CN111553422A - Automatic identification and recovery method and system for surgical instruments - Google Patents
Automatic identification and recovery method and system for surgical instruments Download PDFInfo
- Publication number
- CN111553422A CN111553422A CN202010352809.6A CN202010352809A CN111553422A CN 111553422 A CN111553422 A CN 111553422A CN 202010352809 A CN202010352809 A CN 202010352809A CN 111553422 A CN111553422 A CN 111553422A
- Authority
- CN
- China
- Prior art keywords
- image
- surgical instruments
- images
- quality
- block
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for automatically identifying and recovering surgical instruments, which are applied to the field of medical imaging, wherein the method comprises the following steps: acquiring an image and quality information of a surgical instrument to be recovered, performing primary processing on the image information, performing image identification on the image subjected to the primary processing to acquire the category of the image, performing quality verification on the corresponding surgical instrument in the classified image, and placing the surgical instrument on a recovery table by a rear grabbing device after the verification is passed; the system comprises a computer unit, an image acquisition device, a quality acquisition device and a recovery platform, wherein the computer unit comprises a database unit, an image processing device, an image recognition device and a quality matching device; the image recognition apparatus includes a block image generation unit, a feature amount calculation unit, a separation hyperplane calculation unit, and a type discrimination unit. According to the medical disinfection bag, the image recognition and the quality matching are combined, so that the recovery time of surgical instruments is saved, the working efficiency is improved, and the recovery accuracy of the surgical instruments in the medical disinfection bag is improved.
Description
Technical Field
The invention relates to the field of medical imaging, in particular to a method and a system for automatically identifying and recycling surgical instruments.
Background
At present, surgical instruments recovered by a hospital disinfection supply center are mainly recovered manually, and the counting and recovery of the used instruments and materials are very complicated work. Surgical instruments need to be manually checked, names of the recovery packages are checked, and the like, but the models of the surgical instruments are complex, the working efficiency is greatly reduced, time and labor are wasted, and mistakes are easy to make. In addition, the instruments can be contacted during manual counting, and pollution is easily caused.
In an image recognition method and an image pickup apparatus of patent publication No. CN101650783B, when recognizing an object using feature amounts of block images of an image picked up by an image pickup apparatus, the object is recognized with high accuracy while suppressing processing cost of a CPU, image data is divided into a plurality of blocks to generate block images, and the feature amounts of the respective block images are calculated using color space information and frequency components of the block images. Further, as the teacher data, the image feature amount for each type is calculated in advance, the separation hyperplane which becomes the boundary of the recognition type is calculated using the feature amount, the image feature amount of the block image is calculated similarly for the newly acquired image, and the type to which the block image belongs is determined using the distance from the separation hyperplane of each type. However, in the patent, only a single image local feature amount is adopted as a standard for recognition and judgment, and the recognition efficiency and the recognition accuracy are low.
Disclosure of Invention
The technical purpose is as follows: aiming at the defects of low efficiency and low precision of image recognition surgical instruments in the prior art, the invention discloses an automatic recognition and recovery method and system for surgical instruments.
The technical scheme is as follows: in order to achieve the technical purpose, the invention adopts the following technical scheme:
an automatic identification and recovery method for surgical instruments comprises the following steps:
s1, acquiring an original image: acquiring clear and complete original images of surgical instruments in the medical disinfection package to be recycled;
s2, performing primary processing on the original image: extracting, coding and compressing an original image, converting the image into a JPEG format according to the resolution and color of the image, and acquiring a primary processing image;
s3, performing image recognition on the primary processing image, and acquiring the category to which the corresponding surgical instrument belongs in the primary image: dividing the primary processing image of the surgical instrument into a plurality of images, acquiring standard images of all classes of the surgical instrument in a database unit, and acquiring a feature vector corresponding to the standard image of each class of the surgical instrument for each image; extracting a separation hyperplane of each surgical instrument type from a database unit, calculating the distance between each feature vector and the separation hyperplane of a standard image of the corresponding surgical instrument type, preliminarily judging the type of a block image according to the distance, and further obtaining the primary classification of the corresponding surgical instrument in the block image;
s4, performing quality verification on the surgical instruments after primary classification: the corresponding surgical instruments in the primarily classified block images are captured through the capturing device, the quality information of the corresponding surgical instruments in the primarily classified block images is obtained through the quality acquisition device, the primarily classified results of the surgical instruments output from the image recognition device are obtained, the obtained quality information of the surgical instruments is compared with the quality of the surgical instruments of the type in the database unit for verification, if the quality error is smaller than a threshold value, the type is judged reasonably, the surgical instruments are placed at the corresponding types on the recovery platform through the capturing device, otherwise, the step S3 is returned, and the feature vectors are recalculated for the block images.
