic image based on an edge template to obtain an inner cover suspected area; step 4) carrying out adaptive histogram equalization (CLAHE) processing on the color initial image (RGB) to enhance the image contrast; step 5) carrying out graph cutting by utilizing the equalized image and the suspected area position of the inner cover, and segmenting to obtain a relatively complete oil tank inner cover area with a basic outline; step 6) carrying out shape fitting on the separated outline of the inner cover area based on morphological characteristics, measuring the pitching degree of the inner cover by using the coincidence degree of the fitting area, and using the center of the fitting area as the center coordinate of the inner cover; and 7) performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray characteristics of the inner cover, and fitting out a straight line in the handle direction to obtain the rotation angle of the inner cover handle.
Referring to fig. 2(a), an initial image to be recognized is obtained in an actual environment in the embodiment; fig. 2(b) is an edge feature image obtained by processing, and the initial image is grayed (RGB-GRAY) and converted into a single-channel grayscale image, the grayscale image is gaussian filtered by using a gaussian convolution kernel of 3 × 3 size, and after noise reduction, the image is edge-extracted by using a canny operator to obtain the edge contour feature of the initial image. As shown in fig. 2(b), the edge feature image should include the complete contour of the inner cover.
Referring to fig. 3(a), the size of the image to be matched is 640 × 480, the size of the circular template image is 240 × 240, the center of the circle is located at the center of the image, the radius pixel value is 110, and the size change of different inner covers is adapted by scaling in the matching process; fig. 3(b) is a matching result image obtained after edge template matching, after a circular template image is created, the template image is subjected to scaling change to adapt to the size change of an inner cover of an oil tank in the image to be matched within a certain range, the scaling adopted in the method provided by the invention is 1.1 to 1.5, template matching is performed on each frame of image to be matched by using the template image obtained under each scaling in sequence, and the matching position with the highest similarity degree is selected as the matching result; performing template matching on an image to be matched, taking (x, y) as the coordinate position of the upper left corner of the template image on the image to be matched, and taking the template image as a sliding window to perform global traversal type matching degree calculation on the image to be matched, wherein x is 0,1,2, …, col _ i-col _ t, y is 0,1,2, …, and row _ i-row _ t; for the matching degree calculation of each position, pixel level matching of the template image and the corresponding region on the image to be matched is carried out by taking (x ', y') as pixel coordinates participating in the matching degree calculation, wherein x 'is 0,1,2, …, col _ t-1, y' is 0,1,2, … and row _ t-1, and the matching process adopts a normalized correlation matching method for matching. As shown, the suspected location of the inner lid area obtained from the matching result can completely include the inner lid area of the fuel tank.
Referring to fig. 4(a), in the embodiment, an image obtained by performing adaptive histogram equalization on an initial image is processed by using an adaptive histogram equalization algorithm (CLAHE), the initial image is divided into a plurality of local regions of M × M, histogram equalization is performed to obtain an accumulative distribution function of each region, the maximum slope of the accumulative distribution is limited by limiting the maximum height of the histogram, the height of the histogram is truncated by a threshold T, the truncated portion is uniformly distributed in the whole gray scale range, and the contrast intensity of the local region is further controlled; the pixel change range of the image is expanded through gray level remapping, so that the image contrast is obviously improved, wherein M is 8, and T is 2; fig. 4(b) is a segmentation result image obtained by using a graph cut algorithm (GrabCut), the graph cut algorithm is used to segment the inner lid object by using the initial image after the adaptive histogram equalization and the inner lid suspected region (rectangular frame region), the color and edge features of the target object are used to perform 5 times of iterative learning on the suspected region and the non-suspected region (background), and the background region and the foreground region (target object) are segmented in the suspected region; obtaining a coding matrix with the same size as the initial image after iterative segmentation, re-coding the coding matrix, representing the background and the possible background by 0, and representing the foreground and the possible foreground by 1; and processing the initial image by using the coding matrix obtained by recoding to finally obtain an image only comprising a segmentation result area, wherein the edge profile of the inner cover area obtained by segmentation is complete, and has local deficiency or residue, but all belong to the algorithm control range.
