Whole-course intelligent straw mushroom cultivation method and system

Abstract

The invention belongs to the technical field of edible mushroom cultivation, and particularly discloses a whole-course intelligent straw mushroom cultivation method and system. According to the invention, an image recognition technology is combined, the growth condition in the straw mushroom house is collected by using the rail robot, and the regulation and control parameter strategy corresponding to the environment parameter data and the mushroom house condition in the straw mushroom growth process is set; data such as the air temperature and humidity, the carbon dioxide content in the air, the illumination intensity and the growth vigor condition of straw mushrooms are automatically collected through various sensing devices, the collected data are transmitted to the control center through network signals, and the control center can intelligently regulate and control environment parameters according to the straw mushroom growth regulation and control parameter strategy. According to the intelligent regulation and control method, the problems that an existing volvariella volvacea cultivation system consumes manpower and is low in efficiency, and the relative steady state of a volvariella volvacea house cannot be guaranteed are solved, it can be guaranteed that the growth of volvariella volvacea hyphae and sporocarps is always in the optimal environment to the maximum extent, and the volvariella volvacea production efficiency and yield are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of edible fungus cultivation, and particularly relates to a full-process intelligent cultivation method and system for straw mushrooms. Background Art

[0002] Volvariella volvacea (Bull. ex. Fr.) Sing, also known as rice straw mushroom and Chinese mushroom, belongs to the Fungi, Basidiomycota, Agaricomycetes, Agaricales, Pluteaceae, and Volvariella genus in taxonomy. It originated in tropical and subtropical regions of China. It is not only delicious and nutritious, but also has important medicinal value.

[0003] Straw mushrooms are high-temperature edible fungi, and the production environment requires high temperature and high humidity, and multi-layer bed cultivation, which is very unfriendly to production managers. In addition, when the external environment temperature is low, the mushroom house should maintain a relatively stable climate environment during the production of straw mushrooms. Frequent entry and exit of people will destroy the relative stability of the straw mushroom house. Therefore, the inspection of straw mushroom houses is also one of the problems that plague straw mushroom cultivation. With the rapid development of computer and image processing technology, information technology has been widely used in the identification and detection of agricultural products, bringing convenience to agricultural life and production. Machine vision is the process of collecting and analyzing images of real scenes through computers and other equipment to obtain target information characteristics. Summary of the invention

[0004] The purpose of the present invention is to solve the problems of the existing straw mushroom cultivation system, which is labor-intensive, inefficient and unable to ensure the relative stability of the straw mushroom house, and propose a straw mushroom full-process intelligent cultivation method and system.

[0005] The technical solution of the present invention is as follows: In the first aspect, a method for the whole process intelligent cultivation of straw mushrooms comprises the following steps:

[0006] S1. Obtaining environmental information of the mushroom house and image data of the mushroom growth;

[0007] S2. Construct a straw mushroom growth recognition model;

[0008] S3. Inputting the straw mushroom growth image data into the straw mushroom growth recognition model, and outputting the straw mushroom growth status;

[0009] S4. According to the environmental information of the straw mushroom house and the growth status of the straw mushrooms, the environmental parameters of the straw mushroom house are adjusted to realize the whole process intelligent cultivation method of straw mushrooms.

[0010] The beneficial effects of the present invention are:

[0011] 1. By constructing a straw mushroom growth recognition model to identify the growth status of straw mushrooms, the cultivation of edible fungi can be accurately judged without the help of manpower, which greatly reduces the time farmers spend on cultivation and improves efficiency.

[0012] Preferably, the environmental information of the straw mushroom house in step S1 includes air temperature and humidity, carbon dioxide content in the air, light intensity, cultivation bed weight and material temperature.

[0013] Preferably, the step S2 specifically includes the following sub-steps:

[0014] S21. Collect image data of mushrooms under different lighting conditions, different shooting angles, different background environments, different growth states, and different pests and diseases;

[0015] S22. Annotate the image data of the straw mushroom, including growth conditions, types of pests and diseases, and strain size;

[0016] S23. Flip, translate, rotate, scale and crop the information-annotated image data to obtain training set data;

[0017] S24. Input the training set data into the neural network model for training to obtain a trained neural network model;

[0018] S25. The trained neural network model is lightweighted through quantization and pruning operations to obtain a straw mushroom growth recognition model.