Preferably, the generating process of the block image in S3 is: and carrying out target marking and background marking on the images of the surgical instruments in the medical disinfection package to be recovered, and segmenting the images containing the target marks into a plurality of non-overlapping sub-regions according to image characteristics, wherein each sub-region is a block image to be detected and processed.
Preferably, the image characteristics include image color, shape, gray scale, texture, and pixels.
Preferably, the calculation method of the feature vector in S3 is as follows: and carrying out image matching on the block images and all the standard images by using an SIFT algorithm, wherein the output SIFT feature vector is the feature vector to be solved.
Preferably, the method for calculating the separation hyperplane of each surgical instrument type in the database unit in S3 includes: the standard image of each surgical instrument type is obtained from the database unit, the separating hyperplane of each type is calculated by the separating hyperplane calculating part in the image recognition device, and the separating hyperplane of each type is stored in the database unit.
An automatic identification and recovery system for surgical instruments realizes the automatic identification and recovery method for surgical instruments, and comprises a computer unit, an image acquisition device, a quality acquisition device and a recovery platform, wherein the image acquisition device, the quality acquisition device and the recovery platform are connected with the computer unit;
the image acquisition device comprises a camera for shooting images of surgical instruments in the medical disinfection bag, monitoring the change of a recovered object in time, capturing image change information and using the change information for computer unit identification;
the quality acquisition device is used for acquiring the quality information of the surgical instruments in the medical disinfection bag to be recovered;
the recovery platform is used for classifying and placing the recovered surgical instruments;
the database unit is used for storing all categories of surgical instruments in the medical disinfection package, image information, quality information and separation hyperplane of the surgical instruments, wherein the standard images of the surgical instruments of all the categories are of block images;
the image processing device is used for receiving the image information input by the image acquisition device and coding and compressing the input image information;
the image recognition device comprises a block image generation part, a characteristic quantity calculation part, a separation hyperplane calculation part and a type discrimination part; wherein, the image generation department of the block, is used for dividing the image information outputted by the image processing unit into several block images, and input all block images into the characteristic quantity calculating department; the feature quantity calculating part is used for calculating a feature vector according to the block image; the separation hyperplane calculating part is used for calculating a separation hyperplane of each type according to the standard image of each type of the surgical instrument stored in the database; the type judging part is used for judging the type of the block image according to the distance between the feature vector and the separation hyperplane of each type;
and the quality matching device is used for acquiring the quality information transmitted by the quality acquisition device and the standard quality information stored in the database unit and comparing the quality information with the standard quality information.
Preferably, the computer unit further comprises a main storage unit, the main storage unit stores information in the database unit, and the main storage unit comprises a RAM and a ROM.
Preferably, the system further comprises a portable terminal comprising a communication module for communicating with the computer unit and a secondary storage unit, the secondary storage unit storing information within the database unit.
Has the advantages that:
according to the automatic identification and recovery system for the surgical instruments, the SIFT algorithm is used for calculating the feature vectors of the block images in the image identification method, the block images are kept invariable in rotation, scale scaling and brightness change, the method has no requirement on the number of feature points and the proportion of effective points, and when the number of the feature points is not large, the SIFT matching algorithm can be combined with the feature vectors in other forms; meanwhile, the image recognition and the quality matching are combined, so that the recovery link is greatly simplified, the labor of a person is simplified, the recovery process is standardized, the recovery time of surgical instruments is saved, the working efficiency is improved, the contact between the person and the polluted surgical instruments is reduced, the pollution is reduced, the safety is greatly improved, and the recovery accuracy of the surgical instruments is improved.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a schematic diagram of the system of the present invention;
FIG. 3 is a standard image of a portion of a class of surgical instruments according to the present invention;
FIG. 4 is an example of an image of a surgical instrument to be retrieved according to the present invention;
FIG. 5 is an example of a block image generated in FIG. 4;
FIG. 6 is an embodiment of block image segmentation;
where a is a region extending in the vertical direction and the horizontal direction, and numeral 1 is a schematic diagram showing a region extending in the vertical direction and a region extending in the horizontal direction;
FIG. 7 is a schematic diagram of calculating block image feature vectors using SIFT in the present invention;
fig. 8 is a schematic diagram illustrating a distance from a point to a planar object in a process of calculating a feature vector by another method in the present invention.
Detailed Description
The invention discloses a method and a system for automatically identifying and recovering surgical instruments, which are further explained and explained with reference to the attached drawings.
