Stone frog breeding growth monitoring method and system based on machine vision

Abstract

The present invention relates to the field of image processing technology, and in particular to a method and system for monitoring the growth of stone frogs based on machine vision, comprising: acquiring a monitoring image; determining a stone frog motion characteristic index of a reference area of ​​a monitoring image of a current frame; determining a stone frog reference area of ​​a monitoring image of a current frame; determining the amount of stone frog enhancement feature information of each gray level of the monitoring image of the current frame; determining a gray-scale mapping relationship of gray levels in a histogram equalization process of the monitoring image of the current frame; performing image enhancement on the monitoring image based on the gray-scale mapping relationship of the gray levels to obtain a target image, and using the target image to monitor the growth status of the stone frog. The present invention improves the recognizability between the stone frog and the stone, improves the image enhancement effect, thereby improving the recognition accuracy of the stone frog in the image, and further improves the comprehensiveness and accuracy of the evaluation of stone frog breeding growth monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for monitoring the growth of stone frog breeding based on machine vision. Background Art

[0002] The stone frog is a kind of frog widely distributed in southern my country. It mostly lives in the water pits of mountain streams at an altitude of 370-1500 meters, in the cracks of rocks beside the streams, or in stone caves. This kind of habitat is characterized by coolness and dimness. With the rapid development of computer technology, computer vision technology has brought innovative solutions to the growth monitoring of stone frogs in the field of stone frog breeding. By reasonably arranging cameras and other image acquisition equipment in the breeding site, the appearance characteristics of stone frogs such as morphology and behavior can be accurately captured using image recognition technology. The deep learning algorithm can be trained based on a large amount of stone frog image data, and then accurately judge the growth status of stone frogs. At the same time, data analysis technology can combine the image information of the breeding environment to analyze the impact of environmental factors on the growth of stone frogs, such as whether the clarity of water quality, the temperature and humidity of the breeding site are suitable for the survival of stone frogs.

[0003] However, in the actual application of stone frog breeding and growth monitoring, since stone frogs are nocturnal, in order to obtain image data of their nocturnal activity status, it is usually necessary to use night vision cameras to shoot and monitor them. However, the skin of stone frogs is relatively rough, and in the breeding environment, their skin color and reflectivity are often similar to those of surrounding stones, soil and other objects, making it easy for stone frogs to blend into the background environment. What is more difficult is that when stone frogs are partially blocked by similar objects such as stones and soil, their outlines become blurred, which greatly increases the difficulty of accurately identifying and judging the state of stone frogs directly from the image. In order to improve this situation, image enhancement technology is considered to improve the recognition of stone frog areas in images. However, most of the existing image enhancement technologies are based on the number distribution of pixels to perform grayscale value stretching enhancement, but the actual characteristics of stone frogs in the image, such as the unique shape and texture characteristics of stone frogs and the relative relationship with the surrounding environment, will affect the recognition of the image. Therefore, when enhancing the image of stone frog breeding monitoring based on the number of pixels, it is often difficult to achieve the ideal enhancement effect, resulting in low recognition accuracy of stone frogs in the image, and further resulting in low comprehensiveness and accuracy of the evaluation of stone frog breeding growth monitoring. Summary of the invention

[0004] In order to solve the technical problem that the recognition accuracy of stone frogs in images is low, which further leads to low comprehensiveness and accuracy of the evaluation of stone frog breeding growth monitoring, the purpose of the present invention is to provide a stone frog breeding growth monitoring method and system based on machine vision.

[0005] In order to solve the above technical problems, the technical solutions adopted are as follows:

[0006] The embodiment of the present invention provides a method for monitoring the growth of stone frog breeding based on machine vision, comprising: acquiring a monitoring image of a breeding environment to be monitored in which the stone frog is located; determining a stone frog motion feature index of a reference area of ​​the monitoring image of the current frame by using difference pixels in a monitoring image adjacent to the monitoring image of the current frame, wherein the reference area represents a superpixel block where the stone frog may appear; determining a stone frog reference area of ​​the monitoring image of the current frame according to the stone frog motion feature index; determining the amount of stone frog enhanced feature information of each gray level of the monitoring image of the current frame according to edge lines, the main axis direction of the edge lines and pixels of each gray level in the stone frog reference area; determining a gray-scale mapping relationship of gray levels of the monitoring image of the current frame in a histogram equalization process according to a first number of pixels of each gray level of the monitoring image of the current frame, a second number of all pixels of the monitoring image of the current frame, a third number of gray levels in the stone frog reference area and the amount of stone frog enhanced feature information of each gray level; performing image enhancement on the monitoring image based on the gray-scale gray-scale mapping relationship to obtain a target image, and using the target image to monitor the growth status of the stone frog.

[0007] Optionally, the frog motion characteristic index of the reference area of ​​the monitoring image of the current frame is determined by using the difference pixels in the monitoring image adjacent to the monitoring image of the current frame, including: segmenting the monitoring image by using a superpixel segmentation algorithm to obtain multiple superpixel blocks; obtaining the difference pixels between the monitoring image of the current frame and the monitoring image of the previous frame adjacent to the monitoring image of the current frame by using a frame difference method, and marking the superpixel blocks with the difference pixels, and using the superpixel blocks with the difference pixels as the reference area of ​​the monitoring image of the current frame; determining the reference area of ​​the monitoring image of the current frame according to the convex hull of the connected domain of each difference pixel in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of each difference pixel in the monitoring image of the previous frame. Dynamic similarity features of the connected domain of the difference pixels in the region; selecting the connected domain of the difference pixels with the largest dynamic similarity features as the reference connected domain of the difference pixels in the reference region of the monitoring image of the current frame in the reference connected domain of the monitoring image of the previous frame; calculating the centroid coordinate variance of the connected domain between the difference pixels in the reference region of the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame; determining the frog motion feature index of the reference region of the monitoring image of the current frame according to the mean value of the dynamic similarity features, the centroid coordinate variance of the connected domain and the mean value of the area difference between the difference pixels in the reference region of the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame.

