A feather intelligent recognition method based on industrial camera

Abstract

The present invention provides a feather intelligent recognition method based on an industrial camera, comprising: step S1, obtaining a test image of a test feather sample; step S2, obtaining a standard image of a standard feather sample, dividing the test image and the standard image layer by layer to calculate the grayscale value of the pixel points in each layer of the image; step S3, selecting a number of central pixels in each layer of the image, taking each central pixel as the center of the circle, taking a preset pixel as the radius to obtain a plurality of circles, and taking a number of pixels on each circle, and obtaining an adaptive threshold value according to the grayscale value of each pixel point; step S4, obtaining the brightness value of each central pixel point and the pixel point, screening each central pixel point according to each brightness value and the adaptive threshold value to obtain matching feature points; step S5, obtaining the mean square error as the intelligent recognition result according to the grayscale value of each feature point of the standard feather sample and the test feather sample. The beneficial effect is that the present invention can reduce the cost of detection measures and improve detection efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of feather texture detection, and in particular to a feather intelligent recognition method based on an industrial camera. Background Art

[0002] Feathers are an important economic trait of poultry and also an important appearance feature. Feathers play an important role in flying, keeping warm, swimming, protecting, and regulating temperature. Feathers have a complex structure, and feather types and compositions vary greatly. In addition, feathers need to grow step by step for three generations before they can finally grow. The growth of feathers is closely related to the health of poultry, and the feathers at each stage are different. The feathers of poultry can be used to determine the growth stage of the poultry and whether it has reached the age of being marketed. Based on this, through literature and production experience, a database of normal feathers is established, and then the feathers to be tested are extracted and compared through industrial cameras, so that the health of the poultry can be determined, which greatly improves the efficiency of disease prevention and disease screening, reduces the level of early specialization in disease screening, and makes the disease detection process tend to be process-based and industrialized.

[0003] In order to increase the efficiency of poultry disease screening, patent application CN117809837A discloses a dual-band improved YOLOv7 poultry disease detection method and system based on light perception, the method comprising: obtaining a poultry disease detection dataset including a plurality of RGB-IR image pairs of poultry, and the RGB-IR image pairs are annotated; using the poultry disease detection dataset to train a poultry disease detection model built based on the YOLOv7 network architecture; the model uses two YOLOv7 architectures with the same structure to form a dual-branch network, and replaces the i-th ELAN module in the backbone network in the two branches with the same feature interaction module to enable interaction between different modal features; the RG of the poultry to be tested is used to train the poultry disease detection model based on the YOLOv7 network architecture; the model uses two YOLOv7 architectures with the same structure to form a dual-branch network, and replaces the i-th ELAN module in the backbone network in the two branches with the same feature interaction module to enable interaction between different modal features; The B-IR image is input into the trained model to identify whether the poultry is sick and the type of disease; the algorithm of this model is relatively complex, the amount of calculation is large, and it is easily affected by external environmental factors; and the patent application CN114610902A discloses a poultry disease diagnosis system based on a knowledge graph, which includes a knowledge graph module, a user submission module, and a disease diagnosis module; wherein the signal input and output end of the knowledge graph module is connected to the first signal output and input end of the disease diagnosis module, and the second signal output and input end of the disease diagnosis module is connected to the signal input and output end of the user submission module; the knowledge graph module is used to construct a poultry disease knowledge graph; the system processing flow is relatively long, and the efficiency of detection is not greatly improved.

[0004] In the above detection method or system, the detection equipment is too complex and the large number of devices used in series leads to high cost of such detection measures and the inability to popularize them on a large scale. In addition, the processing process is cumbersome, which easily leads to a decrease in detection efficiency. Summary of the invention

[0005] The technical problem to be solved by the present invention is to reduce the cost of detection measures and improve detection efficiency. In order to overcome the defects of the above-mentioned prior art (or related technology), the present invention provides a feather intelligent recognition method based on an industrial camera.

