Intelligent feeding method, device, equipment, medium and product

Abstract

The invention discloses an intelligent feeding method, device and equipment, a medium and a product, and relates to the field of aquaculture, the method comprises the following steps: collecting water quality information of an aquaculture water area and a fish depth image; the water quality information comprises water temperature, dissolved oxygen, pH value and flow velocity; based on the fish depth image, utilizing a computer vision model to estimate the number of fish schools and the weight of a single fish, and determining the total amount of fish biomass; the fish biomass amount and the water quality information are fused, and the primary bait feeding amount is determined; according to the residual bait image after primary feeding, whether secondary feeding is needed or not is judged, the secondary bait feeding amount is determined according to the fish feeding behavior image and the fish feeding audio in the primary feeding process, and real-time, automatic and accurate feeding can be achieved.

Description

Technical Field

[0001] The present application relates to the field of aquaculture, and in particular to an intelligent feeding method, device, equipment, medium and product. Background Art

[0002] Aquaculture is the industry that utilizes available waters for breeding (including planting) to cultivate aquatic economic animals and plants using aquaculture techniques and facilities, tailored to the aquatic environment and living habits of the aquatic species. Fish farming accounts for over 50% of this sector, primarily through manual feeding or feeding machines, and offers significant potential for increasing output value.

[0003] In aquaculture, feeding machines are commonly used to improve feeding accuracy and reduce labor. These simple feeding machines often manually set feed quantities, feed at arbitrary locations, and feed at random times. These machines fail to consider the water quality of the feeding area, the species being fed and the total number of fish being raised, the amount of residual feed, or the feeding behavior and sounds of the fish, indicating their feeding appetite. This can lead to problems such as feed waste, inaccurate feeding times, and a mismatch between feed amounts and fish feed requirements. Long-term, insufficient feeding can lead to low aquaculture yields, which in turn impacts aquaculture profitability. This waste of feed also increases the cost of water purification, hindering the development of the aquaculture industry. Summary of the Invention

[0004] The purpose of this application is to provide an intelligent feeding method, device, equipment, medium and product to solve the problems of waste of bait and low feeding accuracy caused by the extensive feeding method of the current feeding machine.

[0005] To achieve the above objectives, this application provides the following solutions.

[0006] In a first aspect, the present application provides an intelligent feeding method, comprising the following steps.

[0007] Collect water quality information and fish depth images of aquaculture waters; the water quality information includes water temperature, dissolved oxygen, pH value and flow rate.

[0008] Based on the fish depth image, a computer vision model is used to estimate the number of fish and the weight of individual fish to determine the total fish biomass.

[0009] The total amount of fish biomass and the water quality information are integrated to determine the initial feeding amount of bait.

[0010] Based on the residual bait image after the first feeding, it is judged whether a second feeding is needed, and the second feeding amount is determined based on the fish feeding behavior image and fish feeding audio during the first feeding process.

[0011] In a second aspect, the present application provides an intelligent feeding system, comprising the following modules.

[0012] The depth image acquisition module is used to collect depth images of fish in aquaculture waters.

[0013] The water quality information acquisition module is used to collect water quality information of aquaculture waters.

[0014] The above-water image acquisition module is used to collect images of fish feeding behavior and residual bait.

[0015] Audio acquisition module, used to collect fish feeding audio.

[0016] The remote data processing module is used to formulate a precise feeding strategy for fish according to the above-mentioned intelligent feeding method; the precise feeding strategy for fish includes calculating the initial bait feeding amount and calculating the secondary bait feeding amount.

[0017] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any one of the above-described intelligent feeding methods.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any one of the above-described intelligent feeding methods.

[0019] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned intelligent feeding methods.

