Meat bird precise feeding method and device based on feed intake

Abstract

The application discloses a kind of meat fowl precision feeding method and device based on feed intake, and the feed intake of different day-old meat fowl is collected, and the feed intake prediction model of single meat fowl based on day-old is established according to the feed intake data;State tracking is carried out to the meat fowl going to the feeding area, and the number of meat fowl going to the feeding area is determined;Then, according to the feed intake prediction model and the number of meat fowl in the feeding area, the predicted feed intake of meat fowl in the feeding area is obtained;Finally, the feeding amount is calculated by the predicted feed intake and the remaining amount of feed, and the calculated feeding amount is used to realize precision feeding.This application realizes precision feeding based on the growth demand of meat fowl, and reduces the waste of feed.

Description

Technical Field

[0001] This invention belongs to the field of intelligent livestock breeding equipment technology, specifically relating to a method and device for precise feeding of poultry based on feed intake. Background Technology

[0002] Precision feeding, as a management method that has emerged in large-scale farming in recent years, plays an important role in improving production efficiency and reducing production costs.

[0003] In large-scale farming, farming methods are further divided into wire mesh floor rearing, ground floor rearing, and cage rearing. In wire mesh floor rearing and cage rearing, poultry are confined to a designated area with a relatively fixed feeding area. However, ground floor rearing allows poultry to feed randomly in both location and time. Since ground floor rearing involves free-range poultry on the ground, feeding facilities should ideally prevent fighting, trampling, and pecking, and existing equipment does not take these behaviors into account.

[0004] Traditional free-range poultry feeding methods include manual and mechanical methods. Manual feeding involves adding feed to the trough by hand, which is labor-intensive. Mechanical feeding uses a conveyor belt to transport feed from the hopper to the trough. However, after feeding, a large amount of feed often remains in the trough. If this untreated feed is not cleaned for a long time, it can pollute the poultry's production environment and even compromise their health. Although mechanical feeding saves labor compared to manual feeding, it is difficult to accurately add feed based on the amount of remaining feed in the trough, making it impossible to ensure that the remaining feed is within a reasonable range. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and apparatus for precise feeding of poultry based on feed intake.

[0006] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0007] A precision feeding method for poultry based on feed intake:

[0008] The feed intake of individual poultry was determined by feeding them separately. Feed intake data was collected from poultry of different ages, and an age-based feed intake prediction model for individual poultry was established based on the feed intake data.

[0009] The status of poultry heading to the feeding area is tracked to determine the number of poultry heading to the feeding area.

[0010] Based on the feed intake prediction model and the number of poultry going to the feeding area, the predicted feed intake of poultry in the feeding area is obtained.

[0011] The feeding amount is calculated based on the predicted feed intake and the amount of feed remaining, and precise feeding is achieved based on the calculated feeding amount.

[0012] Furthermore, the process of the feed intake prediction model is as follows:

[0013] The feed intake data is obtained by subtracting the amount of feed remaining from the amount of feed fed.

[0014] First, k-fold cross-validation was used to classify feed intake data. Then, forward stepwise regression was used to establish an age-based feed intake prediction model for a single poultry. Finally, grid search and k-fold cross-validation were used to adjust the hyperparameters.

[0015] Furthermore, the state tracking and determination of the number of poultry heading to the feeding area is achieved using the StrongSORT algorithm.

[0016] Furthermore, the predicted feed intake of poultry within the feeding area is:

[0017] C qy =C yc ×N

[0018] Where: C qy It is the predicted feed intake of poultry within the foraging area; C yc It is the predicted feed intake per bird obtained from the feed intake prediction model; N is the number of poultry going to the feeding area.

[0019] Furthermore, after feeding is completed, the feed intake prediction model is updated in real time, specifically as follows:

[0020] After feeding, the amount of feed remaining in the feed troughs in the feeding area is measured. The feed intake for this feeding is obtained by combining the amount of feed precisely fed before feeding. The average feed intake is then obtained based on the number of poultry in the feeding area. A threshold is set for the average feed intake. If it is not lower than the threshold, the requirements for precise feeding are met. If it is lower than the threshold, the feed intake data for this feeding is recorded and used together with the feed intake data collected in the poultry house before feeding to build a prediction model. The next feeding is based on the new prediction model for precise feeding, and the cycle is repeated.

