A method for screening attractants based on inert element labeling and computer vision

Abstract

The present invention relates to the field of animal feed, and in particular to an attractant screening method based on an inert element labeling method and computer vision, comprising: placing an attractant-containing feed into a corresponding aquaculture water body, setting an above-water camera outside the aquaculture water body, and setting an underwater camera inside the aquaculture water body; preliminarily screening the feed based on the inert element content in the test animal to obtain an attractant preference value; obtaining the motion trajectory, feeding intensity, response speed, and feeding time of the test animal based on the image acquired by the camera, and obtaining the feeding level of the corresponding feed based on the motion trajectory, feeding intensity, response speed, and feeding time; and screening the attractant in combination with the attractant preference value and the feeding level. The prepared feed can be screened in the aquaculture water body based on the inert element content to obtain the attractant preference value, and then the feeding situation is recorded by a camera to obtain the feeding level to improve the accuracy of feed screening.

Description

Technical Field

[0001] The invention relates to the field of animal feed, and in particular to a method for screening attractants based on an inert element labeling method and computer vision. Background Art

[0002] The smell and taste of artificial compound feeds are quite different from the natural feeds of fish and shrimps, which often leads to low enthusiasm of fish and shrimps to eat artificial compound feeds, or even refusal to eat. Fish and shrimps have difficulty in feeding artificial feeds during the seedling stage, poor feeding during the growth period, insufficient energy intake, low growth performance, and poor disease resistance, which seriously affect the economic benefits of aquaculture. The most important factor affecting fish feeding is the chemical composition of the feed. To solve this problem, specific attractants need to be added when developing artificial compound feeds to promote feeding, improve growth performance and disease resistance.

[0003] Therefore, the screening of attractants for fish and shrimp feeds is extremely important. However, since fish and shrimp live in an aquatic environment, their feeding conditions are difficult to observe, which brings great difficulties to the study of fish and shrimp feeding and the screening of attractants. Summary of the invention

[0004] The purpose of the present invention is to provide a method for screening attractants based on inert element labeling and computer vision, aiming to improve the screening efficiency of aquatic animal feed attractants.

[0005] To achieve the above-mentioned purpose, the present invention provides an attractant screening method based on inert element labeling and computer vision, comprising placing a feed containing an attractant into a corresponding aquaculture water body, setting an above-water camera outside the aquaculture water body, and setting an underwater camera inside the aquaculture water body;

[0006] The feed was initially screened based on the inert element content in the test animals to obtain the attractant preference value;

[0007] The movement trajectory, feeding intensity, response speed and feeding time of the test animals are obtained based on the images obtained by the camera, and the feeding level of the corresponding feed is obtained based on the movement trajectory, feeding intensity, response speed and feeding time;

[0008] The attractants were screened based on the attractant preference value and feeding level.

[0009] The specific steps of preliminarily screening the feed based on the inert element content in the test animal to obtain the attractant preference value are:

[0010] Inert element markers are added when preparing various feeds containing attractants;

[0011] Throwing feed into the culture water with a specified number of test animals;

[0012] Recycling of residual feed after a specified period of time;

[0013] Recover the feces of experimental animals and pre-treat them;

[0014] Plasma emission chromatography was used to determine the inert element content of dried feces and generate the preference value of the test animals for the attractant.

[0015] Wherein, the inert element marker is any one of yttrium oxide, lanthanum oxide, samarium oxide, neodymium oxide and ytterbium oxide.

[0016] The specific method of recovering the feces of the test animals and pre-treating them is: using a scoop net to collect the feces with intact capsules, storing them in a refrigerator at -20°C, and then drying them at 70°C.

[0017] The specific steps of obtaining the motion trajectory of the test animal based on the image acquired by the camera are:

[0018] The background difference method is used to perform the initial segmentation of fish motion images;

[0019] The OTSU threshold segmentation algorithm is used to perform secondary segmentation on the initially segmented image to obtain a processed image;

[0020] The processed images were processed using the first-order moment to obtain the real-time position of the test animal;

[0021] Extract the outline and skeleton of the test animal, calculate the angle between the fish body and the image coordinate axis, and realize automatic detection of the fish movement direction according to the direction determination criteria;

[0022] The motion segmentation technology is combined with the CAMShift algorithm to track fish moving targets in video image sequences and extract the trajectory of the experimental animals' movement.