[0020] According to the specific embodiments provided in this application, this application discloses the following technical effects: This application comprehensively considers multiple factors such as the total amount of fish organisms, water quality information, residual bait, fish feeding behavior and fish feeding audio, and formulates a multi-modal precise fish feeding strategy, namely the calculation process of the initial bait feeding amount, the judgment process and calculation process of the secondary bait feeding amount, arranges the equipment and feeds accurately, realizes real-time, automatic and precise feeding, solves the problems of bait waste in feeding machines, inaccurate feeding time, mismatch between feeding amount and fish bait demand in fishery farming, and avoids the problem of insufficient and wasteful feeding of bait. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a flow chart of the intelligent feeding method provided in one embodiment of the present application.

[0023] Figure 2 A schematic diagram of a depth image acquisition module provided in one embodiment of the present application.

[0024] Figure 3 This is a schematic diagram of a water quality information acquisition module provided in one embodiment of the present application.

[0025] Figure 4 This is a schematic diagram of an underwater image acquisition module provided in one embodiment of the present application.

[0026] Figure 5 This is a schematic diagram of an audio acquisition module provided in one embodiment of the present application.

[0027] Figure 6 A schematic diagram of a remote data processing module provided in one embodiment of the present application.

[0028] Figure 7 This is a schematic diagram of a feeding module provided in one embodiment of the present application.

[0029] Figure 8 This is a schematic diagram of the structure of the intelligent feeding system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0032] The embodiment of the present application provides an intelligent feeding method, which is executed by a computer device, specifically a computer device such as a terminal or a server, and can also be executed by the terminal and the server together. In the embodiment of the present application, Figure 1 As shown, the method includes the following steps.

[0033] S1: Collect water quality information of aquaculture waters and fish depth images; the water quality information includes water temperature, dissolved oxygen, pH value and flow rate.

[0034] S2: Based on the fish depth image, a computer vision model is used to estimate the number of fish and the weight of individual fish to determine the total number of fish.

[0035] S3: Integrate the total amount of fish biomass and the water quality information to determine the initial feeding amount of bait.

[0036] S4: judging whether a second feeding is needed based on the residual bait image after the first feeding, and determining the second feeding amount based on the fish feeding behavior image and fish feeding audio during the first feeding process.

[0037] This application uses the relationship between the water quality information of the aquaculture water area and the optimal water quality environment and the total fish biomass calculated from the fish depth image captured by the underwater binocular camera as input to calculate the first bait feeding amount, that is, the initial bait feeding amount; then, based on the residual bait image captured by the above-water camera, the amount of residual bait is calculated to determine whether a second feeding is needed; based on the collected fish feeding behavior images and the fish feeding audio collected by the hydrophone, the feeding desire of the fish in the aquaculture water area is comprehensively judged, and then the amount of bait required for the second feeding, that is, the secondary bait feeding amount, is calculated.

[0038] The device and method involved in the present application ensure that the farmed fish obtain basic bait without a large amount of bait residue through the first feeding, and meet the amount of bait required by the farmed fish through the second feeding, thereby solving the problem of insufficient bait, ensuring the healthy growth of the farmed fish, and improving the farming efficiency.

[0039] The intelligent feeding method provided in this application is implemented based on a depth image acquisition module, a water quality information acquisition module, an above-water image acquisition module, an audio acquisition module, a remote data processing module and a feeding module.

[0040] like Figure 2 As shown, the depth image acquisition module of the present application may include the following steps.

[0041] In S011, within a breeding cycle, 5% of the farmed fish in the breeding waters are pre-tagged. The tag can be placed on the dorsal fin. This biological tag will not affect the fish's normal physiological behaviors such as feeding, defecation, and swimming, and only serves as a marking.

[0042] In S012, according to the physiological habits of fish, an underwater binocular camera is fixed on one side of the wall of the fish pond so that the lens faces the inside of the fish pond.

[0043] In S013, an underwater binocular camera is used to capture biological images of farmed fish within the aquaculture waters, namely, fish depth images. The captured images primarily include dimensional information such as length, width, and area of the farmed fish carrying biological tags, which can be used to calculate their weight. Furthermore, the depth image acquisition module should also capture fish density information within its monitoring area to determine the fish population.

[0044] In S014, the fish depth image acquired by the underwater binocular camera is wirelessly transmitted to a remote computing unit for processing.