[0021] Furthermore, the process for obtaining the remaining feed amount is as follows:

[0022] The remaining feed is detected by ultrasonic waves, and the returned digital signal is processed to obtain a two-dimensional image. Then, the two-dimensional image is reconstructed into a three-dimensional model. The volume of the three-dimensional model is then calculated, and the volume is multiplied by the density of the feed to determine the amount of feed remaining.

[0023] Furthermore, the remaining feed amount is determined. If the remaining feed amount is zero, the predicted feed intake in the feeding area is multiplied by k, which is the amount of feed for this precise feeding. After the poultry finishes feeding, the remaining feed amount is detected again. If the remaining feed amount is still zero, the next k value is selected and the cycle continues until the remaining feed amount is not zero. If the remaining feed amount is not zero, the predicted feed intake is subtracted from the remaining feed amount, which is the amount of feed for this precise feeding, where k = 0.1F + 1.1, and F is the number of cycles.

[0024] A precision feeding device for poultry based on feed intake includes:

[0025] Tracks are installed inside the poultry house;

[0026] The hopper is located above the track and can run along the track;

[0027] Solenoid valve, located inside the conveying pipe of the hopper;

[0028] The ultrasonic ranging module, located in the forward direction of the hopper, is used to detect the amount of feed remaining.

[0029] The camera, located in the direction of the feed hopper's movement, is used to detect the number of poultry in the feeding area;

[0030] The material feeding pipe is located below the track.

[0031] The feed trough is placed at the lower end of the feed pipe;

[0032] The control module is used to control the movement of the hopper and to predict and accurately feed the poultry in the feeding area.

[0033] In the above technical solution, a plate with wheels is provided below the hopper, and a battery is provided on the plate. An ultrasonic ranging module, a control module and a camera are also installed on the plate.

[0034] The above technical solution also includes an auxiliary wheel, which is connected to the wheel and has its tail end embedded in the track.

[0035] The above technical solution also includes an infrared transmitter, which is installed inside the track at the top of the feeding pipe.

[0036] The beneficial effects of this invention are as follows: By placing the feeding device at the top, it reduces interference from poultry with the facility; it collects feed intake data from individually fed poultry, establishing an age-based feed intake prediction model for each bird; it tracks the status of poultry heading to the feeding area, determining the number of poultry going there; based on the feed intake prediction model and the number of poultry in the feeding area, it obtains the predicted feed intake for the poultry in that area; it calculates the feeding amount from the predicted feed intake and the remaining feed, achieving precise feeding based on the calculated amount; after feeding, the feed intake prediction model is updated in real time based on the remaining feed. This invention provides precise feeding based on the growth needs of poultry, reducing feed waste and avoiding environmental pollution, thus not affecting the health of the poultry. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the overall structure of the feeding device described in this invention;

[0038] Figure 2 This is a flowchart illustrating the precise material feeding process described in this invention;

[0039] Figure 3 This is a flowchart illustrating the real-time update of the prediction model described in this invention.

[0040] In the diagram: 1. Hopper; 2. Solenoid valve; 3. Infrared transmitter; 4. Auxiliary wheel; 5. Feeding pipe; 6. Feed trough; 7. Ultrasonic ranging module; 8. Battery; 9. Control module; 10. Camera; 11. Laser rangefinder; 12. Track; 13. Board. Detailed Implementation

[0041] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0042] like Figure 1 As shown, the present invention provides a precision feeding device for poultry based on feed intake, comprising a hopper 1, a solenoid valve 2, an infrared transmitter 3, an auxiliary wheel 4, a feeding pipe 5, a feed trough 6, an ultrasonic ranging module 7, a storage battery 8, a control module 9, a camera 10, a laser rangefinder 11, a track 12, and a plate 13.

[0043] The feeding device in this embodiment adopts a suspended rail installation method. A track 12 is installed inside the poultry house. The track 12 is made of I-beam steel and is filled with sponge as a shock-absorbing material to reduce the noise generated when the feeding device moves on the track. The feeding device runs on the track. Since the track 12 is long and straight, there is no need to consider factors such as turning, so four-wheel drive is used. The auxiliary wheel 4 of the feeding device is intermittently pressureless lubricated with the track. Lubricant is applied manually to the edge of the auxiliary wheel 4 to reduce friction between the auxiliary wheel 4 and the track. The thickness of the auxiliary wheel 4 is the same as the thickness of the recess of the I-beam steel track, so that it is exactly in the recess of the I-beam steel. The radius of the auxiliary wheel 4 is 1.5 times the flange width of the I-beam steel track.