[0008] Optionally, determining the dynamic similarity feature of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame based on the convex hull of the connected domain of each difference pixel in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of each difference pixel in the monitoring image of the previous frame includes: calculating the Euclidean distance between the centroid pixel coordinates of the convex hull of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame and the centroid pixel coordinates of the convex hull of the connected domain of the difference pixels in the monitoring image of the previous frame; calculating the Hausdorff distance between the convex hull of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of the difference pixels in the monitoring image of the previous frame; and determining the inverse of the first product of the Hausdorff distance and the Euclidean distance as the dynamic similarity feature.

[0009] Optionally, determining the stone frog motion characteristic index of the reference area of ​​the monitoring image of the current frame based on the dynamic similarity feature mean, the connected domain centroid coordinate variance and the area difference mean between the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of frames before the current frame includes: normalizing the dynamic similarity feature mean to obtain a normalized value; calculating the second product of the normalized value, the connected domain centroid coordinate variance and the area difference mean, and superimposing the second products to obtain the stone frog motion characteristic index of the reference area of ​​the monitoring image of the current frame.

[0010] Optionally, determining the stone frog reference area of ​​the monitoring image of the current frame according to the stone frog motion feature index includes: selecting a target reference area with a largest stone frog motion feature index; obtaining a target grayscale level in the target reference area; and selecting a superpixel block containing the target grayscale level from the superpixel block of the monitoring image of the current frame as the stone frog reference area.

[0011] Optionally, determining the amount of stone frog enhanced feature information of each gray level of the monitoring image of the current frame according to the edge line in the stone frog reference area, the main axis direction of the edge line and the pixel points of each gray level in the stone frog reference area includes:

[0012] The edge lines and the main axis directions of the edge lines in the stone frog reference area are obtained, and the reference pixel points of the edge lines are determined based on the edge lines and the main axis directions; the distance between the reference pixel points of the current edge line and the reference pixel points of other edge lines is calculated, and the minimum distance is selected as the gap distance feature of the current edge line; the stone frog occlusion interference factor of the stone frog reference area is determined according to the fourth number of pixel points in the stone frog reference area, the gap distance feature of the edge line and the image grayscale mean of the edge line; the area of ​​the convex hull of the pixel points of each gray level in the stone frog reference area is determined, and the pixel points of the edge lines of the stone frog reference area are clustered to obtain cluster clusters; the cluster cluster with the smallest number of pixel points is selected, and the Euclidean distance between each pixel point in the cluster cluster and the nearest pixel point in other cluster clusters is obtained, and the Euclidean distance is used as the enhanced feature distance; the stone frog enhanced feature information amount of each gray level of the current monitoring image is determined according to the stone frog occlusion interference factor, the enhanced feature distance and the area.

[0013] Optionally, determining the stone frog occlusion interference factor of the stone frog reference area based on a fourth number of pixel points in the stone frog reference area, the gap distance characteristics of the edge lines, and the image grayscale mean of the edge lines includes: calculating the reciprocal of the third product between the image grayscale mean of the edge lines and the gap distance characteristics of the edge lines, and superimposing the reciprocals of the third products of each edge line to obtain a superimposed value; determining the fourth product between the superimposed value and the fourth number as the stone frog occlusion interference factor of the stone frog reference area.

[0014] Optionally, determining the amount of stone frog enhanced feature information at each grayscale level of the current monitoring image based on the stone frog occlusion interference factor, enhanced feature distance and area includes: calculating the fifth product of the area of ​​the convex hull of the grayscale pixel points in the stone frog reference area and the average value of the enhanced feature distances at each grayscale level, and a first ratio between the stone frog occlusion interference factor of the stone frog reference area and the fifth product; superimposing the first ratios of the grayscale levels of each stone frog reference area to obtain the amount of stone frog enhanced feature information at each grayscale level of the current monitoring image.

[0015] Optionally, determining the grayscale mapping relationship of the grayscale of the monitoring image of the current frame during the histogram equalization process based on the first number of pixels of each grayscale level of the monitoring image of the current frame, the second number of all pixels of the monitoring image of the current frame, the third number of grayscale levels in the frog reference area, and the frog enhanced feature information amount of each grayscale level includes: calculating a first difference between the third number and a predetermined value; superimposing the frog enhanced feature information amount of each grayscale level to obtain the superimposed enhanced feature information amount; calculating a second ratio between the frog enhanced feature information amount of each grayscale level and the superimposed enhanced feature information amount, and a third ratio between the first number of pixels of each grayscale level and the second number; superimposing the sixth product between each second ratio and each third ratio and multiplying it by the first difference to obtain a seventh product; and processing the seventh product using a rounding function to obtain a grayscale mapping relationship.

[0016] In the second aspect, an embodiment of the present invention provides a stone frog breeding and growth monitoring system based on machine vision, comprising: a processor and a memory; wherein the memory is used to store computer programs that can be run on the processor; the processor is used to execute the program stored in the memory to implement the steps of the stone frog breeding and growth monitoring method based on machine vision as mentioned in the first aspect.

[0017] The present invention has the following beneficial effects: after obtaining the monitoring image of the breeding environment to be monitored where the stone frog is located, the stone frog motion characteristic index representing the stone frog characteristics in the reference area is determined by analyzing the difference pixel points in each frame of the monitoring image, and the stone frog reference area is further selected according to the stone frog motion characteristic index. The image composition in each stone frog reference area is analyzed, that is, the edge line, the main axis direction of the edge line and the pixel points of each gray level in the stone frog reference area are analyzed, so as to evaluate the stone frog enhanced feature information amount of each gray level, and the monitoring image is gray-scale stretched and enhanced based on the stone frog enhanced feature information amount, the pixel points in the monitoring image and the gray level, so as to achieve the purpose of image enhancement. Therefore, the embodiment of the present invention improves the recognizability between the stone frog and the stone by evaluating the specific image features of the environment where the stone frog is located and the actual feature stretching gray level of the stone frog in the monitoring image, and improves the image enhancement effect, thereby improving the recognition accuracy of the stone frog in the image, and further improving the comprehensiveness and accuracy of the evaluation of the stone frog breeding growth monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are 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.