[0045] The fish depth image information obtained in this way is used as a parameter variable for calculating the feeding amount.

[0046] like Figure 3 As shown, the water quality information acquisition module of this application includes the following steps.

[0047] In S021, multiple water quality sensors are used to collect water quality information such as temperature, dissolved oxygen, pH value, and flow rate of the aquaculture waters. Since water quality information varies in different areas of the aquaculture waters, to at least address this issue, underwater sensors should be deployed in as many different locations as possible in the aquaculture waters.

[0048] In S022, the water quality information collected by the water quality sensor is stored locally and backed up.

[0049] In S023, the water quality information is transmitted to a remote data processing module via wireless means.

[0050] Therefore, before calculating the initial feeding amount of bait, the real-time water quality information can be used as a parameter variable to calculate the feeding amount.

[0051] The following explains the impact of water quality information and total fish biomass on feed feeding. According to the feeding patterns of fish, fish have different feeding desires under different water temperature, dissolved oxygen, pH value and flow rate conditions in aquaculture waters, that is, there is a relationship between the required feed and the change of water temperature, dissolved oxygen, pH value and flow rate; in addition, the feed required for one feeding is related to the total biomass in the aquaculture waters. The amount of feed required by fish gradually increases with the number of fish in the aquaculture waters and the biomass obtained by the weight of the fish. Therefore, the initial feed amount in this application needs to consider two variable factors: water quality information and total fish biomass.

[0052] This application requires a second feeding. The amount of the second feeding is determined based on the residual bait situation in the aquaculture waters after the first feeding and the real-time feeding desire of the farmed fish. Therefore, the steps for obtaining these two types of information are as follows.

[0053] like Figure 4 As shown, the water image acquisition module of the present application includes the following steps.

[0054] In S031, to simplify the process of acquiring images of the remaining bait, based on the design of the track-type intelligent baitcasting machine, this application adopts the method of fixing the water camera directly below the baitcasting machine or on the outer wall of the baitcasting machine. In this way, the water camera can capture 360° images of the water surface and can move with the baitcasting machine.

[0055] In S032 and S033, a camera fixed directly below the bait throwing machine collects images of the remaining bait and the feeding behavior of the fish school during the initial feeding process.

[0056] In S034, the image information collected by the water camera module is transmitted to the remote data processing module via wireless means.

[0057] Therefore, before calculating the second feeding amount, the real-time image of the residual bait and the image of the fish feeding behavior can be used as two parameter variables for judging and calculating the second feeding amount.

[0058] like Figure 5 As shown, the audio acquisition module of the present application includes the following steps.

[0059] When fish eat, the teeth in their jaws rub against the bait, making a sound. Therefore, collecting feeding audio can also be used as a criterion for judging the fish's desire to eat.

[0060] In S041, the hydrophone device is placed 20 cm to 50 cm underwater and above the bottom of the pond, directly below the feeding center of the bait feeder.

[0061] In S042, the hydrophone device is turned on during the feeding process to collect the sound data of the fish feeding. During the feeding sound data collection process, the hydrophone device also collects the sound of the bait being spread. The spread sound is noise that needs to be identified and removed.

[0062] In S043, the audio of the school of fish feeding is transmitted wirelessly to a remote data processing module.

[0063] Therefore, before calculating the second feeding amount, the fish feeding sound data obtained in real time can be used as a parameter variable for calculating the second feeding amount.

[0064] In an exemplary embodiment, S2 may be replaced by the following steps.

[0065] Identify fish images with biological tags in the fish depth image.

[0066] The weight of the tagged fish with the biological tags in the fish image is analyzed using a computer vision model.

[0067] Based on the weight of the tagged fish, the average weight of all tagged fish was determined.

[0068] The number of fish in the aquaculture waters is determined based on the proportion of labeled fish to all fish in the aquaculture waters.

[0069] The total amount of fish in the aquaculture waters is determined based on the average weight and the number of the fish.