[0044] In this embodiment, the hopper 1 is located above the track, and its height is slightly less than the height of the feed hopper's discharge port, allowing it to be fed directly from below the hopper. Below the hopper 1 is a plate 13 with four wheels mounted on it. These wheels travel on the track 12 and are connected by connecting rods; auxiliary wheels 4 are connected to both sides of the connecting rods. A battery 8 (for power supply) is mounted on the plate 13. The battery 8 is a fixed lead-acid battery, which can be replaced with a fully charged battery when depleted. The control module 9 is located next to the battery 8 and includes a microcontroller, a WiFi module, a Bluetooth module, and a switching power supply. A laser rangefinder 11 is located in the forward direction of the feeding device and is used to determine the distance between the hopper 1 and the wall, preventing collisions. A solenoid valve 2 is located inside the feed hopper 1's delivery pipe, and its opening size can be adjusted according to the predicted feed intake. An ultrasonic ranging module 7 is also located in the forward direction of the feeding device, below the plate 13, and is mainly used to detect the remaining feed in the feed trough 6. Camera 10 is located next to ultrasonic ranging module 7 and is used to detect the number of poultry in the feeding area.

[0045] A gap is left between the feeding device and the I-beam rail 12. M circular holes are drilled in the I-beam rail 12 to accommodate the feeding pipe 5. The feeding pipe 5 is placed into these holes and welded to the I-beam rail 12. The feed trough 6 is placed directly below the feeding pipe 5, at the same height as the poultry's feeding height. The infrared emitter 3 is located inside the I-beam rail 12 at the top of the feeding pipe 5. It uses the infrared light emitted to determine whether it is blocked by the bracket of the auxiliary wheel 4 to accurately position the feed hopper 1, ensuring that the feed hopper 1's conveying pipe is aligned with the feeding pipe 5.

[0046] The working principle of the poultry precision feeding device based on feed intake of this invention is as follows: When the feeding device is not working, the feed hopper 1 is located at the starting position on one side of the track 12, according to... Figure 1The direction shown is the rightmost end; camera 10 collects video of the poultry house area and sends it to control module 9. Control module 9 predicts the feed intake of poultry going to the first feeding area. Ultrasonic ranging module 7 collects the remaining feed value in feed trough 6 in the first feeding area and sends it to control module 9 to determine the remaining feed in feed trough 6, and then obtains the precise feed amount in feed trough 6 in the first feeding area; based on the data fed back by infrared transmitter 3, when hopper 1 reaches directly above the feed pipe 5 in the first feeding area, control solenoid valve 2 to open and put the precise feed amount into feed trough 6 in the first feeding area; and so on, until the feed amount in feed trough 6 in the Mth feeding area is completed.

[0047] Figure 2 It is a flowchart of precise material feeding, which includes the feeding process and the movement process.

[0048] During the feeding process, the number of poultry in the poultry house area is first detected. Since poultry feeding is random, the StrongSORT algorithm is used to track the state of poultry going to the feeding area, determine the number of poultry going to the feeding area, and finally predict the amount of feed consumed by poultry.

[0049] The process of using a neural network to detect the number of poultry is as follows: Before the feeding device is put into use, camera 10 collects video information of poultry moving to the feeding area and sends it as historical data to control module 9. Control module 9 decomposes the video information into images, labels the images to form a tag file, and inputs the dataset formed by the tag file and images into the neural network for model training. When in use, the video collected in real time by camera 10 is input into the trained neural network, and the number of poultry is output.

[0050] The StrongSORT algorithm is used to track the state of poultry heading to the foraging area and determine the number of poultry heading to the foraging area. The specific process is as follows: First, video of poultry heading to the foraging area is acquired; second, a StrongSORT target detector is used to detect targets in each video frame; third, a high-confidence threshold and a low-confidence threshold are set, and the detected targets of the StrongSORT target detector are filtered using the high-confidence threshold and the low-confidence threshold; finally, the detected targets in the current video frame with a confidence level greater than the low-confidence threshold are compared with the tracked targets in the previous frame using an IoU operation to obtain l. iou And match the detected target that satisfies the following formula with the tracked target of the previous frame;

[0051] -ln(1-0.8×l iou )-ln(1-score)≥1.2

[0052] Where: score is the confidence level of the detected target in the current video frame image.

[0053] Finally, the matching results are fused with the detected targets filtered using a high confidence threshold to obtain the tracking targets in the current video frame image and determine the number of poultry heading to the feeding area.