[0019] Preferably, the straw mushroom growth recognition model comprises a backbone network module, a feature pyramid module and a classification+center point+regression module connected in sequence;

[0020] The backbone network module is used to receive the straw mushroom growth image data, perform feature extraction on the straw mushroom growth image data, and output the straw mushroom growth image features;

[0021] The feature pyramid module is used to receive the features of the straw mushroom growth image, perform feature fusion according to the features of the straw mushroom growth image, and output feature maps of different scales;

[0022] The classification+center point+regression module is used to receive the feature maps of different scales and output the target detection results.

[0023] Preferably, the backbone network module comprises a C3 layer, a C4 layer and a C5 layer connected in sequence; the input end of the C3 layer is the input end of the backbone network module; the output ends of the C3 layer, the C4 layer and the C5 layer are also connected to the feature pyramid module;

[0024] The C3 layer is used to receive the image data of the straw mushroom growth, perform preliminary feature extraction, and output the preliminary features;

[0025] The C4 layer is used to receive the preliminary features and perform high-level feature extraction to output high-level features;

[0026] The C5 layer is used to receive the high-level features and extract abstract features, and output the abstract features;

[0027] The feature pyramid module includes a P7 layer, a P6 layer, a P5 layer, a P4 layer and a P3 layer connected in sequence; another input end of the P5 layer is connected to the C5 layer; another input end of the P4 layer is connected to the C4 layer; another input end of the P3 layer is connected to the C3 layer; the output ends of the P7 layer, the P6 layer, the P5 layer, the P4 layer and the P3 layer are also connected to the classification + center point + regression module; the P3 layer is obtained by the C3 layer through 1x1 convolution and lateral connection; the P4 layer is obtained by the C4 layer through 1x1 convolution and lateral connection; the P5 layer is obtained by the C5 layer through 1x1 convolution and lateral connection; the P6 layer is obtained by applying a 3x3 convolution with a step size of 2 on the P5 layer; the P7 layer is obtained by applying a 3x3 convolution with a step size of 2 on the P6 layer;

[0028] The classification+center point+regression module includes a first Head unit, a second Head unit, a third Head unit, a fourth Head unit and a fifth Head unit; the input end of the first Head unit is connected to the P7 layer; the input end of the second Head unit is connected to the P6 layer; the input end of the third Head unit is connected to the P5 layer; the input end of the fourth Head unit is connected to the P4 layer; the input end of the fifth Head unit is connected to the P3 layer; the output end of the first Head unit, the output end of the second Head unit, the output end of the third Head unit, the output end of the fourth Head unit and the output end of the fifth Head unit are the output ends of the classification+center point+regression module;

[0029] The first Head unit, the second Head unit, the third Head unit, the fourth Head unit and the fifth Head unit all include a classification branch, a center point branch and a regression branch;

[0030] The classification branch is used to receive the feature maps of different scales, and perform target classification according to the feature maps of different scales, and output the category probability distribution of each anchor in the straw mushroom growth image data;

[0031] The center point branch is used to receive the feature maps of different scales, and perform target center point position detection according to the feature maps of different scales, and output the center point offset of each anchor in the straw mushroom growth image data;

[0032] The regression branch is used to receive the feature maps of different scales, and perform target bounding box size and bounding box position detection according to the feature maps of different scales, and output the target bounding box parameters of each anchor in the straw mushroom growth image data.

[0033] The beneficial effects of the above preferred solution are:

[0034] By annotating and preprocessing the image data of straw mushrooms, the richness of training data can be increased, thereby enhancing the generalization of the straw mushroom growth recognition model, and lightweighting the straw mushroom growth recognition model, further improving the operating efficiency of the straw mushroom growth recognition model, and being able to accurately and efficiently identify the growth status of straw mushrooms.

[0035] In the second aspect, a straw mushroom full-process intelligent cultivation system, a data acquisition module, is used to obtain the environmental information of the straw mushroom house and the straw mushroom growth image data;

[0036] A model building module is used to build a straw mushroom growth recognition model;

[0037] A straw mushroom growth recognition module, used for inputting the straw mushroom growth image data into the straw mushroom growth recognition model, and outputting the straw mushroom growth status;

[0038] The parameter adjustment control module is used to receive the environmental information of the straw mushroom house and the growth status of the straw mushrooms, and adjust the environmental parameters of the straw mushroom house according to the environmental information of the straw mushroom house and the growth status of the straw mushrooms.

[0039] Preferably, the data acquisition module comprises:

[0040] Temperature and humidity sensors are used to monitor and collect air temperature and humidity data in the straw mushroom house;

[0041] Light sensors and photosensitive sensors are used to monitor and collect light intensity data in the straw mushroom house;

[0042] Carbon dioxide concentration sensor, used to monitor and collect the carbon dioxide content in the straw mushroom house;

[0043] A pressure sensor is used to monitor and collect water loss data of the culture medium of the straw mushroom;

[0044] The material temperature probe is buried in the matrix to monitor and collect the temperature change data of the straw mushroom culture matrix;

[0045] Track robot used to collect image data of straw mushroom growth.