[0070] In an exemplary embodiment, the process of analyzing the weight of the tagged fish with biological tags in the fish image using a computer vision model may be replaced by the following steps.

[0071] The body length of the tagged fish is calculated using a computer vision model and key point detection technology.

[0072] The weight of the tagged fish with biological tags was calculated based on the species of the tagged fish and the relationship between body weight and body length during the growth period of the species.

[0073] In an exemplary embodiment, S3 may be replaced by the following steps.

[0074] Obtain fish species from aquaculture waters.

[0075] The water quality information is compared with the most suitable water quality information for breeding the fish species to determine the water quality feeding coefficient.

[0076] Determine the bait coefficient based on the bait type and fish growth cycle.

[0077] The preset required bait feeding amount is determined according to the water quality feeding coefficient, the bait coefficient and the total amount of fish organisms.

[0078] The initial bait feeding amount is determined according to the preset required bait feeding amount.

[0079] In an exemplary embodiment, S4 may be replaced by the following steps.

[0080] The amount of residual bait at the set time is determined based on the residual bait image after the first feeding at the set time.

[0081] Determine whether the amount of residual bait at the set time is greater than the residual bait amount threshold; if so, determine not to perform secondary feeding; if not, based on multimodal fusion technology and deep learning models, fuse the fish feeding behavior images and fish feeding audio during the initial feeding process to determine the fish's feeding desire level during the initial feeding process; the feeding desire levels include strong, medium, weak and none.

[0082] The secondary feed amount is determined according to the feeding desire level and the initial feed amount.

[0083] like Figure 6As shown, the remote data processing, bait amount calculation and real-time control method of feeding by a bait feeding machine of the present application includes the following steps.

[0084] After the remote data processing module receives the water quality information, image and audio data (if any), it is passed to the computer, and the computer calls the algorithm model to perform calculations on this information. Due to the size of the aquaculture water area and the computing power limitations of the computer used, the number of computers is not limited to one or several, and this application does not impose any restrictions.

[0085] In S051, the remote computer receives data from each data acquisition module.

[0086] In S052, the computer receives a depth image of fish in the aquaculture waters obtained by an underwater binocular camera, identifies and captures images of fish with biological tags, analyzes the obtained fish images by a debugged and well-arranged algorithm model, calculates information such as the size and weight of the tagged fish through a program, and then calculates the average weight of the tagged fish. The number of fish in the aquaculture waters is then calculated based on the ratio of the tagged fish to all the fish in the aquaculture waters, and finally the overall biomass is calculated. The fish biomass calculation method is as follows.

[0087] Using key point detection technology, the body length of the tagged fish is calculated, and the body length of different tagged fish is defined as Ln. Then, based on the relationship between weight and body length of different fish species and growth periods, the weight of the tagged fish is calculated, and the weight of different tagged fish is defined as M. n Calculate the average weight of tagged fish Represents the average weight of farmed fish. The number of farmed fish N is obtained by analyzing the density information of the depth image.

[0088] In S053, the total biomass of farmed fish is defined as B, and the total biomass of the farmed fish pond is obtained. .

[0089] In S054, the computer receives the water quality environmental parameters in the aquaculture water area obtained by various sensors and calculates the water quality feeding coefficient.

[0090] The water quality feeding coefficient is defined as µ. The collected water quality information of the aquaculture pond is fuzzified. According to the relationship with the parameters of the most suitable aquaculture environment for the fish species, the water quality feeding coefficient µ under this environment is obtained.

[0091] For example, the most suitable water temperature for breeding of spotted sea bream is 25℃-30℃, the most suitable water temperature for breeding is 15℃-32℃, and the rest are stress water temperatures. The coefficient of the effect of temperature on feeding amount is defined as µ t， If this is the optimum water temperature, then µ t =1; if the water temperature is suitable for aquaculture but not the optimal temperature, then µ tis 0.90; if it is a stress water temperature, then µ t Therefore, the respective coefficients of temperature, dissolved oxygen, pH value and flow rate on different fish species are determined, and the coefficients of the influence of temperature, dissolved oxygen, pH value and flow rate on feeding can be multiplied to obtain the water quality feeding coefficient µ.