[0023] The specific steps of obtaining the feeding intensity of the test animal based on the image acquired by the camera are:

[0024] Preprocess the images acquired by the camera;

[0025] Extracting texture feature variable information of feeding images based on gray-level co-occurrence matrix;

[0026] Constructing a data set based on texture feature variable information;

[0027] The vector machine classifier is trained based on the data set to obtain an intensity recognition model;

[0028] The feeding intensity was obtained based on the intensity recognition model.

[0029] The specific method of setting an above-water camera outside the aquaculture water body and setting an underwater camera inside the aquaculture water body is:

[0030] The lens of the above-water camera is facing the center of the aquaculture water body, and the underwater camera is placed at a depth of 50CM from the water surface and at the edge of the feeding range.

[0031] Among them, the specific steps of obtaining the feeding level of the corresponding feed based on the motion trajectory, feeding intensity, response speed and feeding time are: generating an envelope diagram of the motion trajectory; obtaining the ratio of the envelope diagram and the feeding area to obtain a first judgment value; obtaining the ratio of the feeding intensity to the standard intensity to obtain a second judgment value; obtaining the ratio of the response speed to the standard speed to obtain a third judgment value; obtaining the ratio of the feeding time to the standard time to obtain a fourth judgment value; and calculating the feeding level based on the weights of the first judgment value, the second judgment value, the third judgment value and the fourth judgment value.

[0032] The present invention provides an attractant screening method based on inert element labeling and computer vision, comprising: placing the attractant-containing feed into the corresponding aquaculture water body, setting an above-water camera outside the aquaculture water body, and setting an underwater camera inside the aquaculture water body; preliminarily screening the feed based on the inert element content in the test animal to obtain an attractant preference value; obtaining the motion trajectory, feeding intensity, response speed and feeding time of the test animal based on the image acquired by the camera, and obtaining the feeding level of the corresponding feed based on the motion trajectory, feeding intensity, response speed and feeding time; and screening the attractant in combination with the attractant preference value and the feeding level. Through the above method, the prepared feed containing different concentrations of attractants or feed of different substances can be quickly screened in the aquaculture water body based on the inert element content to obtain the attractant preference value, and then the feeding situation is recorded by the camera to obtain the feeding level, and the feed is screened in combination with the attractant preference value to improve the accuracy of feed screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 The present invention is a flow chart of an attractant screening method based on inert element labeling and computer vision.

[0035] Figure 2 The present invention is a flow chart for the preliminary screening of feed based on the content of inert elements in the test animals.

[0036] Figure 3 The present invention is a flow chart of obtaining the motion trajectory of the test animal based on the image acquired by the camera.

[0037] Figure 4 It is a flow chart of obtaining the feeding level of corresponding feeds according to the present invention.

[0038] Figure 5 The present invention is a flow chart for obtaining the feeding intensity of the test animal based on the image acquired by the camera.

[0039] Figure 6 This is a diagram of the placement of cameras in aquaculture water bodies. DETAILED DESCRIPTION

[0040] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0041] See also Figure 1 to Figure 6 The present invention provides a method for screening attractants based on inert element labeling and computer vision, comprising:

[0042] S101 puts the feed that has passed the initial screening into the aquaculture water body, sets an above-water camera outside the aquaculture water body, and sets an underwater camera inside the aquaculture water body;

[0043] The specific method is: the lens of the above-water camera is facing the center of the aquaculture water body, and the underwater camera is placed at a depth of 50CM from the water surface and at the edge of the feeding range.

[0044] See also Figure 5 ,The above-water camera and the underwater camera are connected to the WIFI communication unit through Cat5e cable. The WIFI communication unit of the camera is linked to the WIFI transmission module of the intelligent monitoring system to transmit the video data to the experimental center. The video data is saved in the intelligent storage server of the experimental center.

[0045] S102 performs preliminary screening of feed based on the inert element content in the test animals to obtain the attractant preference value;

[0046] The specific steps are:

[0047] S201 Add inert element markers when preparing feeds with multiple attractants;

[0048] The feeds with multiple attractants can be the same type of feed but contain attractants of different concentrations, or different types of feed and contain attractants of the same concentration, so that the attractants and feeds can be screened at the same time. The inert element marker is any one of yttrium oxide, lanthanum oxide, samarium oxide, neodymium oxide and ytterbium oxide.