[0046] Preferably, the rail robot comprises a charging dock, a rail robot body, a track, an anti-intrusion gate detection mechanism, an automatic gate and an anti-collision monitoring mechanism;

[0047] The charging dock is connected to one end of the track; the track robot body moves along the track; the automatic gate is located in the middle of the track and is perpendicular to the track; anti-intrusion gate detection mechanisms are provided on both sides of the automatic gate; the anti-collision monitoring mechanism is located at the other end of the track.

[0048] Preferably, the rail robot body is provided with an external interface; the external interface includes a preset setting interface, a control interface and an information collection interface.

[0049] Preferably, the preset setting interface includes an anti-fog preheating setting interface, an operation plan setting interface, a robot preset point setting interface, a pan / tilt preset point setting interface and a video setting interface;

[0050] The control interfaces include robot motion control interface, gate control interface and pan / tilt control interface;

[0051] The information collection interfaces include a body temperature interface, a walking condition interface, a gate condition interface, a battery condition interface, a video preview interface, a fault information interface and an alarm information interface.

[0052] The beneficial effects of the present invention are:

[0053] 1. The intelligent control system provided by the present invention can make accurate judgments on the cultivation of edible fungi without the help of manpower, which greatly reduces the time farmers spend on cultivation and improves efficiency.

[0054] 2. The whole-process intelligent control system provided by the present invention can replace the human eye and reflect the inspection results to the control center, which can reduce manpower input and improve the systematicness and accuracy of the work. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 The figure shows a flow chart of the whole intelligent cultivation method of straw mushrooms provided in Example 1 of the present invention.

[0056] Figure 2 Shown is a schematic diagram of the structure of the straw mushroom growth identification model provided in Example 1 of the present invention.

[0057] Figure 3 Shown is a graph showing the results of identifying the growth stages and diseases of straw mushrooms provided in Example 1 of the present invention.

[0058] Figure 4 The figure shows the disease identification result diagram identified by the straw mushroom growth identification model provided in Example 1 of the present invention.

[0059] Figure 5 Shown is a schematic structural diagram of the track robot provided in Example 2 of the present invention.

[0060] Figure 6 Shown is a schematic diagram of the main interface of the straw mushroom full-process intelligent cultivation system provided in Example 2 of the present invention.

[0061] Figure numerals: 1—charging dock, 2—track robot body, 3—track, 4—anti-intrusion gate detection mechanism, 5—automatic gate, 6—anti-collision monitoring mechanism. DETAILED DESCRIPTION

[0062] Now, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the accompanying drawings are only exemplary and are intended to explain the principles and spirit of the present invention, rather than to limit the scope of the present invention.

[0063] Embodiment 1:

[0064] like Figure 1 As shown, a method for the whole process intelligent cultivation of straw mushrooms comprises the following steps:

[0065] S1. Obtaining environmental information of the mushroom house and image data of the mushroom growth;

[0066] S2. Construct a straw mushroom growth recognition model;

[0067] S3. Inputting the straw mushroom growth image data into the straw mushroom growth recognition model, and outputting the straw mushroom growth status;

[0068] S4. According to the environmental information of the straw mushroom house and the growth status of the straw mushrooms, the environmental parameters of the straw mushroom house are adjusted to realize the whole process intelligent cultivation method of straw mushrooms.

[0069] In this embodiment, the environmental information of the straw mushroom house in step S1 includes air temperature and humidity, carbon dioxide content in the air, light intensity, cultivation bed weight and material temperature.

[0070] In this embodiment, step S2 specifically includes the following sub-steps:

[0071] S21. Collect image data of mushrooms under different lighting conditions, different shooting angles, different background environments, different growth states, and different pests and diseases;

[0072] S22. Annotate the image data of the straw mushroom, including growth conditions, types of pests and diseases, and strain size;

[0073] S23. Flip, translate, rotate, scale and crop the information-annotated image data, and use the GAN network to generate more data samples to construct training set data;

[0074] S24. Through the transfer learning system, the training set data is input into the neural network model for training, specifically: adding an online hard example mining module (OHEM), deformable convolution, improving the model loss function (Focal-Loss), increasing or decreasing the model depth, increasing the model receptive field (Non-Local) and increasing the richness of training samples (Data Augmentation), to obtain a trained neural network model;

[0075] S25. The trained neural network model is lightweighted through quantization and pruning operations to obtain a straw mushroom growth recognition model, and the straw mushroom growth recognition model is deployed on the Coral-TPU edge computing device; the quantization specifically includes: converting the weights of float32 floating-point numbers in the neural network model parameters into int8 integers to improve the operating efficiency of the neural network model; the pruning specifically includes: disconnecting the connections between neurons that are not frequently used in the neural network model to reduce the number of parameters of the neural network model.