[0019] Figure 1 A flow chart of a method for monitoring the growth of stone frog breeding based on machine vision provided by one embodiment of the present invention;

[0020] Figure 2 A schematic structural diagram of a stone frog breeding and growth monitoring system based on machine vision provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a stone frog breeding growth monitoring method and system based on machine vision proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0022] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0023] The specific scheme of the method for monitoring the growth of stone frog breeding based on machine vision provided by the present invention is described in detail below in conjunction with the accompanying drawings.

[0024] Embodiment 1:

[0025] See also Figure 1 , which shows a flow chart of a method for monitoring the growth of stone frog breeding based on machine vision provided by an embodiment of the present invention, including:

[0026] Step S101, obtaining a monitoring image of the breeding environment to be monitored where the stone frog is located.

[0027] Specifically, in the embodiment of the present invention, taking into account the characteristics of the nocturnal activities of the stone frog, a high-definition camera with infrared night vision function is set in the breeding environment to be monitored to eliminate the influence of light on the activities of the stone frog, ensure all-weather monitoring, and perform continuous shooting, and adjust the frame rate as needed to capture the daily activities and growth details of the stone frog. In the embodiment of the present invention, the camera's image acquisition frame rate is set to 30FPS. Further, the camera transmits the monitoring image to the central processing system in real time through a wired or wireless network to ensure the immediacy and integrity of the data. Exemplarily, the stone frog is usually a greenhouse simulating a breeding environment in a stream or pond, which meets the conditions of providing water, land, stone caves and other hiding places, feeding, etc. Among them, the adult frog is weak and stable during the non-breeding period. When the stone frog is blocked by the stone cave, the stone frog's skin may be similar to the color and reflectivity of the surrounding stones and soil. The image collected by the night vision camera is prone to blurred outlines of the stone frog, which makes it difficult to distinguish in the image.

[0028] Among them, the monitoring images collected for night-time breeding of stone frogs are infrared images. Infrared images are usually single-channel images, expressed in the form of grayscale images. The grayscale value of a pixel represents the intensity of near-infrared light reflected by the area corresponding to the pixel. Stone frogs are usually raised in greenhouses to simulate the breeding environment in streams or ponds, satisfying the conditions of water, land, stone caves, etc., hiding places, feeding, etc. Among them, adult frogs are weak and stable in their activities during the non-breeding period. When stone frogs are blocked by stone caves, the skin of stone frogs may be similar in color and reflectivity to the surrounding stones and soil. The monitoring images collected by night vision cameras are prone to blurred outlines of stone frogs, making them difficult to distinguish in the monitoring images. Therefore, it is necessary to combine the grayscale level of the stone frog image in the monitoring image with stretching enhancement processing to highlight its contrast with similar environments such as stone caves.

[0029] Step S102, determining the frog motion feature index of the reference area of ​​the monitoring image of the current frame by using the difference pixel points in the monitoring image adjacent to the monitoring image of the current frame.

[0030] Among them, the reference area represents the superpixel block where the stone frog may appear.

[0031] Specifically, the grayscale value of the pixel in the infrared night vision monitoring image is reflected by the reflection intensity of the infrared light source on the surface of the receiving object. In the stone frog breeding scene, the reflection intensity of the infrared light source on the surface of the material in different areas is different. In order to distinguish different areas in the monitoring image, the monitoring image needs to be processed in blocks. Among them, the reference area represents the super-pixel block where the stone frog part may appear in the image. In the breeding scene, stone frogs usually come out of hiding places such as stone caves to forage for feed at night, so the area where changes appear in the image during the monitoring process may represent a moving stone frog. However, in the breeding scene, the moving area may also include natural changes such as water surface fluctuations and plants swaying in the wind. Therefore, the embodiment of the present invention also needs to further analyze the reference area.

[0032] More specifically, as an optional embodiment of the present invention, the frog motion feature index of the reference area of ​​the monitoring image of the current frame is determined by using the difference pixel points in the monitoring image adjacent to the monitoring image of the current frame, including: segmenting the monitoring image by using a superpixel segmentation algorithm to obtain a plurality of superpixel blocks; obtaining the difference pixel points between the monitoring image of the current frame and the monitoring image of the previous frame adjacent to the monitoring image of the current frame by using a frame difference method, and marking the superpixel blocks with the difference pixel points, and using the superpixel blocks with the difference pixel points as the reference area of ​​the monitoring image of the current frame; determining the reference area of ​​the monitoring image of the current frame according to the convex hull of the connected domain of each difference pixel point in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of each difference pixel point in the monitoring image of the previous frame. dynamic similarity features of the connected domain of difference pixels in the reference area of ​​the monitoring image of the current frame; selecting the connected domain of difference pixels with the largest dynamic similarity features as the reference connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame in the reference connected domain of the monitoring image of the previous frame; calculating the centroid coordinate variance of the connected domain between the difference pixels in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame; determining the frog motion feature index of the reference area of ​​the monitoring image of the current frame according to the mean value of dynamic similarity features, the centroid coordinate variance of the connected domain and the mean value of area difference between the difference pixels in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame.

[0033] Specifically, the embodiment of the present invention uses a superpixel segmentation algorithm to process the monitoring image and divide the monitoring image into multiple superpixel blocks. In the embodiment of the present invention, the Simple Linear Iterative Clustering (SLIC) algorithm is selected to perform superpixel segmentation on the monitoring image. The number of superpixel blocks can be obtained by the elbow method. After segmentation, the superpixel blocks are composed of adjacent pixels with similar texture, color, brightness and other characteristics. Further, the frame difference method is used to obtain the difference pixel points between the monitoring image of the current frame and the monitoring image of the previous frame, and the superpixel blocks with difference pixel points are marked as reference areas in the monitoring image. The reference area may include stone frogs or naturally changing moving objects. In order to distinguish the stone frog image partially blocked by the stone cave, it is first necessary to select the reference area of ​​the current movement, select the area with similar grayscale distribution in the monitoring image through the grayscale distribution of the reference area, and analyze the image features of the stone frog covered in other areas, and evaluate the amount of information of the covered stone frog carried by different grayscale levels. Among them, the infrared night camera usually actively projects an infrared light source and receives the reflected infrared light for imaging. For moving water surface ripples, when the light source is fixed, when the ripple movement exceeds the camera receiving angle, it no longer has a brighter reflective feature in the picture, but the new ripples entering the camera receiving angle will still have a reflective feature. During the monitoring process, the water surface ripples are similar to swaying plants and show a certain area of ​​fluctuating reflective features. However, the foraging movement of the stone frog does not have this feature, so it can be analyzed through the motion features of the adjacent frame difference images in the past multiple frames of monitoring images.