[0054] Based on the feed intake prediction model and the number of poultry moving into the feeding area, the predicted feed intake of poultry in the feeding area is obtained:

[0055] C qy =C yc ×N

[0056] Where: C qy It is the predicted feed intake of poultry within the foraging area; C yc It is the predicted feed intake per bird obtained from the feed intake prediction model; N is the number of poultry going to the feeding area.

[0057] The process of establishing the feed intake prediction model is as follows: First, a batch of broiler poultry in the brooding period is selected, and each poultry is fed individually. Before feeding each day, the feed to be fed is weighed and recorded as the feed input W1. After feeding, the remaining feed in the feed trough for individual feeding is weighed and recorded as the remaining feed W2. The feed intake of a single poultry is W3 = W1 - W2. The poultry are fed on time every day, and feed intake data are collected at different ages of the poultry. The collected data are then processed to prepare for the next step of building the prediction model.

[0058] In this embodiment, the classification method used for the prediction model is k-fold cross-validation. First, the obtained feed intake data is divided into k parts (k = 5N; N ≥ 1). Then, one part of the k parts is taken as the test set without repetition each time, and the remaining k-1 parts are used as the training set to train the prediction model. This will result in k evaluation models. Finally, the average performance of the k evaluations obtained from the above two steps is taken as the final evaluation result.

[0059] This embodiment uses a forward stepwise regression algorithm to process feed intake data. First, all weights are set to 1. Then, the feed intake data is standardized so that its distribution satisfies zero mean and unit variance. Next, iteration is performed. The iteration process begins by setting the current minimum error to positive infinity. For each feed intake data feature, the weights are increased or decreased, changing the coefficients of a linear regression equation to obtain a new weight. The error between the feed intake data obtained from the linear regression equation under the new weight and the actual feed intake data is calculated. If the error is less than the current minimum error, the optimal weight is set to equal the current weight, until the error is minimized. This algorithm constructs a feed intake prediction model every 100 iterations. Ten-fold cross-validation is used to compare these feed intake prediction models, and the model with the smallest error is ultimately selected.

[0060] Before building the prediction model, hyperparameters such as the learning rate were set, so they need to be adjusted. This embodiment uses a combination of grid search and k-fold cross-validation for hyperparameter adjustment. First, the evaluation index is determined. Then, for each set of hyperparameter values, the average performance of k evaluations is obtained using cross-validation on the training set. Finally, the performance of each combination of hyperparameter values ​​is compared to find the optimal combination. At this point, the age-based feed intake prediction model for a single poultry bird is complete.

[0061] In this embodiment, the movement process employs position-based PID control, with the specific flow as follows: The laser rangefinder 11 is activated, emitting a laser beam to the wall (position 0) to obtain real-time position parameters. These real-time position parameters are compared with the predicted position parameters to determine the direction of travel. In this embodiment, the control module 9 uses a position-based PID control algorithm to control the speed in segments, as shown in the following control formula:

[0062]

[0063] Where: u(k) is the control variable; K p e(k) is the scaling factor; e(k) is the error between the target position value and the actual position value; K I K is the integral coefficient; D is the differential coefficient.

[0064] For position control, the controller parameters are adjusted experimentally. First, a PI controller is used with some conservative parameters. Then, a step input signal is given. The system performance information, such as overshoot and settling time, can be obtained from the output waveform of the controlled variable. Based on the relationship between the PID parameters and system performance, the PID parameters are repeatedly adjusted. When far from the predicted position, hopper 1 maintains a relatively fast approach speed. Just before reaching the predicted position, it decelerates and slowly approaches until the support of the auxiliary wheel 4 blocks the infrared emitter 3 at the predicted position, indicating that hopper 1 has reached the designated position.

[0065] After obtaining the predicted feed intake within the feeding area and the position of hopper 1, it is necessary to detect the remaining feed in feed trough 6 after the last feeding. In this embodiment, ultrasonic testing is used, and the specific process is as follows:

[0066] First, the remaining feed in feed trough 6 is ultrasonically detected, and the returned signal is a digital signal. The digital signal is then processed according to the frequency domain method. That is, according to the principle of Fourier superposition, the original echo signal is arranged and aligned according to certain rules in the frequency domain, then superimposed, and then transformed into time domain data to obtain a two-dimensional image.