[0092] In S055, the bait coefficient of the bait type used is defined as F cr The feed coefficient is the ratio of feed amount to biomass, and the preset feed amount is defined as F q .

[0093] The preset feeding amount F can be calculated q =µ×F cr ×B.

[0094] Since feeding only once often results in waste or insufficient bait, precise feeding is required. Therefore, this application designs the initial bait feeding amount to be less than the preset feeding amount.

[0095] In S056, the initial bait feeding amount F is set. q1 =80%×F q .

[0096] In S057, the time when the farmed fish has completed 90% of its first feeding is defined as T1 (obtained from the statistics of the farmed fish feeding process). This T1 time is the set time. The remote computer uses a counting model to count the residual bait images obtained at T1 time after the first feeding. The trained YOLO model or other counting models can be used to count the residual bait to obtain the amount of residual bait at the current moment.

[0097] In S058, the residual bait quantity threshold is set to G, and the bait quantity at time T1 calculated using the computer vision model is defined as G1. The following feeding rules can be set: if at time T1, G1>G, it means that there is a lot of residual bait left in this feeding, and the second feeding is not performed; if G1≤G, it means that the fish are eating quickly and have a strong appetite, and the second feeding is performed.

[0098] In S059, a remote computer fuses the images and sound information of fish feeding behavior in the aquaculture waters through multimodal fusion and deep learning models. Based on the fish feeding characteristics and behavioral influence rules, the feeding desire of the fish during the feeding process at time T1 is judged. The feeding desire can be divided into four levels: strong, medium, weak, and none.

[0099] In S060, by referring to the fish feeding desire given by the multimodal fusion algorithm, the second feeding rule is set as follows: the bait amount is F qIf the feeding desire is strong at time T1, the second feeding amount is F q2 =30%×F q If the feeding desire at time T1 is medium, the second feeding amount is F q2 =20%×F q If the feeding desire at time T1 is weak, the second feeding amount is F q2 =10%×F q If the feeding desire is zero at time T1, the secondary feeding amount F q2 =0, that is, no second feeding is performed in this state.

[0100] In S061, a decision is made by computer comprehensive calculation to determine whether to feed for the second time. If feeding for the second time is carried out, the amount of feed fed for the second time is F q2 .

[0101] In S062, the remote data processing module transmits the control decisions of the first feeding and the second feeding to the feeding module via the wireless network.

[0102] like Figure 7 The feeding module of this application includes the following steps.

[0103] In S071, the bait throwing machine receives the control decision of the remote data processing module.

[0104] In S072, the feeding machine completes the loading work. During the initial feeding process, the feeding machine should carry 120% of the preset feeding amount at one time to cope with the second feeding and reduce the additional loading work in the second feeding operation.

[0105] In one embodiment, the amount of bait fed can be verified by subtracting the amount of bait remaining at the end of feeding from the amount of bait loaded before the feeding machine starts feeding, that is, the feeding machine can be equipped with a pressure sensor to automatically calculate the feeding amount.

[0106] In S072, the feeding machine moves to the designated feeding location. As the controllable feeding machine delivers bait to each aquaculture area, it can receive control signals and move to the corresponding location in the aquaculture waters. The feeding machine can be mounted on a feeding track, which should be constructed based on the actual conditions of the aquaculture waters. The feeding machine can move freely on the feeding track or on an unfixed track.

[0107] In S073 and S074, the feeding machine receives the control signal from the remote data processing module and performs the first feeding and the second feeding according to the control signal.

[0108] In S075, the baitcasting machine completes the feeding task and returns to its original position to facilitate the next filling and feeding.