[0034] Based on the above analysis, the embodiment of the present invention first obtains the connected domain of the difference pixels of the monitoring images of adjacent frames, and performs convex hull detection on the connected domain of the difference pixels, and extracts the shape information of the connected domain of the difference pixels. Then, the difference pixels between the monitoring image of the previous frame of the monitoring image of the current frame and the monitoring image of the previous frame (that is, the monitoring image of the previous frame of the monitoring image of the current frame) are obtained, and the convex hull detection is performed on the connected domain of the difference pixels between the monitoring image of the previous frame and the monitoring image of the previous frame. The centroid pixel coordinates of the connected domain convex hull of the difference pixels in the reference area of ​​the monitoring image of the current frame, the centroid pixel coordinates of the connected domain convex hull of the difference pixels in the monitoring image of the previous frame, the connected domain convex hull of the difference pixels in the reference area of ​​the monitoring image of the current frame, and the connected domain convex hull of the difference pixels in the monitoring image of the previous frame are obtained.

[0035] More specifically, as an optional embodiment of the present invention, determining the dynamic similarity feature of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame based on the convex hull of the connected domain of each difference pixel in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of each difference pixel in the monitoring image of the previous frame includes: calculating the Euclidean distance between the centroid pixel coordinates of the convex hull of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame and the centroid pixel coordinates of the convex hull of the connected domain of the difference pixels in the monitoring image of the previous frame; calculating the Hausdorff distance between the convex hull of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of the difference pixels in the monitoring image of the previous frame; and determining the inverse of the first product of the Hausdorff distance and the Euclidean distance as the dynamic similarity feature.

[0036] Specifically, the embodiment of the present invention specifically uses the following formula to calculate the dynamic similarity feature:

[0037]

[0038] In the above formula, Represents the dynamic similarity features of the connected domain of difference pixels in the reference area of ​​the monitoring image of the current frame. It represents the Euclidean distance between the centroid pixel coordinates of the connected domains of the two convex hulls, that is, the Euclidean distance between the centroid pixel coordinates of the connected domain convex hull of the difference pixel in the reference area of ​​the monitoring image of the current frame and the centroid pixel coordinates of the connected domain convex hull of the difference pixel in the monitoring image of the previous frame. It represents the Hausdorff distance between two convex hulls, that is, the Hausdorff distance between the convex hull of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame and the convex hull of the connected domain of the difference pixels in the monitoring image of the previous frame. By combining the positional relationship and shape relationship between convex hulls, their dynamic similarity features are evaluated.

[0039] More specifically, the embodiment of the present invention selects the dynamic similarity features of the connected domain of the difference pixels between the monitoring image of the previous frame and the current reference area of ​​the monitoring image of the current frame The largest connected domain is used as the reference connected domain of the difference pixel point connected domain of the current reference area of ​​the current monitoring image in the previous frame of the monitoring image. Then repeat the above steps to obtain the reference connected domain of the monitoring image of each frame in the past period of time. In the embodiment of the present application, considering that the time of a stone frog's jump is short, after considering the motion error before or after the jump, the reference connected domain of each monitoring image frame in the past 1 second is calculated.

[0040] More specifically, a connected domain of a certain difference pixel point of the monitoring image of the current frame is selected, and a reference connected domain of the corresponding monitoring images of each frame before is selected, and the centroid coordinate variance of the connected domain of a certain difference pixel point and the reference connected domain of each monitoring image frame is calculated. According to the above analysis, the natural image motion has the characteristics of regional fluctuation, while the foraging motion of the stone frog does not have this characteristic and its motion is unidirectional. Therefore, the centroid coordinate variance of each connected domain The larger it is, the higher the probability that the connected domain of the difference pixels in the monitoring image of the current frame represents the motion characteristics of the stone frog.

[0041] In addition, since the jumping trajectory of the stone frog is a parabola-like process, for a camera with a fixed imaging angle, its displacement during the movement is usually uneven; for the natural change of reciprocating motion, the difference in the displacement of the movement trajectory at different frame times is relatively low, so the shape and geometry differences of the connected domains of the difference pixels of the monitoring images of different frames can be used for analysis and evaluation. Specifically, the area difference between the connected domain of the difference pixels of the monitoring image of a certain frame and the reference connected domain corresponding to the monitoring image of the previous frame is calculated. ,According to the above analysis, the larger the value is, the greater the probability that the connected domain of ,the difference pixels of the monitoring image of the current frame represents the ,motion characteristics of the stone frog.

[0042] More specifically, the frog motion feature index of the reference area of ​​the current frame monitoring image is calculated by combining the position difference and area difference of the connected domain of each difference pixel point in the selected multi-frame monitoring image in the past period of time. . As an optional embodiment of the present invention, according to the dynamic similarity feature mean, the connected domain centroid coordinate variance and the area difference mean between the connected domain of the difference pixel points in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame, determining the frog motion feature index of the reference area of ​​the monitoring image of the current frame includes: normalizing the dynamic similarity feature mean to obtain a normalized value; calculating the second product of the normalized value, the connected domain centroid coordinate variance and the area difference mean, and superimposing the second products to obtain the frog motion feature index of the reference area of ​​the monitoring image of the current frame.