Referring to fig. 5(a), in the embodiment, the binarized area image is obtained by performing threshold segmentation on the segmented area image, the grayscale image of the oil tank inner cover area obtained by segmentation is performed, and the threshold segmentation is performed on the grayscale image by a fixed threshold 5 to obtain the binarized area image; fig. 5(b) is a binary area image which is processed and only remains an inner cover area, the binary area image is firstly subjected to 3-step closing operation to fill discrete black dots in the inner cover area, then the filled image is subjected to 21-step opening operation to remove an interference area outside the inner cover area, only the inner cover area of the oil tank is remained, and image results show that the inner cover area contour does not deform to a large extent and the residual interference area due to segmentation is successfully removed.
Referring to fig. 6, in the embodiment, the image of the pitch degree determination result obtained based on the morphological feature is obtained, the edge feature of the inner lid region obtained after the processing is extracted, ellipse fitting is performed on the edge, an approximate inner lid region ellipse side line is obtained, and a circumscribed circle of the ellipse side line is further obtained; calculating the area of a difference area between the elliptic contour and the circumscribed circle contour obtained by fitting, and taking the area proportion (the percentage of the difference area in the circumscribed circle contour area) as an evaluation index of the pitching degree of the inner cover, wherein the index is closer to 0 and represents that the pitching angle of the inner cover is smaller; the result of the pitch degree judgment shows that the evaluation index is about 0.023, and the elliptic contour and the circumscribed circle contour obtained by fitting are basically overlapped.
Referring to fig. 7(a) and 7(b), the images of the tank inner lid region obtained by the segmentation are grayed, edge extraction is performed on the grayed images by using canny operators with upper and lower thresholds of 120 and 80, and hough line detection with a detection threshold of 75 is performed on the edge profile characteristics of the inner lid region. Fig. 7(a) is an edge feature of the inner cap region obtained by edge extraction in the embodiment, and fig. 7(b) is a detection straight line result image obtained by hough straight line detection, and the detected straight lines are basically concentrated in two parallel side regions of the inner cap handle.
Referring to fig. 8(a), a dark area of a gray scale image of an inner cap area is obtained by adaptive threshold segmentation for a shadow area image of a handle edge of the inner cap area image, and a shadow area formed by focusing pixels on the handle edge is obtained; fig. 8(b) is a shadow area image retained by using the handle direction pseudo area, a 15-step closing operation is performed on the straight line set obtained by using hough straight line detection to obtain a handle direction pseudo area of the inner cover of the oil tank, a pixel area along the handle direction in the shadow area is retained by using the handle direction pseudo area in an image difference mode, and finally a pixel point set distributed along the handle direction is obtained.
Referring to fig. 9(a) and 9(b), a straight line is fitted to the pixel regions distributed along the grip direction to obtain a grip direction straight line, and thus a grip rotation angle. Fig. 9(a) is a result image of straight line fitting performed on the image of the shaded area in the handle direction in the embodiment, fig. 9(b) is a detection result image of a straight line in the handle rotation direction, the pixel coordinates of the center point of the inner lid area and the detected handle rotation angle are displayed in the upper left corner of the detection result image, and the detected result is along the handle direction to achieve the expected detection expectation in visual effect.
Referring to fig. 10, the positioning accuracy of the tank lid identification is evaluated by using a position accuracy indicator pra (position registration accuracy), which measures the deviation degree of the center point of the tank lid relative to the ideal center point position obtained by identification, and the accuracy of angle detection is measured by using a rotation angle error Δ θ (difference between the identification angle and the theoretical angle). The calculation formula is as follows:
Figure BDA0003378454870000141
in the formula (3), Distance (P)1,P0) Identifying the pixel distance between the central point of the area and the central point of the marked ideal area; threshold _ distance is the maximum pixel deviation distance allowed by the error; theta1The rotation angle of the inner cover of the oil tank is obtained by image recognition; theta0The theoretical rotating angle of the inner cover of the oil tank is marked manually.
And (4) selecting 89 oil tank inner cover samples with ideal identification information (the central point position and the standard inner cover rotation angle) generated by simulation software in actual experiments, and comparing a sensing result with a theoretical result according to the evaluation index calculation mode to obtain a corresponding result curve graph. Fig. 10(a) is a graph of the position accuracy indicator PRA, the average position accuracy of the proposed method is 0.058, and the maximum pixel deviation distance allowed by the error is 30, i.e. the average position deviation is within 2 pixels; fig. 10(b) is a graph of angle recognition error, and the average error of the rotation angle recognition of the inner lid in the method of the present invention is 0.3726 degrees, which is less than 1 degree.
In conclusion, the self-adaptive identification and segmentation method provided by the invention can be used for carrying out self-adaptive identification and segmentation on the inner cover of the oil tank with local shape difference aiming at different illumination environments such as poor local illumination, and the posture parameters sensed on the basis have higher precision and better stability.