[0076] In this embodiment, if Figure 2 As shown, the straw mushroom growth recognition model includes a backbone network module, a feature pyramid module, and a classification+center point+regression module connected in sequence;

[0077] The backbone network module is used to receive the straw mushroom growth image data, perform feature extraction on the straw mushroom growth image data, and output the straw mushroom growth image features;

[0078] The feature pyramid module is used to receive the features of the straw mushroom growth image, perform feature fusion according to the features of the straw mushroom growth image, and output feature maps of different scales;

[0079] The classification+center point+regression module is used to receive the feature maps of different scales and output the target detection results.

[0080] The backbone network module includes a C3 layer, a C4 layer and a C5 layer connected in sequence, and the C3 layer, the C4 layer and the C5 layer are all composed of a convolutional neural network (CNN); the input end of the C3 layer is the input end of the backbone network module; the output ends of the C3 layer, the C4 layer and the C5 layer are also connected to the feature pyramid module;

[0081] The C3 layer is used to receive the image data of the straw mushroom growth, perform preliminary feature extraction, and output the preliminary features;

[0082] The C4 layer is used to receive the preliminary features and perform high-level feature extraction to output high-level features;

[0083] The C5 layer is used to receive the high-level features and extract abstract features, and output the abstract features;

[0084] The feature pyramid module includes a P7 layer, a P6 layer, a P5 layer, a P4 layer and a P3 layer connected in sequence; another input end of the P5 layer is connected to the C5 layer; another input end of the P4 layer is connected to the C4 layer; another input end of the P3 layer is connected to the C3 layer; the output ends of the P7 layer, the P6 layer, the P5 layer, the P4 layer and the P3 layer are also connected to the classification + center point + regression module; the feature pyramid module divides the detection target into five categories according to the size of the detection target, namely the first category, the second category, the third category, the fourth category and the fifth category;

[0085] The P3 layer is obtained by the C3 layer through 1x1 convolution and lateral connection, and is used to detect smaller first-category targets;

[0086] The P4 layer is obtained by the C4 layer through 1x1 convolution and lateral connection, and is used to detect medium-sized second-category targets;

[0087] The P5 layer is obtained by the C5 layer through 1x1 convolution and lateral connection, and is used to detect larger third-category targets, and generates the P6 and P7 layers through downsampling;

[0088] The P6 layer is obtained by applying a 3x3 convolution with a stride of 2 on the P5 layer, and is used to detect the fourth type of objects with a larger size;

[0089] The P7 layer performs another 3x3 convolution with a stride of 2 on the basis of the P6 layer to detect the fifth type of target of the largest size; in FCOS, these layers adapt to the detection of targets of different sizes through different downsampling rates (stride). The strides of the P3, P4, P5, P6 and P7 layers are 8, 16, 32, 64 and 128 respectively, which can cover the range of targets from very small to very large. Each feature layer is input to the detection head (Head) and predictions are performed for classification, center and box regression tasks. This design allows FCOS to effectively perform target detection in a single stage without the need for a traditional anchor mechanism, thereby simplifying the target detection process and improving efficiency.

[0090] The classification+center point+regression module includes a first Head unit, a second Head unit, a third Head unit, a fourth Head unit and a fifth Head unit; the input end of the first Head unit is connected to the P7 layer; the input end of the second Head unit is connected to the P6 layer; the input end of the third Head unit is connected to the P5 layer; the input end of the fourth Head unit is connected to the P4 layer; the input end of the fifth Head unit is connected to the P3 layer; the output end of the first Head unit, the output end of the second Head unit, the output end of the third Head unit, the output end of the fourth Head unit and the output end of the fifth Head unit are the output ends of the classification+center point+regression module;

[0091] The first Head unit, the second Head unit, the third Head unit, the fourth Head unit and the fifth Head unit all include a classification branch, a center point branch and a regression branch;

[0092] The classification branch is used to receive the feature maps of different scales, and classify the target according to the feature maps of different scales, that is, to determine which category it belongs to (for example, straw mushroom or other), and output the category probability distribution of each anchor in the straw mushroom growth image data, wherein the category probability distribution is usually a vector, in which each element corresponds to the probability of a category;

[0093] The center point branch is used to receive the feature maps of different scales, and detect the target center point position according to the feature maps of different scales by predicting the offset relative to the anchor, and output the center point offset of each anchor in the straw mushroom growth image data, which is usually a two-dimensional vector representing the offset of the center point relative to the anchor;

[0094] The regression branch is used to receive the feature maps of different scales, and perform target bounding box size and bounding box position detection according to the feature maps of different scales, which involves predicting the width and height of the target, as well as possible center point offset, and outputting target bounding box parameters of each anchor in the straw mushroom growth image data, including width, height and center point offset.