[0043] Specifically, the following formula can be used to calculate the frog motion feature index of the reference area of ​​the monitoring image of the current frame:

[0044]

[0045] In the above formula, Indicates the frog motion feature index of the reference area of ​​the monitoring image of the current frame. n indicates the number of connected domains of difference pixels in the current reference area. It represents the mean value of the dynamic similarity features of the connected domain of the i-th difference pixel point corresponding to the reference connected domain in each frame of the monitoring image. Represents the mean area difference between the connected domain of the i-th difference pixel point of the monitoring image of each frame and the corresponding reference connected domain. Represents the coordinate variance of the centroid of the connected domain between the connected domain of the i-th difference pixel and the corresponding reference connected domain. Represents the normalization function, which is used to Normalization is performed. The proportion of connected domains with high dynamic similarity characteristics between different pixels in the calculation process is enlarged to further improve the calculation accuracy of the current stone frog motion feature index. The larger the value is, the closer the motion characteristics of the current reference area in the past period of time are to the motion characteristics of the stone frog foraging and jumping, and the higher the probability that it represents the stone frog image.

[0046] Step S103, determining the frog reference area of ​​the monitoring image of the current frame according to the frog motion feature index.

[0047] Specifically, the target reference area with the largest stone frog motion feature index can be selected, the target grayscale in the target reference area can be obtained, and finally the superpixel block containing the target grayscale can be selected from the superpixel block of the monitoring image of the current frame as the stone frog reference area.

[0048] More specifically, we can select the stone frog movement characteristic index The largest reference area is obtained and the gray level therein is obtained, and then the super-pixel blocks containing its gray level in the monitoring image are selected, and these super-pixel blocks are represented as the stone frog reference area.

[0049] Step S104, determining the amount of frog enhancement feature information of each gray level of the monitoring image of the current frame according to the edge line in the frog reference area, the main axis direction of the edge line and the pixel points of each gray level in the frog reference area.

[0050] Specifically, the purpose of enhancing the monitoring images of stone frog farming is to highlight the outline of the stone frog that is blocked by other surface features such as stone holes and areas similar to the stone frog surface. Therefore, when evaluating the stone frog information, the stone frog area blocked by the interference area should be given priority. When performing superpixel segmentation on the monitoring images of stone frog farming, local areas with similar color and texture features will be divided into the same superpixel block. The stone frog image blocked by the interference area is usually composed of part of the stone frog's body area and the stone area, and its image area is usually larger than that of the general stone frog image. Therefore, the number of pixels in different stone frog reference areas is obtained. , the number of pixels in the stone frog reference area It can characterize the area characteristics of the stone frog reference area. According to the above analysis, the larger the value, the higher the probability that the stone frog in the current superpixel block is blocked by stones. However, the evaluation based only on the number of pixels in the stone frog reference area fails to take into account the situation where multiple stone frogs are divided into the same superpixel block when they are close to each other, so further analysis is needed. For the stone frog reference area blocked by stone caves, it is necessary to further analyze the stone cave structure and the general stone frog image features. The stone cave is a hiding place composed of multiple stones of different sizes, with a gap texture of stone accumulation. The gaps between the stones formed are smaller than the gaps between the spots on the surface of the general stone frog. This feature can be used to further improve the accuracy of the analysis of occlusion interference factors in different stone frog reference areas.

[0051] Based on the above analysis, as an optional embodiment of the present invention, determining the stone frog enhanced feature information of each gray level of the monitoring image of the current frame according to the edge line, the main axis direction of the edge line and the pixel points of each gray level in the stone frog reference area includes: obtaining the edge line and the main axis direction of the edge line in the stone frog reference area, and determining the reference pixel points of the edge line based on the edge line and the main axis direction; calculating the distance between the reference pixel points of the current edge line and the reference pixel points of other edge lines, and selecting the minimum distance as the gap distance feature of the current edge line; according to the fourth number of the pixel points of the stone frog reference area The stone frog occlusion interference factor of the stone frog reference area is determined by the amount, the gap distance characteristics of the edge line and the image grayscale mean of the edge line; the area of ​​the convex hull of the pixel points of each gray level in the stone frog reference area is determined, and the pixel points of the edge line of the stone frog reference area are clustered to obtain cluster clusters; the cluster cluster with the smallest number of pixels is selected, and the Euclidean distance between each pixel point in the cluster cluster and the nearest pixel point in other cluster clusters is obtained, and the Euclidean distance is used as the enhanced feature distance; the stone frog enhanced feature information amount of each gray level of the current monitoring image is determined according to the stone frog occlusion interference factor, the enhanced feature distance and the area.

[0052] Specifically, the embodiment of the present invention can use the edge detection algorithm (computational theory of edge detection, Canny) to obtain the edge line in a certain stone frog reference area. The principal component analysis algorithm (principal components analysis, PCA) is used to extract the main axis direction of each edge line in the current stone frog reference area. The pixel points tangent to each edge line and its main axis direction are selected as the reference pixel points of the current edge line, and then the distance between the reference pixel points of each edge line is calculated, and the minimum value is selected as the gap distance feature of the current edge line. After selecting the main axis direction and its tangent point, the distance between the reference pixel points of other edge lines is calculated. After selecting the minimum distance, the distance between the texture edges in the current reference stone frog reference area can be reflected. According to the above analysis, the smaller the distance value, the smaller the gap area formed by spots or stones in the current stone frog reference area, and the higher the probability of being blocked and disturbed by stones in the current stone frog reference area.

[0053] More specifically, as an optional embodiment of the present invention, determining the stone frog occlusion interference factor of the stone frog reference area according to the fourth number of pixel points in the stone frog reference area, the gap distance characteristics of the edge lines, and the image grayscale mean of the edge lines includes: calculating the reciprocal of the third product between the image grayscale mean of the edge lines and the gap distance characteristics of the edge lines, and superimposing the reciprocals of the third products of each edge line to obtain a superimposed value; determining the fourth product between the superimposed value and the fourth number as the stone frog occlusion interference factor of the stone frog reference area.