As shown in fig. 2, an automatic identification and recovery system for surgical instruments comprises a computer unit, and an image acquisition device and a quality acquisition device connected with the computer unit, wherein the computer unit comprises a database unit, an image processing device, an image identification device and a quality matching device; the image recognition device comprises a block image generation part, a characteristic quantity calculation part, a separation hyperplane calculation part and a type discrimination part;
the database unit is used for storing all categories of surgical instruments in the medical disinfection package, standard image information, standard quality information and separation hyperplane of the surgical instruments, wherein the standard images of the surgical instruments in all categories are of the type of block images, and are shown in fig. 3. The surgical instruments include spoon, atrium retractor, nerve retractor, valve holding forceps, pedicel forceps, auricle forceps and aorta blocking forceps.
The image acquisition device is used for acquiring the image information of the surgical instruments to be recovered, and comprises a camera which is used for shooting the images of the surgical instruments in the medical disinfection bag, monitoring the change of the recovered objects in time, capturing the image change information and using the change information for the identification of the computer unit.
The mass acquisition device is used for acquiring the mass information of the surgical instrument to be recovered, and the acquisition of the mass can be realized by a high-precision electronic scale.
And the image processing device is used for receiving the image information input by the image acquisition device and coding and compressing the input image information.
The image recognition device is used for recognizing the type of the input image information; the block image generating part is used for dividing the image information output by the image processing device into a plurality of blocks of images, namely dividing the image information of the surgical instruments in the medical disinfection package to be recovered into a plurality of blocks of images, and inputting all the blocks of images into the characteristic quantity calculating part; a feature amount calculation section for calculating a feature vector from the block image; a separating hyperplane calculating part for calculating a separating hyperplane of each type according to the standard image of each type of the surgical instrument stored in the database; and a type judging section for judging the type of the block image based on the distance between the feature vector output from the feature amount calculating section and the separation hyperplane for each type.
The quality matching device is used for further verifying the matching result of the classified image information output by the image recognition device according to the quality, namely acquiring the quality information transmitted by the quality acquisition device and standard quality information stored in the database unit, and comparing the quality information with the standard quality information.
The computer unit further comprises a main storage unit, the main storage unit stores information in the database unit, and the main storage unit comprises a RAM and a ROM.
The system also comprises a portable terminal, wherein the portable terminal comprises a communication module used for communicating with the computer unit and an auxiliary storage unit, and the auxiliary storage unit stores the information in the database unit.
The invention combines image recognition and quality matching, greatly simplifies the recovery link, simplifies the labor of people, standardizes the recovery process, saves the recovery time of surgical instruments, improves the working efficiency, reduces the contact of personnel and polluted surgical instruments, reduces the pollution, greatly improves the safety and simultaneously improves the recovery accuracy of the surgical instruments.
As shown in fig. 1, an automatic identification and recovery method for surgical instruments includes the following steps:
s1, acquiring an original image: acquiring clear and complete original images of surgical instruments in the medical disinfection package to be recycled;
s2, performing primary processing on the original image: extracting, coding and compressing an original image, converting the image into a JPEG format according to the resolution and color of the image, and acquiring a primary processing image;
s3, performing image recognition on the primary processing image, and acquiring the category to which the corresponding surgical instrument belongs in the primary image: dividing the primary processing image of the surgical instrument into a plurality of images, acquiring standard images of all classes of the surgical instrument in a database unit, and acquiring a feature vector corresponding to the standard image of each class of the surgical instrument for each image; extracting the separation hyperplane of each surgical instrument type from the database unit, calculating the distance between each feature vector and the separation hyperplane of the standard image of the corresponding surgical instrument type, and preliminarily judging the type of the block image according to the distance so as to obtain the primary classification of the corresponding surgical instrument in the block image.
The separation hyperplane calculation method of each surgical instrument type in the database unit comprises the following steps: the standard image of each surgical instrument type is obtained from the database unit, the separating hyperplane of each type is calculated by the separating hyperplane calculating part in the image recognition device, and the separating hyperplane of each type is stored in the database unit.