Claims (10)

1. A method for recognizing an inner cover of an oil tank and sensing pose parameters based on vision is characterized by comprising the following processes:
acquiring an initial image containing an inner cover of an automobile oil tank;
carrying out graying processing on the initial image to obtain a grayscale image of the initial image;
carrying out Gaussian filtering on the gray level image of the initial image to obtain an edge characteristic image containing the edge profile of the inner cover of the complete oil tank;
carrying out template matching based on an edge template on the edge characteristic image to obtain a suspected area of an inner cover of the oil tank;
carrying out self-adaptive histogram equalization processing on the initial image to enhance the contrast of the initial image; carrying out graph cutting by using the image with enhanced initial image contrast and the suspected area position of the inner cover of the oil tank to obtain an outline image of the area of the inner cover of the oil tank;
carrying out shape fitting on the outline image of the inner cover region of the oil tank based on morphological characteristics, measuring the pitching degree of the inner cover of the oil tank by using the coincidence degree of the fitting region, and using the center of the fitting region as the center coordinate of the inner cover of the oil tank; and performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray characteristics of the inner cover of the oil tank, fitting out a straight line in the handle direction of the inner cover of the oil tank, and obtaining the rotation angle of the inner cover handle, thereby realizing the identification and pose parameter perception of the inner cover of the oil tank.
2. The method for identifying the oil tank inner cover and sensing the pose parameters based on the vision as claimed in claim 1, wherein when a gray scale image of an initial image is subjected to Gaussian filtering to obtain an edge feature image containing a complete oil tank inner cover edge profile:
and performing Gaussian filtering on the gray level image of the initial image by adopting a Gaussian convolution kernel, and then performing edge extraction on the gray level image by adopting a canny operator to obtain an edge characteristic image containing the edge profile of the inner cover of the complete oil tank.
3. The method for identifying the inner cover of the oil tank and sensing the pose parameters based on the vision as claimed in claim 1, wherein the process of performing template matching based on an edge template on the edge feature image to obtain the suspected area of the inner cover of the oil tank comprises the following steps:
creating an initial template image according to the image size of an edge characteristic image containing the edge profile of the complete oil tank inner cover and the size range of the oil tank inner cover;
scaling the initial template image, adapting to the size change of an inner cover of an oil tank in the edge characteristic image within a preset range, sequentially performing template matching on each frame of image to be matched by using the template image obtained under each scaling, and selecting a matching position with the highest similarity degree as a matching result;
carrying out template matching on the edge characteristic image, and carrying out overall traversal type matching degree calculation on the image to be matched by taking the template image as a sliding window according to the coordinate position of the upper left corner of the circular edge template image on the edge characteristic image; and calculating the matching degree of each position, taking the coordinates of the matching points on the template image as pixel coordinates participating in the calculation of the matching degree, and performing pixel-level matching on the circular edge template image and the corresponding area on the edge characteristic image to obtain the suspected area of the inner cover of the oil tank.
4. The visual-based fuel tank inner cover identification and pose parameter sensing method according to claim 3, wherein for the calculation of the matching degree of each position, the coordinates of the matching points on the template image are taken as the pixel coordinates involved in the calculation of the matching degree, and when the pixel-level matching of the circular edge template image and the corresponding area on the edge feature image is performed:
matching is carried out by adopting a normalized correlation matching method, and a calculation formula of the matching degree is as follows:
Figure FDA0003378454860000021
in the formula, (x, y) is the coordinate of the matching position on the image to be matched; (x ', y') are coordinates of matching points on the template image; col _ t is the size of the template image in the width direction; row _ t is the size of the template image in the height direction; t (x ', y') is the pixel value of the corresponding position on the template image; i (x + x ', y + y') is the pixel value of the corresponding position on the image to be matched; and R (x, y) is the matching degree result on the (x, y) coordinate position.
5. The vision-based oil tank inner cover identification and pose parameter sensing method according to claim 1, characterized in that adaptive histogram equalization processing is performed on the initial image to enhance the contrast of the initial image; the image with enhanced initial image contrast and the suspected area position of the oil tank inner cover are used for image cutting, and the process of obtaining the area outline of the oil tank inner cover comprises the following steps:
processing the initial image by adopting a self-adaptive histogram equalization method, dividing the initial image into a plurality of local regions, cutting off the height of a histogram of the local regions by using a preset threshold value, and controlling the contrast intensity of the local regions; segmenting an oil tank inner cover object by using an initial image after self-adaptive histogram equalization and an oil tank inner cover suspected area by using a graph cutting algorithm, performing iterative learning on the suspected area and the background area for a preset number of times by using the color and edge characteristics of the oil tank inner cover object, and segmenting a background area and a foreground area in the oil tank inner cover suspected area; obtaining a coding matrix with the same size as the initial image after iterative segmentation, re-coding the coding matrix, representing the background and the possible background by 0, and representing the foreground and the possible foreground by 1; and processing the initial image by using the coding matrix obtained by recoding to finally obtain an image only comprising the segmentation result area of the inner cover of the oil tank, wherein the image is an outline image containing the area of the inner cover of the oil tank.
6. The method for identifying the oil tank inner cover and sensing the pose parameter based on the vision as claimed in claim 1, wherein the process of performing shape fitting on the outline image of the oil tank inner cover area based on morphological characteristics and measuring the pitching degree of the oil tank inner cover by using the coincidence degree of the fitting area comprises the following steps:
graying the outline image of the inner cover area of the oil tank, and performing threshold segmentation on the grayed image by using a fixed threshold value to obtain a binary area image;
closing the binaryzation area image by a preset step, filling discrete black points in an inner cover area of the oil tank, opening the filled image by the preset step, removing an interference area outside the inner cover area of the oil tank, and only keeping the inner cover area of the oil tank;
extracting edge characteristics of the inner cover area of the oil tank, and performing ellipse fitting on the edge to obtain an approximate elliptical side line of the inner cover area of the oil tank and further obtain a circumscribed circle of the elliptical side line;
calculating the area of the difference area between the elliptical side line and the circumscribed circle of the elliptical side line, taking the area proportion as the evaluation index of the pitching degree of the inner cover of the oil tank, wherein the calculation formula of the evaluation index of the pitching degree of the inner cover of the oil tank is as follows:
Figure FDA0003378454860000031
in the formula, result _ position is a pitching degree evaluation index; s (Area _ circle-Area _ oval) is the Area of a phase difference Area between an elliptic side line obtained by fitting and a circumscribed circle of the elliptic side line; s (Area _ circle) is the circumscribed circle Area.
7. The method for identifying the oil tank inner cover and sensing the pose parameters based on the vision as claimed in claim 1, wherein the process of fitting a double-layer straight line based on the shape edge characteristics and the shadow gray scale characteristics of the oil tank inner cover to fit a straight line in the direction of the oil tank inner cover handle to obtain the rotation angle of the oil tank inner cover handle comprises the following steps:
graying the outline image of the inner cover area of the oil tank, extracting the edge of the obtained gray image of the inner cover area of the oil tank by adopting a canny operator with preset upper and lower threshold values, and carrying out Hough line detection with the detection threshold value as a preset value on the edge outline characteristic of the inner cover area of the oil tank to obtain a straight line set of the inner cover area of the oil tank;
performing preset-step closing operation on the linear set of the inner cover area of the oil tank to obtain a suspected handle direction area of the inner cover of the oil tank;
obtaining a dark area of a gray level image of an inner cover area of the oil tank by utilizing self-adaptive threshold segmentation, and obtaining a shadow area formed by focusing pixel points on the edge of a handle;
and reserving a pixel area along the handle direction in the shadow area by using the handle direction suspected area, performing straight line fitting on the pixel area to obtain a handle direction straight line, and obtaining a handle rotation angle according to the handle direction straight line.
8. The utility model provides an oil tank inner cup discernment and position appearance parameter perception system based on vision which characterized in that includes:
an image acquisition module: the initial image containing the inner cover of the automobile oil tank is obtained;
a graying processing module: the gray level processing module is used for carrying out gray level processing on the initial image to obtain a gray level image of the initial image;
a noise reduction module: the method comprises the steps of performing Gaussian filtering on a gray scale image of an initial image to obtain an edge characteristic image containing the edge profile of an inner cover of the complete oil tank;
a suspected area obtaining module: the edge characteristic image matching device is used for carrying out template matching based on an edge template on the edge characteristic image to obtain a suspected area of an inner cover of the oil tank;
regional profile image extraction module of oil tank inner cup: the system is used for carrying out adaptive histogram equalization processing on the initial image and enhancing the contrast of the initial image; carrying out graph cutting by using the image with enhanced initial image contrast and the suspected area position of the inner cover of the oil tank to obtain an outline image of the area of the inner cover of the oil tank;
oil tank inner cup position appearance identification module: the shape fitting device is used for carrying out shape fitting on the outline image of the inner cover area of the oil tank based on morphological characteristics, measuring the pitching degree of the inner cover of the oil tank by utilizing the coincidence degree of the fitting area, and taking the center of the fitting area as the center coordinate of the inner cover of the oil tank; and performing double-layer straight line fitting based on the shape edge characteristics and the shadow gray characteristics of the inner cover of the oil tank, fitting out a straight line in the handle direction of the inner cover of the oil tank, and obtaining the rotation angle of the inner cover handle, thereby realizing the identification and pose parameter perception of the inner cover of the oil tank.
9. An electronic device, comprising:
one or more processors;
a storage device having one or more programs stored thereon;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the vision-based in-tank lid identification and pose parameter awareness method of any of claims 1-7.
10. A storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the vision-based tank cap identification and pose parameter sensing method according to any one of claims 1 to 7.
CN202111425958.1A 2021-11-26 2021-11-26 Vision-based oil tank inner cover identification and pose parameter sensing method Active CN114359597B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111425958.1A CN114359597B (en) 2021-11-26 2021-11-26 Vision-based oil tank inner cover identification and pose parameter sensing method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111425958.1A CN114359597B (en) 2021-11-26 2021-11-26 Vision-based oil tank inner cover identification and pose parameter sensing method