[0067] Next, the pixels in the 2D image sequence are mapped to their corresponding positions in the 3D imaging space through coordinate transformation, and the pixel values ​​before mapping are assigned to voxels. Then, each voxel in the 3D imaging space is traversed, and the corresponding pixel value in the 2D image sequence is found based on the position of the empty voxel. Finally, the empty voxels are interpolated using the Adaptive Distance Weighted (ADW) algorithm to obtain the 3D model. The formula for Adaptive Distance Weighted is as follows:

[0068]

[0069]

[0070]

[0071] Where: I(V) c I(V) represents the pixel value of the current voxel; p k ) represents the value of the k-th pixel within the voxel's neighborhood; W k For weights; d k σ represents the distance from the current voxel to its neighboring pixels. 2 represents the variance of the gray values ​​of all pixels within the neighborhood; μ represents the mean of the gray values ​​of all pixels within the neighborhood; a and b are both custom adjustable weight coefficients.

[0072] Then, the volume of the 3D model is calculated. In this embodiment, the volume calculation adopts the quasi-Monte Carlo method. The specific steps are as follows: First, construct the octree of the model and classify the nodes of the octree. Second, generate random points in each boundary node and detect whether each random point is located inside the model. Finally, count the number of random points located inside the model and calculate the volume.

[0073] Finally, based on the actual feed density and the obtained volume, the amount of feed remaining in feed trough 6 is calculated. The remaining feed is then assessed. If the remaining feed is zero, there are two possibilities: the poultry are fully fed and there is no feed left, or the poultry are not fully fed and there is no feed left. Therefore, the case of zero remaining feed needs separate explanation. When the remaining feed is zero, the predicted feed intake in the feeding area is multiplied by k (where k = 0.1F + 1.1, F is the number of cycles), which is the amount of feed precisely fed this time. After the poultry finish feeding, the remaining feed is checked again. If the remaining feed value is still zero, the next k value is selected and the cycle continues until the remaining feed is not zero. If the remaining feed is not zero, the amount of feed remaining is subtracted from the predicted feed intake in the feeding area to obtain the amount of feed precisely fed this time.

[0074] After precise feed feeding, the feed intake prediction model needs to be updated in real time. The process is as follows: Figure 3 As shown, camera 10 first obtains the signal of poultry leaving the feeding area. Ultrasonic ranging module 7 then performs another feed quantity detection to obtain the remaining feed in feed trough 6. Based on the previously precisely fed feed amount and the remaining feed in feed trough 6, the total feed intake of poultry in the feeding area is calculated. Then, based on the number of poultry in the feeding area, the average feed intake is calculated. A threshold is then applied to the average feed intake. If it is not lower than the threshold, the goal of precise feeding has been achieved. If it is lower than the threshold, it indicates that there is too much feed remaining in feed trough 6. The average feed intake is then added to the feed intake data obtained from separate feedings, and a new prediction model is established. The next feeding will use the new model to obtain the predicted feed intake of poultry in the feeding area and perform precise feeding, followed by another feed quantity detection. This cycle repeats to achieve precise feeding and reduce feed waste.