[0095] In this embodiment, the cultivation cycle of straw mushroom is 13 days in total, as shown in Table 1, wherein ↑ represents improvement, and the growth stage of straw mushroom is divided into 7 stages: cultivation (afternoon of the first day after inoculation to the fourth day after inoculation), stimulation (afternoon of the fifth day after inoculation), recovery (evening of the fifth day after inoculation), original machine formation (from the sixth day after inoculation to the seventh day after inoculation), original machine turning black (the eighth day after inoculation), growth (from the ninth day after inoculation to the twelfth day after inoculation) and harvesting and maturation (the thirteenth day after inoculation); the straw mushroom growth state described in step S3 includes straw mushroom growth stage identification and disease identification, and the diseases include fungus infection mushroom, navel mushroom, small ghost cap and fungus on the surface of cultivation material; the straw mushroom growth image data is input into the straw mushroom growth identification model, and the straw mushroom growth image data is feature extracted and identified through the backbone network module, feature pyramid module and classification + center point + regression module connected in sequence, and the straw mushroom growth state stage and fungus infection mushroom, navel mushroom, small ghost cap and fungus on the surface of cultivation material in the straw mushroom house are output, such as Figure 3 and Figure 4 As shown, Figure 3 Figures a, b, c and d in the middle are the straw mushrooms identified at different growth stages. Figure 4 Figures a, b, c and d in the middle are the results of identifying diseases such as the identified straw mushrooms and mushrooms infected with assorted fungi, navel mushrooms, small ghost caps and assorted fungi on the surface of the cultivation material. The green frame represents the straw mushroom, and the red frame represents diseases such as the mushrooms infected with assorted fungi, navel mushrooms, small ghost caps and assorted fungi on the surface of the cultivation material. Then, according to the environmental information of the straw mushroom house, the growth state stage of the straw mushrooms and the mushrooms infected with assorted fungi, navel mushrooms, small ghost caps and assorted fungi on the surface of the cultivation material, the environmental parameters of the straw mushroom house are adjusted; the wind speed, material temperature, air temperature and humidity and carbon dioxide content in the air of the straw mushroom house are set differently at different stages; as shown in Table 2, when mushrooms infected with assorted fungi are identified, the mushrooms infected with assorted fungi are picked to avoid infection; when navel mushrooms are identified, the carbon dioxide content in the air of the straw mushroom house is reduced to less than 1000; when small ghost caps are identified, the material temperature is adjusted to be greater than 32°C; when assorted fungi on the surface of the cultivation material are identified, the material temperature is adjusted to be greater than 32°C, so as to realize the full-process intelligent cultivation method of straw mushrooms.

[0096] Table 1 Growth regulation strategies of straw mushrooms

[0097]

[0098]

[0099] Table 2 Regulation of growth and disease of straw mushroom

[0100] Fungus infection mushroom Pick as soon as possible to avoid infection Navel Mushroom Reduce the CO2 concentration in the mushroom house to less than 1000 Coprinus Material temperature is greater than 32℃ Bacteria on the surface of the cultivation material Material temperature is greater than 32℃

[0101] Embodiment 2:

[0102] On the basis of Example 1, the present invention provides a whole-process intelligent cultivation system of straw mushrooms, which is used to implement the whole-process intelligent cultivation method of straw mushrooms as described in Example 1, and the system comprises:

[0103] A data acquisition module is used to obtain the environmental information of the straw mushroom house and the straw mushroom growth image data;

[0104] Model building module, used to build a straw mushroom growth recognition model;

[0105] A straw mushroom growth recognition module, used for inputting the straw mushroom growth image data into the straw mushroom growth recognition model, and outputting the straw mushroom growth status;

[0106] The parameter adjustment control module is used to receive the environmental information of the straw mushroom house and the growth status of the straw mushrooms, and adjust the environmental parameters of the straw mushroom house according to the environmental information of the straw mushroom house and the growth status of the straw mushrooms.