[0006] The present invention provides a feather intelligent recognition method based on an industrial camera, wherein a conveyor belt is pre-configured for continuously conveying test feather samples, and an industrial camera is mounted on the conveyor belt. The feather intelligent recognition method comprises the following steps:

[0007] Step S1, controlling the start of the conveyor belt to transfer the test feather sample to the shooting position of the industrial camera, and controlling the industrial camera to shoot a test image of the test feather sample;

[0008] Step S2, obtaining a standard image of a standard feather sample, dividing the test image and the standard image layer by layer using a B-Spline image pyramid method, and calculating the grayscale value of pixels in each layer of the image;

[0009] Step S3, selecting a number of central pixel points in each layer of the image, taking each central pixel point as the center of the circle and a preset pixel as the radius to obtain a plurality of circles, and taking a number of the pixel points on each of the circles, and obtaining an adaptive threshold value according to the grayscale value of each of the pixel points;

[0010] Step S4, obtaining each of the central pixels and the brightness value of each of the pixels, and screening each of the central pixels according to the brightness value and the adaptive threshold to obtain matching feature points;

[0011] Step S5, calculating a mean square error according to the grayscale value of each feature point of the standard feather sample and the grayscale value of each feature point of the test feather sample as an intelligent recognition result.

[0012] Compared with the prior art, the feather intelligent recognition method based on industrial camera in this application has the following advantages:

[0013] The hardware devices used in this application include a conveyor belt and an industrial camera. The control logic of the conveyor belt and the industrial camera is simple and easy to use. Compared with detection equipment with too high complexity, it can reduce costs and simplify operations. Grayscale value calculation, adaptive threshold calculation, feature point screening and mean square error calculation of pixel points of standard images and test images are performed, which simplifies the processing process and improves detection efficiency. The smaller the mean square error, the higher the similarity between the standard image and the test image, and the larger the mean square error, the lower the similarity between the standard image and the test image. Using the mean square error to characterize the feather intelligent recognition result is more accurate.

[0014] In a possible implementation manner, before executing step S1, the method includes:

[0015] The zero point position of the conveying point is corrected so that the conveyor belt is in a horizontal state.

[0016] In a possible implementation, in step S1, controlling the start of the conveyor belt includes:

[0017] The conveyor belt is controlled to run at a preset constant speed, and when the test feather sample is transmitted to the shooting position of the industrial camera, the conveyor belt is controlled to stop running for 3 seconds, and the conveyor belt is controlled to start after 3 seconds.

[0018] In a possible implementation manner, an LED device is mounted next to the industrial camera, and in step S1, when the test feather sample is transmitted to the shooting position of the industrial camera, the step further includes:

[0019] The LED device is controlled to light up for supplementary lighting.

[0020] Compared with the prior art, the above solution can reduce the interference caused by the environment and improve the quality of the captured standard images and test images.

[0021] In a possible implementation, in step S2, the grayscale value of each pixel is obtained by the following calculation formula:

[0022] ,

[0023] in,

[0024] represents the grayscale value;

[0025] Represents the horizontal coordinate of the pixel point in the image;

[0026] Represents the vertical coordinate of the pixel point in the image;

[0027] Indicates the B-Spline order;

[0028] Indicates the number of layers of the B-Spline image pyramid;

[0029] Represents the spline coefficient array;

[0030] represents the scaling function.

[0031] In a possible implementation, in step S2, the scaling function is obtained by the following calculation formula:

[0032] ,

[0033] in,

[0034] represents the scaling function;

[0035] Represents the horizontal coordinate of the pixel point in the image;

[0036] Represents the vertical coordinate of the pixel point in the image;

[0037] Indicates the B-Spline order;

[0038] Indicates the number of layers of the B-Spline image pyramid;

[0039] express Impulse response of order B-Spline;

[0040] Represents a two-dimensional B-Spline basis function.

[0041] In a possible implementation, in step S2, the spline coefficient array is obtained by the following calculation formula:

[0042] ;

[0043] ;

[0044] ;

[0045] in,

[0046] Represents the spline coefficients of the first layer image;

[0047] Represents the spline coefficients of the subsequent layer image;

[0048] Indicates the B-Spline order;

[0049] Indicates the number of layers of the B-Spline image pyramid.

[0050] In a possible implementation, in step S3, a circle is obtained with the central pixel point as the center and 3 pixels as the radius, and 16 pixels on the circle are taken, and the adaptive threshold is obtained by the following calculation formula:

[0051] ;

[0052] in,

[0053] represents the adaptive threshold;

[0054] represents the preset adaptive parameters;

[0055] Represents the gray value of the central pixel;

[0056] represents the maximum value among the gray values;

[0057] represents the minimum value among the gray values;

[0058] Represents the average value of the remaining grayscale values ​​excluding the maximum and minimum values.