[0109] In addition, since only one feeding machine is used to feed different areas of the aquaculture waters, although the investment cost of the feeding machine can be reduced, if the aquaculture area is large or the total number of aquaculture areas is too large, using only one feeding machine to feed all aquaculture areas may result in excessive feeding time, which affects the feeding time and causes an error between the feeding amount and the required bait amount. In addition, the workload of a single feeding machine is heavy, which can easily lead to an increased probability of damage. Therefore, to at least solve this problem, in one embodiment, the feeding system can include multiple feeding machines. When the area of a single aquaculture water is small, each feeding machine can be responsible for one aquaculture area, and multiple feeding machines can feed multiple aquaculture areas simultaneously. When the area of aquaculture water is large, multiple feeding machines can be responsible for one aquaculture area.

[0110] This application focuses on the proposal of precise feeding strategies for fish and the construction of feeding systems. It does not provide an overview of the mechanical structure of the feeding machine. The mechanical structure of the feeding machine should meet the requirements of the aquaculture waters. For example, factory aquaculture may require the feeding machine to have highly flexible robotic arms and automatic bait filling functions; pond aquaculture may require the feeding machine to have the ability to move on tracks; cage aquaculture may require the feeding machine to be connected to the cage feed port, etc.

[0111] Specific adjustments can be made to the different application environments of this application.

[0112] For example: in factory farming or cage farming, underwater binocular cameras can be fixed at different depths to collect data from as large a farming area as possible; in pond farming, gates can be set up and underwater binocular cameras can be placed on both sides of the gates.

[0113] For smaller breeding areas, wireless transmission and receiving equipment can be replaced by wired equipment, taking economic conditions into consideration and ensuring functionality.

[0114] For example, in cases where the aquaculture water area is too large, such as in offshore cage aquaculture, the bait thrower is difficult to be arranged on a fixed track or moved automatically. While ensuring that the wireless signal can be received and the program can be encoded, it can be built on a bait throwing boat to form a ship-mounted bait thrower.

[0115] While ensuring that the system configuration architecture of this application remains unchanged, the calculation models used can be replaced as the models are improved and developed.

[0116] When the demand for computer computing power is high and the timeliness of the feeding process is high, the remote data processing module can also be placed on the cloud service platform for information processing and calculation.

[0117] This application can effectively use computer vision technology to process the depth image information obtained by the underwater binocular camera, and wirelessly transmit the water quality information obtained by the sensor to realize the fishery Internet of Things. For the initial feeding amount of bait, not only the aquaculture species and biomass of the aquaculture waters are considered, but also the influence of various water quality factors on the feeding amount is combined to ensure the rationality of the initial feeding amount of bait and reduce the occurrence of waste of residual bait. The system collects the residual bait image obtained after the first feeding, the fish feeding behavior image and the fish feeding audio data during the first feeding process, and accurately determines the bait supplementary feeding amount for the second feeding through multimodal information fusion of the remote computing unit to ensure that the aquaculture objects have sufficient food intake, thereby increasing the maximum output and improving the aquaculture income.

[0118] Based on the same inventive concept, embodiments of the present application also provide an intelligent feeding system for implementing the aforementioned intelligent feeding method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more intelligent feeding system embodiments provided below can be found in the above-described limitations of the intelligent feeding method and will not be further elaborated here.

[0119] In an exemplary embodiment, Figure 8 As shown, the present application provides an intelligent feeding system including the following modules.

[0120] The depth image acquisition module is used to collect depth images of fish in aquaculture waters.

[0121] The water quality information acquisition module is used to collect water quality information of aquaculture waters.

[0122] The above-water image acquisition module is used to collect images of fish feeding behavior and residual bait.

[0123] Audio acquisition module, used to collect fish feeding audio.

[0124] The remote data processing module is used to formulate a precise feeding strategy for fish according to the above-mentioned intelligent feeding method; the precise feeding strategy for fish includes calculating the initial bait feeding amount and calculating the secondary bait feeding amount.

[0125] In practical applications, the depth image acquisition module includes an underwater binocular depth camera for acquiring depth images of fish.

[0126] The water quality information acquisition module includes a water temperature sensor, a dissolved oxygen sensor, a pH sensor and a flow rate sensor for collecting water quality information.