[0054] Specifically, the embodiment of the present invention specifically uses the following formula to calculate the stone frog occlusion interference factor of the stone frog reference area:

[0055]

[0056] In the above formula, Represents the stone frog occlusion interference factor of the stone frog reference area. Indicates the number of edge lines in the current stone frog reference area. Indicates The gap distance characteristics of the edge lines. Indicates The image grayscale mean of the edge lines. Indicates the number of pixels in the reference area of ​​the frog. The infrared night camera realizes imaging by collecting actively projected infrared reflections, but the stone crevices do not have a surface that can directly reflect the projected infrared light. In the imaging, its grayscale characteristics are relatively low compared to the surface of the frog. The calculation process of summing the gap distance features is adjusted, and after combining with the number of pixels in the frog reference area, Z is used to represent the frog occlusion interference factor of the current frog reference area.

[0057] More specifically, after obtaining the stone frog occlusion interference factors of different stone frog reference areas, for the stone frog reference area with a larger stone frog occlusion interference factor, it is necessary to amplify the reference weight thereof when evaluating the stone frog enhanced feature information amount of different gray levels. In the stone frog image disturbed by stone occlusion, in the stone frog reference area, the stone plays a shielding role on the stone frog, so in the current stone frog reference area, the gray level representing the stone frog feature is relatively concentrated in its spatial distribution compared with the gray level representing the stone feature, and the area of ​​aggregation is relatively small. Therefore, the embodiment of the present invention calculates the area of ​​the convex hull of the pixel points of each gray level in the current stone frog reference area. According to the above analysis, the smaller the area of ​​the convex hull, the higher the probability that the current gray level represents the stone frog image feature. In addition, it should be noted that the stone frog image disturbed by stone occlusion that needs to be enhanced should ensure that a considerable part of its body can be captured by the camera, otherwise its shielded part is too large and does not have the monitoring condition. For the monitoring image after superpixel segmentation, the edges of the superpixel block are usually composed of the edges of the stone frog and the stone. For the image of the stone frog in the stone cave and blocked by the stone, the edge of the superpixel block is usually composed of a larger number of edges representing the characteristics of the stone and a smaller number of edge pixels representing the characteristics of the stone frog.

[0058] Furthermore, the embodiment of the present invention uses the k-means clustering algorithm (k-means clustering, k-means) to cluster the pixel points of the edge line of the current stone frog reference area, wherein the clustering relationship is the Euclidean distance between the gray value of the pixel point and the coordinates of the pixel point, and the number of clusters is set to 2. After the clustering is completed, a cluster with a smaller number of pixels is selected, and the Euclidean distance between the coordinates of each pixel point in this cluster and the coordinates of the nearest pixel point of another cluster is obtained, and it is used as the enhanced feature distance. The smaller the value, the closer the distance between the current pixel point and the pixel point of another cluster is, and the more it needs to be stretched and enhanced to highlight it.

[0059] More specifically, as an optional embodiment of the present invention, determining the amount of stone frog enhanced feature information at each gray level of the current monitoring image based on the stone frog occlusion interference factor, enhanced feature distance and area includes: calculating the fifth product of the area of ​​the convex hull of the gray level pixel points in the stone frog reference area and the average value of the enhanced feature distances of each gray level, and the first ratio between the stone frog occlusion interference factor of the stone frog reference area and the fifth product; superimposing the first ratio of the gray levels of each stone frog reference area to obtain the amount of stone frog enhanced feature information at each gray level of the current monitoring image.

[0060] Specifically, the embodiment of the present invention specifically uses the following formula to calculate the amount of stone frog enhanced feature information of each gray level of the current monitoring image:

[0061]

[0062] In the above formula, represents the amount of enhanced feature information of the frog at each gray level of the current monitoring image. u represents the number of frog reference areas in the monitoring image. represents the frog occlusion interference factor of the vth frog reference area, Represents the area of ​​the convex hull of the medium grayscale pixels in the vth stone frog reference area. It represents the average value of the enhanced feature distance of the pixels of each gray level in the vth stone frog reference area in the monitoring image. It is worth noting that if the current gray level does not have an edge pixel, the value is 1. According to the above analysis, by As the weight, the enhanced feature distance and the area of ​​the convex hull are then used to evaluate the amount of enhanced feature information of stone frogs at different gray levels. The larger the value, the greater the amount of stone frog information contained when the current gray level is blocked by stones in the cave.

[0063] Step S105, determines the grayscale mapping relationship of the grayscale of the monitoring image of the current frame during the histogram equalization process according to the first number of pixels of each grayscale level of the monitoring image of the current frame, the second number of all pixels of the monitoring image of the current frame, the third number of grayscale levels in the stone frog reference area, and the amount of stone frog enhanced feature information of each grayscale level.

[0064] Specifically, after obtaining the enhanced characteristic information of the frog at each gray level, the monitoring image is enhanced by histogram equalization to adjust the gray level mapping relationship. In the embodiment of the present invention, the enhanced characteristic information of the frog at all gray levels is added to obtain the superimposed enhanced characteristic information of the ath monitoring image. .

[0065] As an optional embodiment of the present invention, according to the first number of pixels of each gray level of the monitoring image of the current frame, the second number of all pixels of the monitoring image of the current frame, the third number of gray levels in the frog reference area and the frog enhanced feature information of each gray level, determining the gray level mapping relationship of the gray levels of the monitoring image of the current frame in the histogram equalization process includes: calculating the first difference between the third number and the predetermined value; superimposing the frog enhanced feature information of each gray level to obtain the superimposed enhanced feature information; calculating the second ratio between the frog enhanced feature information of each gray level and the superimposed enhanced feature information, and the third ratio between the first number and the second number of pixels of each gray level; superimposing the sixth product between each second ratio and each third ratio and multiplying it with the first difference to obtain a seventh product; processing the seventh product using a rounding function to obtain a gray level mapping relationship.

[0066] Specifically, the embodiment of the present invention specifically uses the following formula to calculate the gray level mapping relationship:

[0067]

[0068] In the above formula, is the gray level mapping relationship of the hth gray level in the histogram equalization process. round[] is the rounding function, and L is the third number of gray levels in the monitoring image. is the first number of pixels of the tth gray level. is the second number of all pixels of the monitoring image of the current frame. Enhance the feature information of the stone frog at the tth gray level. is the amount of superimposed enhanced feature information of the a-th monitoring image.

[0069] Step S106, performing image enhancement on the monitoring image based on the grayscale mapping relationship to obtain a target image, and using the target image to monitor the growth status of the stone frog.