[0059] In a possible implementation, in step S4, when screening the matched feature points, for each of the central pixel points, the brightness value of the central pixel point is added to the adaptive threshold to obtain an upper limit value, and the brightness value of the central pixel point is subtracted from the adaptive threshold to obtain a lower limit value, and then it is determined whether there are 12 consecutive pixels on the circle of the central pixel point whose brightness values ​​are between the upper limit value and the lower limit value:

[0060] If yes, the central pixel is used as the feature point;

[0061] If not, delete the central pixel.

[0062] In a possible implementation, in step S5, the mean square error is obtained by the following calculation formula:

[0063] ;

[0064] in,

[0065] MSE represents the mean square error;

[0066] n represents the number of each gray value;

[0067] Indicates the number of the tested feather samples said grayscale value;

[0068] Indicates the standard feather sample The grayscale value. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0070] First, those skilled in the art should understand that these implementations are only used to explain the technical principles of the embodiments of the present application, and are not intended to limit the protection scope of the embodiments of the present application. Those skilled in the art can make adjustments to them as needed to adapt to specific application scenarios.

[0071] The present application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0072] See also Figure 1 The embodiment of the present application discloses a feather intelligent recognition method based on an industrial camera. A conveyor belt is pre-configured for continuously conveying test feather samples. An industrial camera is mounted on the conveyor belt. The feather intelligent recognition method includes the following steps:

[0073] Step S1, controlling the start of a conveyor belt to transport a test feather sample to a shooting position of an industrial camera, and controlling the industrial camera to shoot a test image of the test feather sample;

[0074] Step S2, obtaining a standard image of a standard feather sample, dividing the test image and the standard image layer by layer using a B-Spline image pyramid method, and calculating the grayscale value of the pixel points in each layer of the image;

[0075] Step S3, select a number of central pixels in each layer of the image, take each central pixel as the center of the circle and a preset pixel as the radius to obtain a plurality of circles, and take a number of pixels on each circle to obtain an adaptive threshold according to the gray value of each pixel;

[0076] Step S4, obtaining each central pixel and the brightness value of each pixel, and screening each central pixel according to each brightness value and an adaptive threshold to obtain matching feature points;

[0077] Step S5, calculating the mean square error according to the grayscale value of each feature point of the standard feather sample and the grayscale value of each feature point of the test feather sample as the intelligent recognition result.

[0078] Specifically, in this embodiment, the feather intelligent identification method of the present invention is simple and feasible, the principle is simple, the operation is basically automated, and the detection process does not involve the process of changing the properties of the test feather samples, and will not cause irreversible damage to the poultry. Therefore, the generation of detection waste can be saved, thereby reducing costs; and the test images captured during each detection process can be retained, which can provide a reference for the improvement of subsequent processes.

[0079] Specifically, in this embodiment, the standard feather samples (healthy poultry feathers) are placed on a conveyor belt, and the computer analysis system runs a collection program for the standard feather samples. The feathers are led by the conveyor belt into the field of view of the industrial camera, and the industrial camera takes pictures. The industrial camera transmits the obtained standard images to the computer in real time, and recognizes them through the input established model and according to the given algorithm. The collection of different standard feather samples is repeated at least 10 times, and the standard images of this type of feathers can be preliminarily sampled. Similarly, if disease screening is performed, samples of this type of diseased feathers should be collected no less than 10 times.

[0080] Preferably, the present invention firstly adopts the B-Spline image pyramid method to divide the image layer by layer, and then sets the adaptive threshold by calculating the gray value of the pixel points in the image, so that the adaptive threshold is automatically adjusted as the environment changes, thereby realizing the effective extraction of image feature points.

[0081] In a preferred embodiment of the present invention, before executing step S1, the following steps are included:

[0082] Correct the zero position of the transfer point so that the conveyor belt is in a horizontal state.

[0083] Specifically, in this embodiment, the conveyor belt needs to be calibrated before use, and the zero point position of the conveyor belt needs to be recalibrated to ensure that the conveyor belt is in a horizontal state before use. The specific operating steps are: use a spirit level on the conveyor belt, and use the spirit level to adjust the conveyor belt to a horizontal state to ensure that the industrial camera can obtain a good shooting effect and ensure the stability of the conveyor belt operation.

[0084] In a preferred embodiment of the present invention, in step S1, when controlling to start the conveyor belt, it includes:

[0085] The conveyor belt is controlled to run at a preset constant speed, and when the test feather sample is transmitted to the shooting position of the industrial camera, the conveyor belt is controlled to stop running for 3 seconds, and the conveyor belt is controlled to start after 3 seconds.