The generation process of the block image comprises the following steps: and performing target marking and background marking on the medical instrument image to be recovered in the medical disinfection packet, and segmenting the image containing the target marking into a plurality of non-overlapping sub-regions according to the image characteristics, wherein each sub-region is a block image to be treated, and the image characteristics comprise image color, shape, gray scale, texture and pixels. The image of the surgical instrument is segmented into sub-regions that do not overlap each other and have respective characteristics, each region being a continuum of pixels, where the characteristics may be color, shape, grayscale, texture, etc. of the image. The image segmentation represents a physically meaningful set of connected regions according to the prior knowledge of the target and the background, namely, the target and the background in the image are marked and positioned, and then the target is separated from the background. Fig. 4 is an example of an image of a surgical instrument to be recovered, which is sequentially from left to right: spoon, atrium drag hook, nerve drag hook, hold lamella pincers, renal pedicle pincers, auricle pincers and aorta occlusion pincers. The block image generating unit divides the target image into vertical 5 blocks × horizontal 3 blocks, and fig. 5 is an example of a block image. The block image generator may divide the target image into 4 vertical blocks × 6 horizontal blocks, or may divide the target image at another ratio as shown in fig. 6, that is, the block image generator may divide the target image after adjusting the longer side of the target image to be equal to or less than a predetermined value.
S4, performing quality verification on the surgical instruments after primary classification: the corresponding surgical instruments in the primarily classified block images are captured through the capturing device, the quality information of the corresponding surgical instruments in the primarily classified block images is obtained through the quality acquisition device, the primarily classified results of the surgical instruments output from the image recognition device are obtained, the obtained quality information of the surgical instruments is compared with the quality of the surgical instruments of the type in the database unit for verification, if the quality error is smaller than a threshold value, the type is judged reasonably, the surgical instruments are placed at the corresponding types on the recovery platform through the capturing device, otherwise, the step S3 is returned, and the feature vectors are recalculated for the block images.
The calculation method of the feature vector in step S3 is: and carrying out image matching on the block images and all the standard images by using an SIFT algorithm, wherein the output SIFT feature vector is the feature vector to be solved. The following description is made of symbols or formulas appearing in the SIFT algorithm:
i (x, y) represents the original image; g (x, y, σ) represents a Gaussian filter, in which
L (x, y, σ) represents an image generated by convolving an original image with a Gaussian filter, that is, L (x, y, σ) represents an image generated by convolving an original image with a Gaussian filter
A series of sigmaiA series of L (x, y, σ) may be generatedi) Image, and apply L (x, y, σ)i) The image is referred to as a scale space representation of the original image; DOG denotes the difference of gaussians
As shown in fig. 7, the specific process of calculating the feature vector by using the SIFT algorithm is as follows:
s31, constructing a Gaussian difference pyramid and a scale space;
the characteristic quantity image is a detail image which is finally constructed for searching characteristic points, the searching of the characteristic points needs to search a spatial local minimum value, namely, a difference image between an upper layer and a lower layer is needed when the local minimum value points are searched on a certain layer, if the characteristic points of the a layer need to be searched, the a +2 layer difference image is needed, and then the 2 nd layer to the a +1 th layer are searched;
each difference image G (x, y, σ) needs to be generated by differentiating the images L (x, y, k σ) and L (x, y, σ) of the two scale spaces, and assuming that a is 3, the difference image that we need has a +2 of 5, which are G (x, y, σ), G (x, y, k σ), and G (x, y, k σ), respectively2σ)、G(x，y，k3σ) and G (x, y, k)4σ)；
Wherein G (x, y, k σ), G (x, y, k)2σ)、G(x，y，k3σ) these three images are the images we use to find local extrema points. Then we need 6 scale space images, a +3, to generate the above difference images, which are L (x, y, σ), L (x, y, k σ), respectively2σ)、L(x，y，k3σ)、L(x，y，k4σ) and L (x, y, k)5σ); therefore, for the scale space, a +3 layers of images are required together to construct a +2 layers of differential feature quantity images.
As shown in the above case, the layer 1 image of the group 1 of the entire scale space is already generated by blurring the original image, that is, the detail information is lost, and the original image is not useful at all. Based on this consideration, we first magnify the image by a factor of 2, so that the details of the original image are hidden. From the above one scenario analysis, we have known that I (x, y) is considered to have been σnWhen I (x, y) is enlarged by 2 times, I (x, y) is obtained as a blurred image of 0.5s(x, y), then can be seen as being 2 σ n1 blurred image. Then is composed ofs(x, y) Gaussian filter for generating group 1 layer 1 images
Can be expressed as
Where FirstLayer (x, y) represents a layer 1 image of group 1 in the entire scale space, Is(x, y) is the image enlarged by bilinear interpolation from I (x, y); general sigma0＝1.6，σn＝0.5。
S32, searching for feature points: finding DOG extreme points; and comparing each pixel point with the surrounding pixel points, when the pixel point is larger than or smaller than all the adjacent points, namely, the pixel point is an extreme point, and after edge influence is removed through a Harris angular point detection algorithm, calculating the amplitude and the amplitude of the gradient of the pixel point for the extreme point.