[0107] In this embodiment, the data acquisition module includes:

[0108] Temperature and humidity sensors are used to monitor and collect air temperature and humidity data in the straw mushroom house;

[0109] Light sensors and photosensitive sensors are used to monitor and collect light intensity data in the mushroom house; light sensors and photosensitive sensors can be directly connected to related supplementary lighting systems, sunshade systems and other equipment, automatically turn on related equipment when necessary, and transmit related data to the user monitoring terminal through wireless transmission technology;

[0110] The carbon dioxide concentration sensor is used to monitor and collect the carbon dioxide content in the mushroom house. When the concentration exceeds the preset threshold range, the carbon dioxide content data is transmitted to the user monitoring terminal through wired or wireless transmission, and the relevant staff makes corresponding adjustments; it can also be automatically controlled, specifically, the exhaust fan and the ventilation fan are automatically turned on by the controller for ventilation. When the carbon dioxide drops to normal, the controller automatically turns off the exhaust fan and the ventilation fan;

[0111] A pressure sensor is used to monitor and collect water loss data of the culture medium of the straw mushroom;

[0112] The material temperature probe is buried in the matrix to monitor and collect the temperature change data of the straw mushroom culture matrix;

[0113] Track robot used to collect image data of straw mushroom growth.

[0114] In this embodiment, if Figure 5 As shown, the rail robot includes a charging dock 1, a rail robot body 2, a track 3, an anti-intrusion gate detection mechanism 4, an automatic gate 5 and an anti-collision monitoring mechanism 6;

[0115] The charging dock 1 is connected to one end of the track 3; the track robot body 2 moves along the track 3; the automatic gate 5 is located in the middle of the track 3 and is perpendicular to the track 3; anti-intrusion gate detection mechanisms 4 are provided on both sides of the automatic gate 5; the anti-collision monitoring mechanism 6 is located at the other end of the track 3.

[0116] The rail robot body 2 is provided with a preset setting interface, a control interface and an information collection interface;

[0117] The preset setting interface includes an anti-fog preheating setting interface, an operation plan setting interface, a robot preset point setting interface, a pan / tilt preset point setting interface and a video setting interface;

[0118] The anti-fog preheating setting interface is used to set the upper threshold of the heating temperature of the machine body. The main control board will detect and control it automatically and stop heating when the temperature reaches the threshold;

[0119] The operation plan setting interface is used to run the walking plan table, and the robot operates autonomously according to the time and inspection points set in the plan table;

[0120] The robot preset point setting interface is used to make the robot pause at a certain position on the track by using the stop mode, and record the position and catalog the walking preset points;

[0121] The pan / tilt preset point setting interface is used to set the capture angle of the collected object using the camera loaded on the pan / tilt, and 10 angles can be preset;

[0122] The video setting interface is used to perform basic configuration of the camera and camera access authentication configuration.

[0123] The control interfaces include robot motion control interface, gate control interface and pan / tilt control interface;

[0124] The robot motion control interface is used to realize the robot's forward, backward and stop operations;

[0125] The gate control interface is used to control the opening and closing of the gates, and needs to open and close the two gates synchronously at the same time;

[0126] The pan / tilt control interface is used to adjust the pan / tilt at preset walking points, and the capture angle can be displayed in real time according to the camera loaded on the pan / tilt.

[0127] The information collection interfaces include body temperature interface, walking condition interface, gate condition interface, battery condition interface, video preview interface, fault information interface and alarm information interface;

[0128] The body temperature interface is used to collect body temperature information and provide it to the main control and central control for calling, wherein the main control part is used for heating control (internal interface).

[0129] The walking condition interface is used for travel record, start, stop and reverse operation feedback information, and the walking condition can be learned through the feedback information.

[0130] The gate working condition interface is used to detect the gate opening and closing degree detection result information. Through this interface, the intrusion behavior of the robot can be effectively avoided.

[0131] The battery operating condition interface is used to record the voltage, current and output power of the battery in real time, providing a basis for battery charging and discharging (internal interface), and is related to the energy consumption condition.

[0132] The video preview interface is used for real-time video display, and the robot walking, walking preset points, capture angle selection and capture preset point setting are completed by loading the preview video.

[0133] The fault information interface is used to record abnormal fault alarms occurring in various working conditions, such as communication failures, abnormal working conditions, etc.

[0134] The alarm information interface is used to record the alarm information of the device itself, such as video authentication anomaly, channel acquisition anomaly and other records.