[0049] S202 Place feed into the aquaculture water containing a specified number of experimental animals;

[0050] S203 recovers the remaining feed after a specified time;

[0051] For example, three tanks with the same number and size of fish and shrimp in each tank can be set up as a replicate. The test fish can be fed with attractant feeds containing inert element markers at different concentrations in the morning and afternoon every day. Each time the fish are overfed, the water and air are stopped, and the remaining bait is collected after 0.5 hours of precise feeding.

[0052] S204 collects and pre-treats feces of test animals;

[0053] The specific method is: three days after the feeding experiment begins, use a scoop net to collect feces with intact capsules, store them in a refrigerator at -20℃, and then dry them at 70℃.

[0054] S205 uses plasma emission chromatography to determine the inert element content of dried feces and generate the test animals' preference value for feed.

[0055] After the fish and shrimp in the experimental group consumed attractant feed containing inert element markers at different concentrations, the inert element content of the dried feces was determined using plasma emission chromatography (ICP-MS). The feed preference value of fish and shrimp for different concentrations of attractant substances was determined by the relative proportion of specific inert elements in the feces.

[0056] S103 obtains the movement trajectory, feeding intensity, response speed and feeding time of the test animal based on the image acquired by the camera, and obtains the feeding level of the corresponding feed based on the movement trajectory, feeding intensity, response speed and feeding time;

[0057] The specific steps of obtaining the motion trajectory of the test animal based on the image acquired by the camera are:

[0058] S301 uses background difference method to perform initial segmentation on fish motion images;

[0059] S302 uses the OTSU threshold segmentation algorithm to perform secondary segmentation on the initially segmented image to obtain a processed image;

[0060] S303 uses the first-order moment to process the processed image to obtain the real-time position of the test animal;

[0061] S304 extracts the outline and skeleton of the test animal, calculates the angle formed by the fish body and the image coordinate axis, and realizes automatic detection of the movement direction of the fish according to the direction determination criterion;

[0062] S305 combines motion segmentation technology with the CAMShift algorithm to track fish moving targets in video image sequences and extract the trajectory of experimental animals' movements.

[0063] The specific steps of obtaining the feeding intensity of the test animal based on the image acquired by the camera are:

[0064] S501 pre-processes the image acquired by the camera;

[0065] Feeding intensity refers to the degree of fullness of the stomach or intestines of fish with food. Images can be acquired using nuclear magnetic resonance imaging or other imaging methods, and then the acquired images are de-noised and converted into grayscale images.

[0066] S502 extracts texture feature variable information of the feeding image based on the gray level co-occurrence matrix;

[0067] Gray-level co-occurrence matrix refers to a common method of describing texture by studying the spatial correlation characteristics of grayscale. Grayscale histogram is the result of statistical analysis of a single pixel on an image having a certain grayscale, while gray-level co-occurrence matrix is ​​the result of statistical analysis of two pixels on an image that have a certain grayscale at a certain distance. Therefore, it is possible to mark based on texture feature variable information combined with actual feeding intensity.

[0068] S503 constructs a data set based on texture feature variable information;

[0069] The above-labeled texture feature variable information is used to generate a data set for training.

[0070] S504 trains a vector machine classifier based on the data set to obtain an intensity recognition model;

[0071] S505 obtains the feeding intensity based on the intensity recognition model.

[0072] The calculation of feeding intensity is still based on the image obtained by the camera. After a series of image processing steps, the gray-level co-occurrence matrix is ​​used to extract the texture feature variable information of the feeding image, including contrast, energy, correlation, inverse difference and entropy. After that, these five feature variables are used as input vectors to construct the data set of the model, and the support vector machine classifier is trained. In order to improve the accuracy of model classification, the grid search method is used to select the optimal penalty coefficient c and kernel function parameter g of the support vector machine classifier. Finally, the trained model is used to divide the feeding intensity of fish and shrimp into four categories: weak, general, medium and strong, and finally the evaluation of fish feeding intensity is realized.