[0070] Specifically, the embodiment of the present invention combines the grayscale mapping relationship of each grayscale level and the CLAHE contrast-limited histogram equalization method to enhance the monitoring image, thereby realizing adaptive enhancement of the stone frog breeding image and reducing the influence of blurred and difficult-to-distinguish edges caused by the stone frog being blocked by the stone cave.

[0071] More specifically, after obtaining the enhanced monitoring image of stone frog breeding, the enhanced target image is input into the trained neural network model, the growth status of the stone frog shown in the current target image is evaluated, and the monitoring result of the growth status of the stone frog is output.

[0072] Among them, the specific training process of the neural network model is as follows:

[0073] First, a large number of high-quality night vision images of stone frogs at different growth stages were collected, including images of stone frogs in different growth states, health conditions and environmental conditions, and the images were divided into training sets and validation sets according to a ratio of 7:3. Then, each image was annotated in detail, marking the location, growth stage (such as young frogs, adult frogs, etc.), and health status (such as health and signs of disease) of the stone frogs. Then, the convolutional neural network (CNN) was used as the training architecture, and the loss function was defined as the cross entropy function. The training process was optimized by the gradient descent method. Finally, the training set was used to train until the loss function converged, and the validation set was input to verify the training results, and the model training was completed.

[0074] After acquiring the monitoring image of the breeding environment to be monitored where the stone frog is located, the embodiment of the present invention determines the stone frog motion characteristic index representing the stone frog characteristics in the reference area by analyzing the difference pixel points in each frame of the monitoring image, and further selects the stone frog reference area according to the stone frog motion characteristic index. The image composition in each stone frog reference area is analyzed, that is, the edge line, the main axis direction of the edge line and the pixel points of each gray level in the stone frog reference area are analyzed, so as to evaluate the stone frog enhanced feature information amount of each gray level, and based on the stone frog enhanced feature information amount, the pixel points and the gray level in the monitoring image, the monitoring image is stretched and enhanced in gray level to achieve the purpose of image enhancement. Therefore, the embodiment of the present invention improves the recognizability between the stone frog and the stone by evaluating the specific image features of the environment where the stone frog is located and the actual feature stretching gray level of the stone frog in the monitoring image, and improves the image enhancement effect, thereby improving the recognition accuracy of the stone frog in the image, and further improving the comprehensiveness and accuracy of the evaluation of the stone frog breeding growth monitoring.

[0075] Embodiment 2:

[0076] Corresponding to the stone frog breeding growth monitoring method based on machine vision provided in the above embodiment, based on the same technical concept, the embodiment of the present invention also provides a stone frog breeding growth monitoring system based on machine vision, and the stone frog breeding growth monitoring system based on machine vision is used to execute the above stone frog breeding growth monitoring method based on machine vision. Figure 2 A schematic diagram of a system for monitoring the growth of stone frogs based on machine vision is provided to implement various embodiments of the present invention. Figure 2As shown in the hardware level, the stone frog breeding growth monitoring system based on machine vision includes a processor, and optionally, an internal bus, a network interface, and a memory. Among them, the memory may include a memory, such as a high-speed random access memory (Random-Access Memory, RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage, etc. Of course, the stone frog breeding growth monitoring system based on machine vision may also include hardware required for other services.

[0077] The processor, network interface and memory can be interconnected through an internal bus, which can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0078] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation commands. The memory may include internal memory and non-volatile memory, and provide instructions and data to the processor.

[0079] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a device for locating the specified user at the logical level. The processor executes the program stored in the memory and is specifically used to execute: Figure 1 The method disclosed in the illustrated embodiment implements the functions and beneficial effects of each method in the previous method embodiments, which will not be described in detail here.

[0080] It should be noted that the stone frog breeding and growth monitoring system based on machine vision provided in the embodiment of the present invention and the stone frog breeding and growth monitoring method based on machine vision provided in the embodiment of the present invention are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned stone frog breeding and growth monitoring method based on machine vision, and has the same or similar beneficial effects, and the repeated parts will not be repeated.