Publications (2)

Publication Number Publication Date
CN114359597A true CN114359597A (en) 2022-04-15
CN114359597B CN114359597B (en) 2023-09-12

Family

ID=81098174

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111425958.1A Active CN114359597B (en) 2021-11-26 2021-11-26 Vision-based oil tank inner cover identification and pose parameter sensing method

Country Status (1)

Country Link
CN (1) CN114359597B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116703784A (en) * 2023-08-02 2023-09-05 济南宝林信息技术有限公司 Heart ultrasonic image vision enhancement method

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110378376A (en) * 2019-06-12 2019-10-25 西安交通大学 A kind of oil filler object recognition and detection method based on machine vision
WO2021000524A1 (en) * 2019-07-03 2021-01-07 研祥智能科技股份有限公司 Hole protection cap detection method and apparatus, computer device and storage medium
CN112734750A (en) * 2021-01-25 2021-04-30 上海交通大学 Tank cover visual detection and positioning method and system, medium and device
CN113436212A (en) * 2021-06-22 2021-09-24 广西电网有限责任公司南宁供电局 Extraction method for inner contour of circuit breaker static contact meshing state image detection

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110378376A (en) * 2019-06-12 2019-10-25 西安交通大学 A kind of oil filler object recognition and detection method based on machine vision
WO2021000524A1 (en) * 2019-07-03 2021-01-07 研祥智能科技股份有限公司 Hole protection cap detection method and apparatus, computer device and storage medium
CN112734750A (en) * 2021-01-25 2021-04-30 上海交通大学 Tank cover visual detection and positioning method and system, medium and device
CN113436212A (en) * 2021-06-22 2021-09-24 广西电网有限责任公司南宁供电局 Extraction method for inner contour of circuit breaker static contact meshing state image detection

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
李轩;刘云清;: "基于似圆阴影的光学遥感图像油罐检测", 电子与信息学报, no. 06 *
段志达;魏利胜;刘小珲;丁坤;: "基于Hough变换圆检测和边缘模板匹配的轴承缺陷检测与定位", 安徽工程大学学报, no. 04 *
甘文龙;罗会信;王中任;: "管道焊接激光视觉跟踪的定位方法研究", 激光与红外, no. 06 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116703784A (en) * 2023-08-02 2023-09-05 济南宝林信息技术有限公司 Heart ultrasonic image vision enhancement method
CN116703784B (en) * 2023-08-02 2023-10-20 济南宝林信息技术有限公司 Heart ultrasonic image vision enhancement method

Also Published As

Publication number Publication date
CN114359597B (en) 2023-09-12

Similar Documents

Publication Publication Date Title
CN109443480B (en) Water level scale positioning and water level measuring method based on image processing
CN108389179B (en) Machine vision-based can cover surface defect detection method
CN107545239B (en) Fake plate detection method based on license plate recognition and vehicle characteristic matching
CN110348451B (en) Automatic box number acquisition and identification method in railway container loading and unloading process
CN109657632B (en) Lane line detection and identification method
CN108256521B (en) Effective area positioning method for vehicle body color identification
CN110378376A (en) A kind of oil filler object recognition and detection method based on machine vision
CN111915704A (en) Apple hierarchical identification method based on deep learning
WO2022205525A1 (en) Binocular vision-based autonomous underwater vehicle recycling guidance false light source removal method
CN111709417B (en) License plate recognition method for mine car
CN115115612B (en) Surface defect detection method and system for mechanical parts
CN112053376B (en) Workpiece weld joint identification method based on depth information
CN115018846B (en) AI intelligent camera-based multi-target crack defect detection method and device
CN116597392B (en) Hydraulic oil impurity identification method based on machine vision
CN112734729B (en) Water gauge water level line image detection method and device suitable for night light supplement condition and storage medium
CN112734750B (en) Tank cover visual detection and positioning method and system, medium and device
CN115147448A (en) Image enhancement and feature extraction method for automatic welding
CN114445707A (en) Intelligent visual fine detection method for defects of bottled water labels
CN109670498A (en) A kind of license plate locating method
CN107516315B (en) Tunneling machine slag tapping monitoring method based on machine vision
CN114359597B (en) Vision-based oil tank inner cover identification and pose parameter sensing method
CN112435272A (en) High-voltage transmission line connected domain removing method based on image contour analysis
CN114241438B (en) Traffic signal lamp rapid and accurate identification method based on priori information
CN115841633A (en) Power tower and power line associated correction power tower and power line detection method
CN113053164A (en) Parking space identification method using look-around image

Legal Events

Date Code Title Description
PB01 Publication
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