[0073] In order to quantify more accurately and combine it with other indicators to comprehensively judge the nature of the attractant, the feeding intensity can be quantified on the basis of the above judgment, that is, the ratio of the area of ​​the texture feature variable and the standard area can be extracted, ranging from 0 to 1, and these ratios can be divided into 4 intervals corresponding to the above 4 categories.

[0074] In the calculation of response speed and feeding time, the leftover bait video recorded by the video system is used for image grayscale transformation, image binarization and edge detection, and other preprocessing to obtain the leftover bait image contour; the leftover bait image is marked by setting the leftover bait contour area threshold, and the number of leftover bait is roughly counted. Then the marked components are analyzed, and it is found that there are a lot of impurities such as feces in the leftover bait, and the main component of the impurities is feces; according to the differences between the leftover bait and feces in image color, contour, and texture, six features are extracted, namely, average grayscale value, perimeter area ratio, convex hull area ratio, skeleton number, contrast, and inverse difference. The extracted feature data are classified and processed using support vector machine and decision tree algorithm respectively to realize leftover bait recognition; finally, the recognition effects of the two algorithms are analyzed. Since the leftover bait recognition model generated by the decision tree is simple, has good real-time performance, and is easy to transplant, the leftover bait recognition model generated by the decision tree is selected to screen and mark the leftover bait in the image, so as to realize the accurate counting of leftover bait in the video.

[0075] The images obtained by above-water and underwater cameras can be used to analyze the image changes in the feeding area before and after feeding, from which the response time of fish and shrimp to attractant feed can be calculated.

[0076] The image obtained by the underwater camera marks the state of the start of bait throwing, and the residual bait counting algorithm is used to set the residual bait remaining threshold in the computer. When the residual bait remaining reaches the preset threshold, the current threshold image is marked, and the feeding time is obtained by comparing the image time of the start of bait throwing and the image time when the residual bait remaining reaches the preset threshold.

[0077] The specific steps of obtaining the feeding level of the corresponding feed based on the movement trajectory, feeding intensity, response speed and feeding time are:

[0078] S401 sets the feeding area, standard intensity, standard speed and standard time, and generates an envelope diagram of the motion trajectory;

[0079] According to the feeding area during actual operation, the stomach volume of the selected target animal, the standard speed and standard time of the target animal when eating the comparison feed, the motion trajectory mainly refers to the horizontal projection of the moving trajectory of the test animal. In order to reduce accidental events, the trajectory density in the envelope diagram can be measured to remove areas with low trajectory density.

[0080] S402 obtains the ratio of the envelope diagram to the feeding area to obtain a first judgment value;

[0081] The first judgment value can be used to determine the degree of attraction of the attractant to fish. If the degree of attraction is high, the envelope diagram will be more concentrated relative to the feeding area, and the value of the first judgment value will be larger.

[0082] S403: obtaining a ratio of the food intake intensity to the standard intensity to obtain a second judgment value;

[0083] The standard intensity is obtained based on the feeding intensity of the comparison feed, and if the effect of the attractant is better, the value of the second judgment value is greater.

[0084] S404: obtaining a ratio of the response speed to the standard speed to obtain a third judgment value;

[0085] The standard speed is obtained based on the response speed of the comparison feed, and if the effect of the attractant is better, the value of the third judgment value is larger.

[0086] S405 calculates the eating level based on the weights of the first judgment value, the second judgment value, the third judgment value, and the fourth judgment value.

[0087] Assume that the first judgment value is A, the second judgment value is B, the third judgment value is C, and the fourth judgment value is D, and their corresponding weights are a1, a2, a3, and a4, respectively. Then the calculation formula for the eating level Y is:

[0088] Y=a1A+a2B+a3C+a4D

[0089] S104 screens attractants based on attractant preference values ​​and feeding levels.

[0090] The attractant preference value and the feeding level are combined to screen the attractant to improve the accuracy of the screening. An optional method is to normalize the two values ​​and then add them together for calculation.