And S33, representing the position, the amplitude and the amplitude of the extreme point by a group of vectors, wherein the vectors are the feature vectors output by the SIFT algorithm.
The feature amount calculation unit may calculate the feature vector of the block image by another method. The feature vector includes image data; the feature quantity image data comprises vector data and image raster data, the space vector data has various geometric attributes and space-time attributes, the data is massive, the concrete attributes of one surface element do not need to be displayed, and the surface element can be completely abstracted into a simple mathematical problem, namely the vector data of a space reference system does not need to be considered.
As shown in fig. 8, the following calculation is adopted in the present scheme: first, a point element is a simple class constructed from coordinates, a surface element is represented by the same coordinate string from the beginning to the end, and the distance between the point element and the surface element, that is, the calculated maximum distance, is calculated. When the maximum distance is calculated, comparison is involved, the coordinate strings need to be traversed, and deletion and insertion operations are not needed, so that the coordinate strings are stored in an array mode.
To calculate the point-to-surface distance, the point-to-surface distance is classified into three categories: the center distance is the distance from a point to the geometric center of a surface; the farthest distance, i.e. the distance from the point to the farthest surface, and the closest distance, i.e. the distance from the point to the closest surface;
in the data structure of the point surface, the coordinate of the geometrical center can be obtained through calculation, and the maximum distance is certain at the inflection point of the boundary, so that the problem can be converted into the distance between the two coordinates. Two attributes are designed here: x and y coordinates; two construction methods; a static method calculates the distance between two coordinates.
The feature quantity calculating unit performs point-plane combination calculation of vector data and image grid data.
The single feature amount here includes a local feature amount, a global feature amount, and a neighboring feature amount; wherein the local characteristic quantity: the characteristics extracted from the local parts of the images can form a vector, and the similarity degree can be calculated between the two images by defining a distance or similarity measurement degree; overall characteristic quantity: the method refers to the overall properties of an image, common global features include color features, texture features and shape features, such as intensity histograms and the like, and since the common global features are pixel-level low-level visual features, the overall feature quantity has the characteristics of good invariance, simplicity in calculation, intuition in representation and the like, but the high feature dimension and the large calculation amount are the fatal weaknesses of the image. Furthermore, global feature descriptions are not applicable to image aliasing and occluded cases. The local features are features extracted from local regions of the image, and comprise edges, corners, lines, curves, regions with special attributes and the like; the adjacent feature quantity: an image belongs to a class if most of the N most similar (i.e., nearest neighbor in feature space) samples in feature space belong to that class. The selected neighbors are all objects that have been correctly classified. The method only determines the category of the sample image to be classified according to the category of the nearest sample image or sample images in the classification decision.
The block image processing apparatus includes a local feature amount and a global feature amount of a block image, and in a second feature amount space having a plurality of features of the block image as coordinate axes, the coordinate position of a feature amount vector formed by combining a plurality of local feature amounts and one or a plurality of arbitrary regions of the second feature amount space, and the global feature amount includes the number of block images obtained by counting for each region the block image having the feature amount vector belonging to the region. A feature amount calculation unit for calculating a local feature amount and an entire feature amount of the block image, the feature amount being calculated from a vector combination of a pixel value of the block image and the feature amount; the feature amount used in combination is not limited to only the pixel value. A new feature quantity representing a complicated feature can be constituted by a combination of the respective feature quantities which are different from each other. The new feature amounts are used, for example, as feature amount vectors (a1, a2, a 3.) in a feature amount space having 1, 2, 3, 4 features as coordinate axes. The feature quantity space is divided into N regions in advance. Thus, in each region, the distance between arbitrary vectors in the region becomes small, and the local feature amounts become similar. Therefore, it is easy to extract the commonality between the plurality of local feature amounts. Since each block image belongs to any one of the N regions, it is determined to which region each of all block images included in all images belongs, the number of block images belonging to an arbitrary region Zi (1 ≦ i ≦ q) is counted, and the count value for each region is used as the overall feature amount. In this way, by using a combination of the respective feature amounts, it is possible to simultaneously perform determination by adding edge information which is a completely different type of feature amount in addition to the above-described combination of colors. For example, in the case of identifying a forceps, it is necessary to capture the forceps and a thin edge as features. The number of patch images of "scale and fine edge", that is, the number of patch images similar to the forceps, is obtained as the feature amount by combining the scale component, which is the feature amount for identifying "edge", and the edge for identifying "fine edge". Fig. 5 may also be considered as a feature space concentrated for identifying, for example, the "forceps" shown in fig. 4.