[0135] In this embodiment, the parameter adjustment control module has three modes of automatic control, manual control and centralized monitoring when adjusting the environmental parameters of the straw mushroom house according to the environmental information of the straw mushroom house and the growth status of the straw mushrooms;

[0136] The automatic control mode runs automatically according to the pre-programmed program based on the set parameters;

[0137] Manual control mode: According to the needs, you can choose the on-site manual control mode and start various modes at will;

[0138] The centralized monitoring mode connects the control boxes of each mushroom house in series through a communication cable, and then transmits it to the monitoring computer in the monitoring center, forming a centralized monitoring system for the mushroom house environment. The monitoring system can display and automatically record the monitoring data (humidity, temperature, carbon dioxide content) of each mushroom house and the working status of peripheral equipment in real time, remotely set the working parameters of each control box, automatically alarm and record the time and cause of the alarm.

[0139] The main interface of the whole process intelligent cultivation system of straw mushrooms proposed by the present invention is as follows: Figure 6As shown in the figure, according to the big data of straw mushrooms, combined with the technology of image recognition, the computer control center performs intelligent control. The intelligent control system can make accurate judgments on the cultivation of edible fungi without the help of manpower, which greatly reduces the time farmers spend on cultivation and improves efficiency. At the same time, the intelligent control of the whole process can also be applied to edible fungi factories. For the factory's warehouse inspection, it can replace the human eye and reflect the inspection results to the control center. While reducing manpower input, it can also improve the systematicness and accuracy of the work.

[0140] The whole-process intelligent cultivation system of straw mushrooms proposed by the present invention installs supporting Internet of Things monitoring software, so that users can obtain real-time information such as temperature, humidity, carbon dioxide content, and operation status of equipment in mushroom houses through the Internet at any place, so that the production of mushrooms can be fully controlled.

[0141] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for the whole process intelligent cultivation of straw mushrooms, characterized in that: The method comprises the following steps: S1. Obtaining environmental information of the mushroom house and image data of the mushroom growth; S2. Construct a straw mushroom growth recognition model; S3. Inputting the straw mushroom growth image data into the straw mushroom growth recognition model, and outputting the straw mushroom growth status; S4. According to the environmental information of the straw mushroom house and the growth status of the straw mushrooms, the environmental parameters of the straw mushroom house are adjusted to realize the whole process intelligent cultivation method of straw mushrooms.

2. The whole process intelligent cultivation method of straw mushroom according to claim 1, characterized in that: The environmental information of the straw mushroom house in step S1 includes air temperature and humidity, carbon dioxide content in the air, light intensity, water loss of the substrate and material temperature.

3. The whole process intelligent cultivation method of straw mushroom according to claim 1, characterized in that: The step S2 specifically includes the following sub-steps: S21. Collect image data of mushrooms under different lighting conditions, different shooting angles, different background environments, different growth states, and different pests and diseases; S22. Annotate the image data of the straw mushroom, including growth conditions, types of pests and diseases, and strain size; S23. Flip, translate, rotate, scale and crop the information-annotated image data to obtain training set data; S24. Input the training set data into the neural network model for training to obtain a trained neural network model; S25. The trained neural network model is lightweighted through quantization and pruning operations to obtain a straw mushroom growth recognition model.

4. The whole process intelligent cultivation method of straw mushroom according to claim 3 is characterized in that: The straw mushroom growth recognition model comprises a backbone network module, a feature pyramid module and a classification+center point+regression module connected in sequence; The backbone network module is used to receive the straw mushroom growth image data, perform feature extraction on the straw mushroom growth image data, and output the straw mushroom growth image features; The feature pyramid module is used to receive the features of the straw mushroom growth image, perform feature fusion according to the features of the straw mushroom growth image, and output feature maps of different scales; The classification+center point+regression module is used to receive the feature maps of different scales and output the target detection results.