[0091] The present invention discloses an attractant screening method based on an inert element labeling method and computer vision, comprising: placing a feed containing an attractant into a corresponding aquaculture water body, setting an above-water camera outside the aquaculture water body, and setting an underwater camera inside the aquaculture water body;

[0092] The feed was initially screened based on the inert element content in the test animals to obtain the attractant preference value;

[0093] The movement trajectory, feeding intensity, response speed and feeding time of the test animals are obtained based on the images obtained by the camera, and the feeding level of the corresponding feed is obtained based on the movement trajectory, feeding intensity, response speed and feeding time;

[0094] The attractants were screened based on the attractant preference value and feeding level.

[0095] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for screening attractants based on inert element labeling and computer vision, characterized in that: include: Put the feed containing the attractant into the corresponding aquaculture water body, set up an above-water camera outside the aquaculture water body, and set up an underwater camera inside the aquaculture water body; The feed was initially screened based on the inert element content in the test animals to obtain the attractant preference value; The movement trajectory, feeding intensity, response speed and feeding time of the test animals are obtained based on the images obtained by the camera, and the feeding level of the corresponding feed is obtained based on the movement trajectory, feeding intensity, response speed and feeding time; The attractants were screened by combining the attractant preference value and the feeding level; The specific steps of preliminarily screening the feed based on the inert element content in the test animal to obtain the attractant preference value are: Inert element markers are added when preparing various feeds containing attractants; Throwing feed into the culture water with a specified number of test animals; Recycling of residual feed after a specified period of time; Recover the feces of experimental animals and pre-treat them; Plasma emission chromatography was used to determine the inert element content of dried feces and generate the preference value of the test animals for the attractant.

2. The method for screening attractants based on inert element labeling and computer vision as claimed in claim 1, characterized in that: The inert element marker is any one of yttrium oxide, lanthanum oxide, samarium oxide, neodymium oxide and ytterbium oxide.

3. The method for screening attractants based on inert element labeling and computer vision as claimed in claim 2, characterized in that: The specific method of recovering the feces of the test animals and pre-treating them is: using a scoop net to collect the feces with intact capsules, storing them in a refrigerator at -20°C, and then drying them at a temperature of 70°C.

4. The method for screening attractants based on inert element labeling and computer vision as claimed in claim 1, characterized in that: The specific steps of obtaining the motion trajectory of the test animal based on the image acquired by the camera are: The background difference method is used to perform the initial segmentation of fish motion images; The OTSU threshold segmentation algorithm is used to perform secondary segmentation on the initially segmented image to obtain a processed image; The processed images were processed using the first-order moment to obtain the real-time position of the test animal; Extract the outline and skeleton of the test animal, calculate the angle between the fish body and the image coordinate axis, and realize automatic detection of the fish movement direction according to the direction determination criteria; The motion segmentation technology is combined with the CAMShift algorithm to track fish moving targets in video image sequences and extract the trajectory of the experimental animals' movement.

5. The method for screening attractants based on inert element labeling and computer vision as claimed in claim 1, characterized in that: The specific steps of obtaining the feeding intensity of the test animal based on the image acquired by the camera are: Preprocess the images acquired by the camera; Extracting texture feature variable information of feeding images based on gray-level co-occurrence matrix; Constructing a data set based on texture feature variable information; The vector machine classifier is trained based on the data set to obtain an intensity recognition model; The feeding intensity was obtained based on the intensity recognition model.

6. The method for screening attractants based on inert element labeling and computer vision as claimed in claim 1, characterized in that: The specific method of setting an above-water camera outside the aquaculture water body and setting an underwater camera inside the aquaculture water body is: The lens of the above-water camera is facing the center of the aquaculture water body, and the underwater camera is placed at a depth of 50CM from the water surface and at the edge of the feeding range.

7. The method for screening attractants based on inert element labeling and computer vision as claimed in claim 1, characterized in that: The specific steps of obtaining the feeding level of the corresponding feed are: Setting a feeding area, a standard intensity, a standard speed, and a standard time, and generating an envelope diagram of the motion trajectory; obtaining a ratio of the envelope diagram to the feeding area to obtain a first judgment value; Obtaining a ratio of the feeding intensity to the standard intensity to obtain a second judgment value; Obtaining a ratio of the response speed to the standard speed to obtain a third judgment value; Obtaining the ratio of the feeding time to the standard time to obtain a fourth judgment value; The eating level is calculated based on the weights of the first judgment value, the second judgment value, the third judgment value, and the fourth judgment value.

Images