The separation hyperplane calculating unit has a function of inputting image feature values for each category and calculating a separation hyperplane for each category. The separated hyperplane calculating unit calculates a separated hyperplane using a dedicated library of a linear SVM (support vector machine) which is widely used as a learning algorithm. In the following, the separating superstraight lines of the feature amount planes in the two-dimensional image feature amounts a1, a2, A3, a4, B1, B2, B3, and B4 are explained in view of easy understanding. As shown in fig. 4, the training data of the correct answer and the training data of the incorrect answer are plotted with the vertical axis representing the image feature amount a and the horizontal axis representing the image feature amount 1. In this case, the separating hyperplane calculating unit learns a separating hyperplane, which is a straight line separating the training data of the correct answer and the training data of the incorrect answer, by using the linear SVM. The learning results are recorded as separate hyperplane data. In addition, in the case of the p-dimension (p > 2: p is an integer), the feature quantity plane is a feature quantity space, and the separation hyperplane is a separation hyperplane. Therefore, the separation hyperplane is a concept including the separation hyperplane, and the feature space is a concept including the feature plane.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (8)
1. An automatic identification and recovery method for surgical instruments is characterized by comprising the following steps:
s1, acquiring an original image: acquiring clear and complete original images of surgical instruments in the medical disinfection package to be recycled;
s2, performing primary processing on the original image: extracting, coding and compressing an original image, converting the image into a JPEG format according to the resolution and color of the image, and acquiring a primary processing image;
s3, performing image recognition on the primary processing image, and acquiring the category to which the corresponding surgical instrument belongs in the primary image: dividing the primary processing image of the surgical instrument into a plurality of images, acquiring standard images of all classes of the surgical instrument in a database unit, and acquiring a feature vector corresponding to the standard image of each class of the surgical instrument for each image; extracting a separation hyperplane of each surgical instrument type from a database unit, calculating the distance between each feature vector and the separation hyperplane of a standard image of the corresponding surgical instrument type, preliminarily judging the type of a block image according to the distance, and further obtaining the primary classification of the corresponding surgical instrument in the block image;
s4, performing quality verification on the surgical instruments after primary classification: the corresponding surgical instruments in the primarily classified block images are captured through the capturing device, the quality information of the corresponding surgical instruments in the primarily classified block images is obtained through the quality acquisition device, the primarily classified results of the surgical instruments output from the image recognition device are obtained, the obtained quality information of the surgical instruments is compared with the quality of the surgical instruments of the type in the database unit for verification, if the quality error is smaller than a threshold value, the type is judged reasonably, the surgical instruments are placed at the corresponding types on the recovery platform through the capturing device, otherwise, the step S3 is returned, and the feature vectors are recalculated for the block images.
2. The automatic identification and recovery method for surgical instruments according to claim 1, wherein: the generation process of the block image in S3 is as follows: and carrying out target marking and background marking on the images of the surgical instruments in the medical disinfection package to be recovered, and segmenting the images containing the target marks into a plurality of non-overlapping sub-regions according to image characteristics, wherein each sub-region is a block image to be detected and processed.
3. The automatic identification and recovery method for surgical instruments according to claim 2, wherein: the image characteristics include image color, shape, grayscale, texture, and pixels.
4. The automatic identification and recovery method for surgical instruments according to claim 1, wherein: the calculation method of the feature vector in the step S3 is as follows: and carrying out image matching on the block images and all the standard images by using an SIFT algorithm, wherein the output SIFT feature vector is the feature vector to be solved.
5. The automatic identification and recovery method for surgical instruments according to claim 1, wherein: the method for calculating the separation hyperplane of each surgical instrument type in the database unit in the step S3 includes: the standard image of each surgical instrument type is obtained from the database unit, the separating hyperplane of each type is calculated by the separating hyperplane calculating part in the image recognition device, and the separating hyperplane of each type is stored in the database unit.