5. The whole process intelligent cultivation method of straw mushroom according to claim 4 is characterized in that: The backbone network module includes a C3 layer, a C4 layer and a C5 layer connected in sequence; the input end of the C3 layer is the input end of the backbone network module; the output ends of the C3 layer, the C4 layer and the C5 layer are also connected to the feature pyramid module; The C3 layer is used to receive the image data of the straw mushroom growth, perform preliminary feature extraction, and output the preliminary features; The C4 layer is used to receive the preliminary features and perform high-level feature extraction to output high-level features; The C5 layer is used to receive the high-level features and extract abstract features, and output the abstract features; The feature pyramid module includes a P7 layer, a P6 layer, a P5 layer, a P4 layer and a P3 layer connected in sequence; another input end of the P5 layer is connected to the C5 layer; another input end of the P4 layer is connected to the C4 layer; another input end of the P3 layer is connected to the C3 layer; the output ends of the P7 layer, the P6 layer, the P5 layer, the P4 layer and the P3 layer are also connected to the classification + center point + regression module; the P3 layer is obtained by the C3 layer through 1x1 convolution and lateral connection; the P4 layer is obtained by the C4 layer through 1x1 convolution and lateral connection; the P5 layer is obtained by the C5 layer through 1x1 convolution and lateral connection; the P6 layer is obtained by applying a 3x3 convolution with a step size of 2 on the P5 layer; the P7 layer is obtained by applying a 3x3 convolution with a step size of 2 on the P6 layer; The classification+center point+regression module includes a first Head unit, a second Head unit, a third Head unit, a fourth Head unit and a fifth Head unit; the input end of the first Head unit is connected to the P7 layer; the input end of the second Head unit is connected to the P6 layer; the input end of the third Head unit is connected to the P5 layer; the input end of the fourth Head unit is connected to the P4 layer; the input end of the fifth Head unit is connected to the P3 layer; the output end of the first Head unit, the output end of the second Head unit, the output end of the third Head unit, the output end of the fourth Head unit and the output end of the fifth Head unit are the output ends of the classification+center point+regression module; The first Head unit, the second Head unit, the third Head unit, the fourth Head unit and the fifth Head unit all include a classification branch, a center point branch and a regression branch; The classification branch is used to receive the feature maps of different scales, and perform target classification according to the feature maps of different scales, and output the category probability distribution of each anchor in the straw mushroom growth image data; The center point branch is used to receive the feature maps of different scales, and perform target center point position detection according to the feature maps of different scales, and output the center point offset of each anchor in the straw mushroom growth image data; The regression branch is used to receive the feature maps of different scales, and perform target bounding box size and bounding box position detection according to the feature maps of different scales, and output the target bounding box parameters of each anchor in the straw mushroom growth image data.

6. A whole-process intelligent cultivation system of straw mushrooms, characterized in that: include: A data acquisition module is used to obtain the environmental information of the straw mushroom house and the straw mushroom growth image data; A model building module is used to build a straw mushroom growth recognition model; A straw mushroom growth recognition module, used for inputting the straw mushroom growth image data into the straw mushroom growth recognition model, and outputting the straw mushroom growth status; The parameter adjustment control module is used to receive the environmental information of the straw mushroom house and the growth status of the straw mushrooms, and adjust the environmental parameters of the straw mushroom house according to the environmental information of the straw mushroom house and the growth status of the straw mushrooms.

7. The whole process intelligent cultivation system of straw mushroom according to claim 6 is characterized in that: The data acquisition module comprises: Temperature and humidity sensors are used to monitor and collect air temperature and humidity data in the straw mushroom house; Light sensors and photosensitive sensors are used to monitor and collect light intensity data in the straw mushroom house; Carbon dioxide concentration sensor, used to monitor and collect the carbon dioxide content in the straw mushroom house; A pressure sensor is used to monitor and collect water loss data of the culture medium of the straw mushroom; The material temperature probe is buried in the matrix to monitor and collect the temperature change data of the straw mushroom culture matrix; Track robot used to collect image data of straw mushroom growth.

8. The whole process intelligent cultivation system of straw mushroom according to claim 7 is characterized in that: The rail robot comprises a charging dock (1), a rail robot body (2), a rail (3), an anti-intrusion gate detection mechanism (4), an automatic gate (5) and an anti-collision monitoring mechanism (6); The charging dock (1) is connected to one end of the track (3); the track robot body (2) moves along the track (3); the automatic gate (5) is located in the middle of the track (3) and is perpendicular to the track (3); anti-intrusion gate detection mechanisms (4) are provided on both sides of the automatic gate (5); and the anti-collision monitoring mechanism (6) is located at the other end of the track (3).

9. The whole process intelligent cultivation system of straw mushroom according to claim 8, characterized in that: The track robot body (2) is provided with an external interface; the external interface comprises a preset setting interface, a control interface and an information collection interface.

10. The whole process intelligent cultivation system of straw mushroom according to claim 9, characterized in that: The preset setting interface includes an anti-fog preheating setting interface, an operation plan setting interface, a robot preset point setting interface, a pan / tilt preset point setting interface and a video setting interface; The control interfaces include robot motion control interface, gate control interface and pan / tilt control interface; The information collection interfaces include a body temperature interface, a walking condition interface, a gate condition interface, a battery condition interface, a video preview interface, a fault information interface and an alarm information interface.

Images