[0127] The above-water image acquisition module includes an industrial camera for acquiring images of changes in residual bait and images of fish behavior.

[0128] The audio acquisition module includes a hydrophone device that collects the sounds of fish feeding.

[0129] The remote data processing module includes a remote computing device that processes fish images, water quality environment parameters, residual bait images, fish behavior images and fish feeding audio.

[0130] Various water quality sensors are placed underwater to measure and quantify water quality parameters in real time, including temperature, dissolved oxygen, pH value, water flow rate and other water quality parameter information; the underwater binocular camera used collects images of fish schools in the breeding area; the industrial camera used collects images of the feeding behavior of farmed fish and the amount of remaining bait in the feeding area; the hydrophone equipment used collects audio data of fish feeding in the breeding area.

[0131] With the diversification of information sources, fish farming environment parameters and fish biomass information are considered as the first feeding strategy factors, and participate in the feeding system's calculation of the preset feeding amount and the initial feed feeding amount.

[0132] Residual bait images, fish feeding behavior and fish feeding audio data are considered as the second feeding strategy factors, participating in the dynamic adjustment of the feeding system and deciding whether to feed the fish for the second time.

[0133] In practical applications, a fuzzy inference system is used to determine the relationship between water quality parameters and the optimal water quality environment for fish. The number and weight of fish are calculated based on the biological images of fish trained by a deep learning system. The water quality parameters of the fish farming environment and the total biological amount of farmed fish are input into the precise feeding system to obtain the preset feeding amount and the first feeding amount.

[0134] A computer vision model is used to count the remaining bait, and the amount of remaining bait is used as the basis for making a second feeding decision.

[0135] Utilizing multimodal fusion technology, a deep learning model is trained and classified using labeled video and audio clips that indicate fish feeding desire (strong, medium, weak, or absent). This method is used to determine fish feeding desire, gain a deeper understanding of their feeding habits and behavioral patterns, and ultimately determine the second feeding amount. This deep learning model can be a convolutional neural network model.

[0136] The present application also includes: a feeding machine device for feeding bait to fish in aquaculture waters.

[0137] This application uses computer vision technology to count and estimate the weight of fish schools, and then estimates the biomass of fish schools, fusing the total fish biomass with the water quality information collected by water quality sensors; uses computer vision to count the amount of residual bait, and uses multimodal algorithms to fuse fish behavior images and feeding audio, to build an intelligent feeding integration system based on the fish farming environment, fish biomass and residual bait, and fish feeding behavior.

[0138] Different feeding strategies are formulated based on the water quality environment requirements corresponding to the species, size and number of farmed fish, and comprehensive water quality parameters and fish habit information.

[0139] In actual applications, a computer control interface built according to a programming language controls the feeding machine through wireless transmission to perform feeding operations.

[0140] In practical application, this application includes the following steps.

[0141] The water quality information acquisition process obtains the water quality parameters in the fish farming pond through multiple water quality sensors.

[0142] The water quality parameter processing step processes water quality parameter data through a fuzzy inference system.

[0143] The depth image acquisition process acquires image data by using an underwater binocular camera set in the fish pond.

[0144] Deep image processing process, processing fish image data through computer vision and deep learning methods.

[0145] The audio data acquisition step is to acquire the audio data of fish feeding by using a hydrophone device placed in the fish farming pond.

[0146] Audio processing step, processing fish feeding audio data through audio noise reduction and filtering.

[0147] The multimodal data fusion process uses multimodal fusion technology and deep learning technology to fuse and process the acquired image and audio data.

[0148] The feeding process involves feeding fish in aquaculture factories using feeding machines.

[0149] Feedback adjustment process, through the control principle, timely adjusts the feeding process within a feeding cycle.

[0150] Stop feeding process, feeding cycle ends, stop feeding, stop feeding in emergency situation.

[0151] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output (I / O) interface, and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store intelligent feeding data. The I / O interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements an intelligent feeding method.