6. An automatic surgical instrument identification and recovery system for realizing the automatic surgical instrument identification and recovery method of any one of claims 1 to 5, wherein: the system comprises a computer unit, an image acquisition device, a quality acquisition device and a recovery platform, wherein the image acquisition device, the quality acquisition device and the recovery platform are connected with the computer unit;
the image acquisition device comprises a camera for shooting images of surgical instruments in the medical disinfection bag, monitoring the change of a recovered object in time, capturing image change information and using the change information for computer unit identification;
the quality acquisition device is used for acquiring the quality information of the surgical instruments in the medical disinfection bag to be recovered;
the recovery platform is used for classifying and placing the recovered surgical instruments;
the database unit is used for storing all categories of surgical instruments in the medical disinfection package, standard images, standard quality information and separation hyperplane of the surgical instruments, wherein the standard images of the surgical instruments of all the categories are of block images;
the image processing device is used for receiving the image information input by the image acquisition device and coding and compressing the input image information;
the image recognition device comprises a block image generation part, a characteristic quantity calculation part, a separation hyperplane calculation part and a type discrimination part; wherein, the image generation department of the block, is used for dividing the image information outputted by the image processing unit into several block images, and input all block images into the characteristic quantity calculating department; the feature quantity calculating part is used for calculating a feature vector according to the block image; the separation hyperplane calculating part is used for calculating a separation hyperplane of each type according to the standard image of each type of the surgical instrument stored in the database; the type judging part is used for judging the type of the block image according to the distance between the feature vector and the separation hyperplane of each type;
and the quality matching device is used for acquiring the quality information transmitted by the quality acquisition device and the standard quality information stored in the database unit and comparing the quality information with the standard quality information.
7. The system of claim 6, wherein: the computer unit further comprises a main storage unit, the main storage unit stores information in the database unit, and the main storage unit comprises a RAM and a ROM.
8. The system of claim 6, wherein: the portable terminal comprises a communication module used for communicating with the computer unit and an auxiliary storage unit, and the auxiliary storage unit stores information in the database unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010352809.6A CN111553422A (en) | 2020-04-28 | 2020-04-28 | Automatic identification and recovery method and system for surgical instruments |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010352809.6A CN111553422A (en) | 2020-04-28 | 2020-04-28 | Automatic identification and recovery method and system for surgical instruments |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111553422A true CN111553422A (en) | 2020-08-18 |
Family
ID=72007744
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010352809.6A Pending CN111553422A (en) | 2020-04-28 | 2020-04-28 | Automatic identification and recovery method and system for surgical instruments |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111553422A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112712016A (en) * | 2020-12-29 | 2021-04-27 | 上海微创医疗机器人（集团）股份有限公司 | Surgical instrument identification method, identification platform and medical robot system |
| CN113380393A (en) * | 2021-06-28 | 2021-09-10 | 南通市第一人民医院 | Method and system for supervising medical instruments for gynecology |
| CN113952048A (en) * | 2021-11-24 | 2022-01-21 | 燕山大学 | Surgical instrument identification device and identification method thereof |
| CN114812697A (en) * | 2022-06-28 | 2022-07-29 | 张家港市欧凯医疗器械有限公司 | Medical tube health detection method and system |
| CN117095803A (en) * | 2023-10-19 | 2023-11-21 | 邦士医疗科技股份有限公司 | Informationized system for rapid inventory of surgical instruments |
| TWI843690B (en) * | 2023-11-23 | 2024-05-21 | 南臺學校財團法人南臺科技大學 | Device for recovery of medical needles |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101650783A (en) * | 2008-08-13 | 2010-02-17 | 株式会社Ntt都科摩 | Image identification method and imaging apparatus |
| CN202277385U (en) * | 2011-10-14 | 2012-06-20 | 上海理工大学 | Automatic identifying and counting system of surgical instrument |
| CN102799854A (en) * | 2011-05-23 | 2012-11-28 | 株式会社摩如富 | Image identification device and image identification method |
| WO2016059382A1 (en) * | 2014-10-15 | 2016-04-21 | Spa Track Medical Limited | Surgical instrument rfid tag reading apparatus and method, and surgical instrument tracking system |