[0086] Specifically, in this embodiment, the conveyor belt moves at a constant speed. When the conveyor belt delivers the test feather sample to the shooting position of the industrial camera, the conveyor belt is controlled to stop the test feather sample at the shooting position for 3 seconds through the feedback signal of the bottom sensor. After 3 seconds, the conveyor belt continues to move at a constant speed until the next test feather sample continues to reach the shooting position, and the shooting is repeated in a cycle.

[0087] In a preferred embodiment of the present invention, an LED device is mounted on the side of the industrial camera, and in step S1, when the test feather sample is transmitted to the shooting position of the industrial camera, it also includes:

[0088] Control the lighting of the LED device for fill light.

[0089] In a preferred embodiment of the present invention, in step S2, the grayscale value of each pixel is obtained by the following calculation formula:

[0090] ,

[0091] in,

[0092] Represents grayscale value;

[0093] Represents the horizontal coordinate of the pixel in the image;

[0094] Indicates the vertical coordinate of the pixel in the image;

[0095] Indicates the B-Spline order;

[0096] Indicates the number of layers of the B-Spline image pyramid;

[0097] Represents the spline coefficient array;

[0098] represents the scaling function.

[0099] In a preferred embodiment of the present invention, in step S2, the scaling function is obtained by the following calculation formula:

[0100] ,

[0101] in,

[0102] represents the scaling function;

[0103] Represents the horizontal coordinate of the pixel in the image;

[0104] Indicates the vertical coordinate of the pixel in the image;

[0105] Indicates the B-Spline order;

[0106] Indicates the number of layers of the B-Spline image pyramid;

[0107] express Impulse response of order B-Spline;

[0108] Represents a two-dimensional B-Spline basis function.

[0109] In a preferred embodiment of the present invention, in step S2, the spline coefficient array is obtained by the following calculation formula:

[0110] ;

[0111] ;

[0112] ;

[0113] in,

[0114] Represents the spline coefficients of the first layer image;

[0115] Represents the spline coefficients of the subsequent layer image;

[0116] Indicates the B-Spline order;

[0117] Indicates the number of layers of the B-Spline image pyramid.

[0118] In a preferred embodiment of the present invention, in step S3, a circle is obtained with the central pixel point as the center and 3 pixels as the radius, and 16 pixels on the circle are taken, and the adaptive threshold is obtained by the following calculation formula:

[0119] ;

[0120] in,

[0121] represents the adaptive threshold;

[0122] represents the preset adaptive parameters;

[0123] Indicates the gray value of the central pixel;

[0124] Represents the maximum value among all grayscale values;

[0125] Represents the minimum value among all gray values;

[0126] Represents the average value of all grayscale values ​​except the maximum and minimum values.

[0127] In a preferred embodiment of the present invention, in step S4, when screening the matching feature points, for each central pixel, the brightness value of the central pixel is added to the adaptive threshold to obtain an upper limit value, and the brightness value of the central pixel is subtracted from the adaptive threshold to obtain a lower limit value, and then it is determined whether there are 12 consecutive pixels on the circle of the central pixel whose brightness values ​​are between the upper limit value and the lower limit value:

[0128] If yes, the central pixel is taken as the feature point;

[0129] If not, delete the center pixel.

[0130] In a preferred embodiment of the present invention, in step S5, the mean square error is obtained by the following calculation formula:

[0131] ;

[0132] in,

[0133] MSE stands for mean square error;

[0134] n represents the number of gray values;

[0135] Indicates the number of feather samples tested. Gray value;

[0136] Indicates the standard feather sample Gray value.

[0137] Specifically, in this embodiment, the key points can be further screened by calculating the characteristic point response value, and the characteristic point response value is calculated by the following calculation formula:

[0138] ;

[0139] in,

[0140] S represents the characteristic point response value;

[0141] The circle representing the feature point The brightness value of each pixel, and ≤16;

[0142] Indicates the brightness value of the feature point;

[0143] represents the adaptive threshold;

[0144] After the feature point response values ​​of all feature points are calculated, the sizes of all feature point response values ​​are compared, the feature point with the largest feature point response value is retained, and the remaining feature points are deleted.