[0152] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the above method when executing the computer program.

[0153] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the above method when executed by a processor.

[0154] In an exemplary embodiment, a computer program product is provided, including a computer program, which implements the above method when executed by a processor.

[0155] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0156] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0157] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0158] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0159] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An intelligent feeding method, characterized in that: The intelligent feeding method comprises: Collecting water quality information and fish depth images of aquaculture waters; the water quality information includes water temperature, dissolved oxygen, pH value and flow rate; Based on the fish depth image, using a computer vision model to estimate the number of fish and the weight of individual fish to determine the total fish biomass; The total amount of fish and the water quality information are integrated to determine the initial feeding amount of bait; Based on the residual bait image after the first feeding, it is judged whether a second feeding is needed, and the second feeding amount is determined based on the fish feeding behavior image and fish feeding audio during the first feeding process.

2. The intelligent feeding method according to claim 1, characterized in that: Based on the fish depth image, a computer vision model is used to estimate the number of fish and the weight of individual fish to determine the total fish biomass, specifically including: Identifying fish images with biological tags in the fish depth images; Analyzing the weight of the tagged fish with the biological tags in the fish image using a computer vision model; Determine the average weight of all tagged fish based on the weight of the tagged fish; Determine the number of fish in the aquaculture waters based on the proportion of labeled fish to all fish in the aquaculture waters; The total amount of fish in the aquaculture waters is determined based on the average weight and the number of the fish.

3. The intelligent feeding method according to claim 2, characterized in that: The weight of the fish tagged with the biological tag in the fish image is analyzed using a computer vision model, specifically including: Calculating the body length of the tagged fish using a computer vision model and key point detection technology; The weight of the tagged fish with biological tags was calculated based on the species of the tagged fish and the relationship between body weight and body length during the growth period of the species.

4. The intelligent feeding method according to claim 1, characterized in that: The total amount of fish and the water quality information are integrated to determine the initial feeding amount of bait, specifically including: Obtain fish species from aquaculture waters; Comparing the water quality information with the optimal water quality information for aquaculture of the fish species to determine a water quality feeding coefficient; Determine the bait coefficient based on bait species and fish growth cycle; Determine the preset required bait feeding amount according to the water quality feeding coefficient, the bait coefficient and the total amount of fish; The initial bait feeding amount is determined according to the preset required bait feeding amount.

5. The intelligent feeding method according to claim 1, characterized in that: Based on the residual bait image after the initial feeding, it is determined whether a second feeding is needed. Based on the fish feeding behavior image and fish feeding audio during the initial feeding process, the second feeding amount is determined. Specifically, it includes: Determine the amount of residual bait at the set time according to the residual bait image after the first feeding at the set time; Determining whether the amount of residual bait at the set time is greater than a threshold value of the amount of residual bait; If so, make sure not to feed twice; If not, the fish's feeding desire level during the first feeding process is determined by fusing the fish's feeding behavior images and audio of the fish's feeding during the first feeding process based on multimodal fusion technology and a deep learning model; the feeding desire level includes strong, medium, weak, and none; The secondary feed amount is determined according to the feeding desire level and the initial feed amount.

6. The intelligent feeding method according to claim 1, characterized in that: Before each feeding, the amount of bait in the feeding machine is calculated automatically based on the pressure data collected by the pressure sensor.

7. An intelligent feeding system, characterized in that: The intelligent feeding system comprises: Depth image acquisition module, used to collect depth images of fish in aquaculture waters; Water quality information acquisition module, used to collect water quality information of aquaculture waters; An above-water image acquisition module is used to collect images of fish feeding behavior and residual bait; Audio acquisition module, used to collect audio of fish feeding; A remote data processing module is used to formulate a precise feeding strategy for fish according to the intelligent feeding method according to claims 1-6; the precise feeding strategy for fish includes calculating the initial bait feeding amount and calculating the secondary bait feeding amount.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent feeding method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intelligent feeding method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the intelligent feeding method according to any one of claims 1 to 6 is implemented.

Images