| CN108294831A (en) * | 2018-03-20 | 2018-07-20 | 中南大学湘雅二医院 | A kind of automatic classification of surgical instrument material and recovery system |
| CN109147170A (en) * | 2018-07-26 | 2019-01-04 | 上海凯景信息技术有限公司 | A kind of self-service cabinet based on image recognition |
| CN110276305A (en) * | 2019-06-25 | 2019-09-24 | 广州众聚智能科技有限公司 | A kind of dynamic commodity recognition methods |
-
2020
- 2020-04-28 CN CN202010352809.6A patent/CN111553422A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101650783A (en) * | 2008-08-13 | 2010-02-17 | 株式会社Ntt都科摩 | Image identification method and imaging apparatus |
| CN102799854A (en) * | 2011-05-23 | 2012-11-28 | 株式会社摩如富 | Image identification device and image identification method |
| CN202277385U (en) * | 2011-10-14 | 2012-06-20 | 上海理工大学 | Automatic identifying and counting system of surgical instrument |
| WO2016059382A1 (en) * | 2014-10-15 | 2016-04-21 | Spa Track Medical Limited | Surgical instrument rfid tag reading apparatus and method, and surgical instrument tracking system |
| CN108294831A (en) * | 2018-03-20 | 2018-07-20 | 中南大学湘雅二医院 | A kind of automatic classification of surgical instrument material and recovery system |
| CN109147170A (en) * | 2018-07-26 | 2019-01-04 | 上海凯景信息技术有限公司 | A kind of self-service cabinet based on image recognition |
| CN110276305A (en) * | 2019-06-25 | 2019-09-24 | 广州众聚智能科技有限公司 | A kind of dynamic commodity recognition methods |
Non-Patent Citations (2)
| Title |
|---|
| RONNY: "SIFT定位算法关键步骤的说明", 《HTTPS://WWW.BBSMAX.COM/A/Q4ZVWX82JK/》 * |
| 侯宏花: "《数字图像处理与分析》", 30 September 2011, 北京理工大学出版社 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112712016A (en) * | 2020-12-29 | 2021-04-27 | 上海微创医疗机器人（集团）股份有限公司 | Surgical instrument identification method, identification platform and medical robot system |
| CN112712016B (en) * | 2020-12-29 | 2024-01-26 | 上海微创医疗机器人（集团）股份有限公司 | Surgical instrument identification method, identification platform and medical robot system |
| CN113380393A (en) * | 2021-06-28 | 2021-09-10 | 南通市第一人民医院 | Method and system for supervising medical instruments for gynecology |
| CN113952048A (en) * | 2021-11-24 | 2022-01-21 | 燕山大学 | Surgical instrument identification device and identification method thereof |
| CN114812697A (en) * | 2022-06-28 | 2022-07-29 | 张家港市欧凯医疗器械有限公司 | Medical tube health detection method and system |
| CN117095803A (en) * | 2023-10-19 | 2023-11-21 | 邦士医疗科技股份有限公司 | Informationized system for rapid inventory of surgical instruments |
| CN117095803B (en) * | 2023-10-19 | 2023-12-15 | 邦士医疗科技股份有限公司 | Informationized system for rapid inventory of surgical instruments |
| TWI843690B (en) * | 2023-11-23 | 2024-05-21 | 南臺學校財團法人南臺科技大學 | Device for recovery of medical needles |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11681418B2 (en) | Multi-sample whole slide image processing in digital pathology via multi-resolution registration and machine learning | |
| CN111553422A (en) | Automatic identification and recovery method and system for surgical instruments | |
| EP3382644A1 (en) | Method for 3d modelling based on structure from motion processing of sparse 2d images | |
| CN112686812B (en) | Bank card inclination correction detection method and device, readable storage medium and terminal | |
| CN109583483B (en) | Target detection method and system based on convolutional neural network | |
| US20230099984A1 (en) | System and Method for Multimedia Analytic Processing and Display | |
| US20070058856A1 (en) | Character recoginition in video data | |
| CN112633297B (en) | Target object identification method and device, storage medium and electronic device | |
| CN113688846B (en) | Object size recognition method, readable storage medium, and object size recognition system | |
| CN114549603B (en) | Method, system, equipment and medium for converting labeling coordinate of cytopathology image | |
| JP2010134957A (en) | Pattern recognition method | |
| CN110414571A (en) | A kind of website based on Fusion Features reports an error screenshot classification method | |
| CN113095445B (en) | Target identification method and device | |
| CN111461101A (en) | Method, device and equipment for identifying work clothes mark and storage medium | |
| CN112257711B (en) | Method for detecting damage fault of railway wagon floor | |
| Shihavuddin et al. | Automated classification and thematic mapping of bacterial mats in the north sea | |
| CN111127556A (en) | Target object identification and pose estimation method and device based on 3D vision | |
| CN111462310A (en) | Bolt defect space positioning method based on multi-view geometry | |
| CN116740135A (en) | Infrared dim target tracking method and device, electronic equipment and storage medium | |
| CN110288040B (en) | Image similarity judging method and device based on topology verification | |
| CN111612045B (en) | Universal method for acquiring target detection data set | |
| CN112418262A (en) | Vehicle re-identification method, client and system | |
| CN117218672A (en) | Deep learning-based medical records text recognition method and system | |
| US20030210818A1 (en) | Knowledge-based hierarchical method for detecting regions of interest | |
| CN111325194B (en) | Character recognition method, device and equipment and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200818 |