[0145] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "in the present embodiment", "specific example", or "some examples" etc. means that the specific features, mechanisms, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, mechanisms, materials or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0146] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A feather intelligent recognition method based on industrial camera, characterized in that: A conveyor belt is pre-configured for continuously conveying the test feather samples, and an industrial camera is mounted on the conveyor belt. The feather intelligent identification method includes the following steps: Step S1, controlling the start of the conveyor belt to transfer the test feather sample to the shooting position of the industrial camera, and controlling the industrial camera to shoot a test image of the test feather sample; Step S2, obtaining a standard image of a standard feather sample, dividing the test image and the standard image layer by layer using a B-Spline image pyramid method, and calculating the grayscale value of pixels in each layer of the image; Step S3, selecting a number of central pixel points in each layer of the image, taking each central pixel point as the center of the circle and a preset pixel as the radius to obtain a plurality of circles, and taking a number of the pixel points on each of the circles to obtain an adaptive threshold value according to the grayscale value of each of the pixel points; Step S4, obtaining the brightness value of each of the central pixels and each of the pixels, and screening each of the central pixels according to the brightness value and the adaptive threshold to obtain matching feature points; Step S5, calculating a mean square error according to the grayscale value of each feature point of the standard feather sample and the grayscale value of each feature point of the test feather sample as an intelligent recognition result; In step S3, a circle is obtained with the central pixel point as the center and 3 pixels as the radius, and 16 pixels on the circle are taken, and the adaptive threshold is obtained by the following calculation formula: ； in, represents the adaptive threshold; represents the preset adaptive parameters; Represents the gray value of the central pixel; represents the maximum value among the gray values; represents the minimum value among the gray values; Represents the average value of the remaining grayscale values ​​excluding the maximum and minimum values.

2. The feather intelligent identification method according to claim 1, characterized in that: Before executing step S1, the method includes: The zero point position of the conveyor belt is corrected so that the conveyor belt is in a horizontal state.

3. The feather intelligent identification method according to claim 1, characterized in that: In the step S1, when controlling to start the conveyor belt, it includes: The conveyor belt is controlled to run at a preset constant speed, and when the test feather sample is transmitted to the shooting position of the industrial camera, the conveyor belt is controlled to stop running for 3 seconds, and the conveyor belt is controlled to start after 3 seconds.

4. The feather intelligent identification method according to claim 1, characterized in that: An LED device is mounted on the side of the industrial camera, and in step S1, when the test feather sample is transmitted to the shooting position of the industrial camera, it also includes: The LED device is controlled to light up for supplementary lighting.

5. The feather intelligent identification method according to claim 1, characterized in that: In step S2, the grayscale value of each pixel is obtained by the following calculation formula: ， in, represents the grayscale value; Represents the horizontal coordinate of the pixel point in the image; Represents the vertical coordinate of the pixel point in the image; Indicates the B-Spline order; Indicates the number of layers of the B-Spline image pyramid; Represents the spline coefficient array; represents the scaling function.

6. The feather intelligent identification method according to claim 5, characterized in that: In step S2, the scaling function is obtained by the following calculation formula: ， in, represents the scaling function; Represents the horizontal coordinate of the pixel point in the image; Represents the vertical coordinate of the pixel point in the image; Indicates the B-Spline order; Indicates the number of layers of the B-Spline image pyramid; express Impulse response of order B-Spline; Represents a two-dimensional B-Spline basis function.

7. The feather intelligent identification method according to claim 5, characterized in that: In step S2, the spline coefficient array is obtained by the following calculation formula: ； ； ； in, Represents the spline coefficients of the first layer image; Represents the spline coefficients of the subsequent layer image; Indicates the B-Spline order; Indicates the number of layers of the B-Spline image pyramid.

8. The feather intelligent identification method according to claim 1, characterized in that: In step S4, when screening the matched feature points, for each central pixel, the brightness value of the central pixel is added to the adaptive threshold to obtain an upper limit value, and the brightness value of the central pixel is subtracted from the adaptive threshold to obtain a lower limit value, and then it is determined whether there are 12 consecutive pixels on the circle of the central pixel whose brightness values ​​are between the upper limit value and the lower limit value: If yes, the central pixel is used as the feature point; If not, delete the central pixel.

9. The feather intelligent identification method according to claim 1, characterized in that: In step S5, the mean square error is obtained by the following calculation formula: ； in, MSE represents the mean square error; n represents the number of each gray value; Indicates the number of the tested feather samples said grayscale value; Indicates the standard feather sample The grayscale value.

Images