[0075] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for precise feeding of poultry based on feed intake, characterized in that: The feed intake of individual poultry was determined by feeding them separately. Feed intake data was collected from poultry of different ages, and an age-based feed intake prediction model for individual poultry was established based on the feed intake data. The status of poultry heading to the feeding area is tracked to determine the number of poultry heading to the feeding area; based on the feed intake prediction model and the number of poultry heading to the feeding area, the predicted feed intake of poultry in the feeding area is obtained. Ultrasonic detection is used to generate two-dimensional images of the remaining feed, and three-dimensional reconstruction is performed based on the two-dimensional images, wherein: The pixels in the 2D image sequence are mapped to their corresponding positions in the 3D imaging space through coordinate transformation, and the pixel values ​​before mapping are assigned to voxels. For each voxel in the 3D imaging space, the corresponding pixel value in the 2D image sequence is queried based on the position of the empty voxel. Then, the empty voxels are interpolated using an adaptive distance-weighted algorithm to obtain the 3D model. The adaptive distance-weighted formula is: In the formula, I(V) c I(V) represents the pixel value of the current voxel; p k W represents the value of the k-th pixel within the voxel's neighborhood. k For weights; d k σ represents the distance from the current voxel to its neighboring pixels. 2 represents the variance of the grayscale values ​​of all pixels within the neighborhood; μ represents the mean of the grayscale values ​​of all pixels within the neighborhood. a Both b and 'b' are custom adjustable weighting coefficients; The volume of the 3D model was calculated using the quasi-Monte Carlo method to obtain the amount of feed remaining. First, an octree of the model was constructed and the nodes of the octree were classified. Then, random points were generated in each boundary node and it was detected whether each random point was located inside the model. Finally, the number of random points located inside the model was counted to calculate the volume. The feeding amount is calculated based on the predicted feed intake and the remaining feed amount, achieving precise feeding. The remaining feed amount is then assessed. If it is zero, the predicted feed intake within the feeding area is multiplied by k, which represents the amount of feed for this precise feeding. After the poultry finish feeding, the remaining feed amount is checked again. If the value is still zero, a new k value is selected, and the cycle continues until the remaining feed amount is not zero. If the remaining feed amount is not zero, the predicted feed intake is subtracted from the remaining feed amount to obtain the amount of feed for this precise feeding. , F This represents the number of loop iterations. The process of the feed intake prediction model is as follows: The feed intake data is obtained by subtracting the amount of feed remaining from the amount of feed fed. First, the feed intake data was classified using k-fold cross-validation. Then, a day-based feed intake prediction model for a single poultry was established using forward stepwise regression. Finally, grid search and k-fold cross-validation were used to adjust the hyperparameters. The tracking and determination of the number of poultry heading to the feeding area is achieved using the StrongSORT algorithm. The specific process is as follows: First, video of poultry heading to the feeding area is acquired; second, a StrongSORT target detector is used to detect targets in each video frame; third, a high-confidence threshold and a low-confidence threshold are set, and the detected targets of the StrongSORT target detector are filtered using these thresholds; finally, the detected targets in the current video frame with a confidence level greater than the low-confidence threshold are compared with the tracked targets in the previous frame using an IoU operation to obtain... l iou And match the detected targets that satisfy the following formula with the tracked targets of the previous frame: in, score The confidence level of the detected target in the current video frame image; Finally, the matching results are fused with the detected targets filtered using a high confidence threshold to obtain the tracking targets in the current video frame image and determine the number of poultry heading to the feeding area.

2. The method for precise feeding of poultry based on feed intake according to claim 1, characterized in that, The predicted feed intake of poultry within the feeding area is: in: C qy It is the predicted feed intake of poultry within the feeding area; C yc It is the predicted feed intake per animal obtained based on the feed intake prediction model; N The number of poultry that went to the feeding area.

3. The method for precise feeding of poultry based on feed intake according to claim 1, characterized in that, After feeding is completed, the feed intake prediction model is updated in real time, specifically as follows: After feeding, the amount of feed remaining in the feed troughs in the feeding area is measured. The feed intake for this feeding is obtained by combining the amount of feed precisely fed before feeding. The average feed intake is then obtained based on the number of poultry in the feeding area. A threshold is set for the average feed intake. If it is not lower than the threshold, the requirements for precise feeding are met. If it is lower than the threshold, the feed intake data for this feeding is recorded and used together with the feed intake data collected in the poultry house before feeding to build a prediction model. The next feeding is based on the new prediction model for precise feeding, and the cycle is repeated.

4. An apparatus for implementing the precise feeding method for poultry based on feed intake as described in any one of claims 1-3, characterized in that, include: Track (12) is installed inside the poultry house; The hopper (1) is located above the track (12) and can run along the track (12); Solenoid valve (2) is located inside the conveying pipe of hopper (1); The ultrasonic ranging module (7) is located in the forward direction of the hopper (1) and is used to detect the amount of feed remaining. The camera (10), located in the forward direction of the hopper (1), is used to detect the number of poultry in the feeding area; The material discharge pipe (5) is located below the track (12); Feed trough (6) is placed at the lower end of feed pipe (5); The control module (9) is used to control the movement of the hopper (1) and to predict and accurately feed the poultry in the feeding area.

5. The apparatus according to claim 4, characterized in that, The hopper (1) is provided with a plate (13) with wheels below it. A battery (8) is provided on the plate (13). An ultrasonic ranging module (7), a control module (9) and a camera (10) are also provided on the plate (13).

6. The apparatus according to claim 4, characterized in that, It also includes an auxiliary wheel (4) and an infrared transmitter (3). The auxiliary wheel (4) is connected to the wheel, and the tail end of the auxiliary wheel (4) is embedded in the track (12). The infrared transmitter (3) is set inside the track at the top of the feed pipe (5).

Images