[0081] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for monitoring the growth of stone frogs based on machine vision, characterized in that: The method for monitoring the growth of stone frog breeding based on machine vision comprises: Acquire monitoring images of the breeding environment to be monitored where the stone frog is located; Determine the frog motion feature index of a reference area of ​​the monitoring image of the current frame by using difference pixels in the monitoring image adjacent to the monitoring image of the current frame, wherein the reference area represents a superpixel block where the frog may appear; Determine the frog reference area of ​​the monitoring image of the current frame according to the frog motion characteristic index; Determine the amount of frog enhancement feature information of each gray level of the monitoring image of the current frame according to the edge line in the frog reference area, the main axis direction of the edge line and the pixel points of each gray level in the frog reference area; Determine the grayscale mapping relationship of the grayscale in the monitoring image of the current frame during the histogram equalization process according to the first number of pixels of each grayscale of the monitoring image of the current frame, the second number of all pixels of the monitoring image of the current frame, the third number of grayscales in the frog reference area, and the frog enhanced feature information amount of each grayscale; Performing image enhancement on the monitoring image based on the grayscale mapping relationship of the grayscale to obtain a target image, and using the target image to monitor the growth status of the stone frog; The method of determining the frog motion characteristic index of the reference area of ​​the monitoring image of the current frame by using the difference pixel points in the monitoring image adjacent to the monitoring image of the current frame comprises: Segmenting the monitoring image using a superpixel segmentation algorithm to obtain a plurality of superpixel blocks; Using a frame difference method to obtain difference pixels between a monitoring image of a current frame and a monitoring image of a previous frame adjacent to the monitoring image of the current frame, and marking a super pixel block having the difference pixels, and using the super pixel block having the difference pixels as a reference area of ​​the monitoring image of the current frame; Determine the dynamic similarity features of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame according to the connected domain convex hull of each difference pixel in the reference area of ​​the monitoring image of the current frame and the connected domain convex hull of each difference pixel in the monitoring image of the previous frame; Selecting the connected domain of the difference pixel points with the largest dynamic similarity feature as the reference connected domain of the difference pixel points in the reference area of ​​the monitoring image of the current frame in the monitoring image of the previous frame; Calculating the centroid coordinate variance of the connected domain between the connected domain of the difference pixel point in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of frames before the current frame; Determine the frog motion feature index of the reference area of ​​the monitoring image of the current frame according to the dynamic similarity feature mean between the connected domain of the difference pixel points in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame, the centroid coordinate variance of the connected domain, and the area difference mean; The method of determining the frog reference area of ​​the monitoring image of the current frame according to the frog motion characteristic index includes: Selecting the target reference area with the largest stone frog motion characteristic index; Acquire a target grayscale level in the target reference area; Selecting a super pixel block containing the target gray level from the super pixel blocks of the monitoring image of the current frame as the stone frog reference area; The method of determining the frog enhancement feature information of each gray level of the monitoring image of the current frame according to the edge line in the frog reference area, the main axis direction of the edge line and the pixel points of each gray level in the frog reference area comprises: Acquire an edge line in the stone frog reference area and a main axis direction of the edge line, and determine a reference pixel point of the edge line based on the edge line and the main axis direction; Calculating the distance between the reference pixel point of the current edge line and the reference pixel points of other edge lines, and selecting the minimum distance as the gap distance feature of the current edge line; Determine a stone frog occlusion interference factor of the stone frog reference area according to a fourth number of pixels of the stone frog reference area, a gap distance feature of the edge line, and an image grayscale mean of the edge line; Determine the area of ​​the convex hull of the pixel points of each gray level in the stone frog reference area, and cluster the pixel points of the edge line of the stone frog reference area to obtain cluster clusters; Selecting a cluster with the smallest number of pixels, and obtaining the Euclidean distance between each pixel in the cluster and the nearest pixel in other clusters, and using the Euclidean distance as the enhanced feature distance; Determine the amount of enhanced feature information of the stone frog at each gray level of the current monitoring image according to the stone frog occlusion interference factor, the enhanced feature distance and the area; Determining the grayscale mapping relationship of the grayscale in the monitoring image of the current frame during the histogram equalization process according to the first number of pixels of each grayscale of the monitoring image of the current frame, the second number of all pixels of the monitoring image of the current frame, the third number of grayscales in the frog reference area, and the frog enhanced feature information amount of each grayscale includes: calculating a first difference between the third quantity and a predetermined value; Superimposing the enhanced characteristic information of the stone frog at each gray level to obtain the superimposed enhanced characteristic information; Calculate a second ratio between the amount of stone frog enhancement feature information at each gray level and the amount of superimposed enhancement feature information, and a third ratio between the first number of pixels at each gray level and the second number; superimposing sixth products of the second ratios and the third ratios and multiplying the superimposed products by the first difference to obtain a seventh product; The seventh product is processed by using a rounding function to obtain the gray level mapping relationship.

2. The method for monitoring the growth of stone frog cultivation based on machine vision according to claim 1, characterized in that: Determining the dynamic similarity features of the connected domain of the difference pixels in the reference area of ​​the monitoring image of the current frame according to the connected domain convex hull of each difference pixel in the reference area of ​​the monitoring image of the current frame and the connected domain convex hull of each difference pixel in the monitoring image of the previous frame includes: Calculating the Euclidean distance between the centroid pixel coordinates of the connected domain convex hull of the difference pixel points in the reference area of ​​the monitoring image of the current frame and the centroid pixel coordinates of the connected domain convex hull of the difference pixel points in the monitoring image of the previous frame; Calculating the Hausdorff distance between the convex hull of a connected domain of difference pixels in the reference area of ​​the monitoring image of the current frame and the convex hull of a connected domain of the difference pixels in the monitoring image of the previous frame; The reciprocal of a first product between the Hausdorff distance and the Euclidean distance is determined as the dynamic similarity feature.

3. The method for monitoring the growth of stone frog cultivation based on machine vision according to claim 1, characterized in that: Determining the frog motion feature index of the reference area of ​​the monitoring image of the current frame according to the dynamic similarity feature mean between the connected domain of the difference pixel points in the reference area of ​​the monitoring image of the current frame and the reference connected domain of the monitoring images of the frames before the current frame, the centroid coordinate variance of the connected domain, and the area difference mean includes: Normalizing the mean of the dynamic similarity feature to obtain a normalized value; The second product of the normalized value, the connected domain centroid coordinate variance and the area difference mean is calculated, and the second products are superimposed to obtain the frog motion characteristic index of the reference area of ​​the monitoring image of the current frame.

4. The method for monitoring the growth of stone frog cultivation based on machine vision according to claim 1, characterized in that: Determining the stone frog occlusion interference factor of the stone frog reference area according to the fourth number of pixel points of the stone frog reference area, the gap distance feature of the edge line, and the image grayscale mean of the edge line includes: Calculating the reciprocal of the third product between the image grayscale mean of the edge line and the gap distance feature of the edge line, and superimposing the reciprocals of the third products of each edge line to obtain a superimposed value; Determine a fourth product between the superposition value and the fourth quantity as the stone frog occlusion interference factor of the stone frog reference area.

5. The method for monitoring the growth of stone frog cultivation based on machine vision according to claim 4, characterized in that: The method of determining the amount of enhanced feature information of the frog at each gray level of the current monitoring image according to the frog blocking interference factor, the enhanced feature distance and the area includes: Calculate the fifth product of the area of ​​the convex hull of the pixel points of the gray level in the stone frog reference area and the average value of the enhanced feature distances of each gray level, and the first ratio between the stone frog occlusion interference factor of the stone frog reference area and the fifth product; The first ratios of the gray levels of the stone frog reference areas are superimposed to obtain the stone frog enhanced feature information of each gray level of the current monitoring image.

6. A stone frog breeding growth monitoring system based on machine vision, characterized in that: include: A processor and a memory; wherein the memory is used to store a computer program that can be run on the processor; A processor is used to execute the program stored in the memory to implement the steps of the stone frog breeding growth monitoring method based on machine vision as described in any one of claims 1 to 5.

Images