CN114839197A - Rice damage detection device and detection method - Google Patents

Abstract

The present application relates to the technical field of paddy internal damage detection, and in particular, to a paddy damage detection device and a detection method. The device adopts a feeding module to drive the paddy to move toward the direction of the sorting module; A light hole is arranged on the light source; the light source module emits light to the light hole; the visual sensor module acquires the image of the rice passing through the sorting module under the light; and the main control unit is used to detect the internal damage of the rice according to the image information. In this way, through the cooperative work of each module, the detection of the internal damage of the paddy can be automatically realized, and the internal damage rate of a large amount of paddy can be accurately obtained.

Description

A kind of rice damage detection device and detection method

technical field

The present application relates to the technical field of internal damage detection of paddy, and in particular, to a paddy damage detection device and a detection method.

Background technique

Rice is one of the most important food crops in my country. In recent years, the rapid development of rice mechanized harvesting and drying technology has greatly improved production efficiency. When using mechanized harvesting and drying of rice, it will inevitably cause grain damage, which includes external damage and internal damage. It is manifested as stress cracks inside the grain, which affects the storage of the grain and the quality of the rice after processing. Especially when the grain is used as a seed, it will affect its germination rate. The traditional detection method is to peel off the rice husks and observe with the aid of a magnifying glass, a microscope and other instruments, which is labor-intensive and inefficient. With the development of science and technology, the non-destructive testing of grain internal damage has attracted the attention of scholars at home and abroad. For example, the related literature reported that the internal image of grain was obtained by computer vision technology, Micro-CT scanning method, acoustic emission method, etc. The treatment realizes the three-dimensional reconstruction of its structure and the characterization of internal damage, which provides an idea for the non-destructive detection of internal damage of rice, but these processing methods still mainly focus on the characterization of the internal damage type of single grain, and it is difficult to achieve continuous continuous production of large quantities of rice. Non-destructive testing, it is difficult to obtain an accurate internal damage rate.

SUMMARY OF THE INVENTION

The present application provides a paddy damage detection device and a detection method to solve the problem in the prior art that it is difficult to accurately obtain the internal damage rate of a large amount of paddy when detecting paddy.

In order to achieve the above object, the application realizes through the following technical solutions:

In a first aspect, the present application provides a paddy damage detection device, comprising: a feeding module, a sorting module, a light source module, a visual sensor module and a main control unit, the sorting module is connected to the feeding module, and the light source module It is arranged below the sorting module, the visual sensor is arranged above the sorting module, and the visual sensor and the light source module are located on the same vertical axis. The main control unit and the visual sensor connect;

The feeding module is used to drive the paddy to move towards the direction of the sorting module;

The sorting module is used for single-grain axial queuing of the rice, and the sorting module is provided with light holes;

The light source module is used for emitting light to the light hole;

The visual sensor module is used for acquiring the image of the rice grains passing through the sorting module under the light;

The main control unit is used for detecting the internal damage of the rice according to the image information.

Optionally, the feeding module includes a base, a first elastic sheet group, a second elastic sheet group, an electromagnetic exciter, a feeding chute, an armature and a hopper, and one end of the first elastic sheet group is connected to the feeding chute, so the The other end of the first elastic sheet group is connected with the base, the second elastic sheet group is arranged opposite to the first elastic sheet group, one end of the second elastic sheet group is connected with the feeding chute, and the first elastic sheet group is connected to the feeding slot. The other end of the second elastic plate group is connected to the base, the armature is arranged on the first elastic plate group, the electromagnetic vibration exciter is arranged on the base, and the electromagnetic vibration exciter is facing the armature The hopper is arranged on the material trough, and is located at one end of the material trough, and the other end of the material trough is connected with the sorting module.

Optionally, the sorting module includes N channels and collecting troughs arranged in parallel, N is a positive integer, the N channels are all V-shaped channels, and each channel includes a first side surface and a second side surface, and all of the N channels are V-shaped channels. An end of the first side face close to the feeding module is provided with a first blanking port, an end of the second side face close to the feeding module is provided with a second blanking port, and the first blanking port is provided. Both the opening and the second blanking opening are arranged in a direction away from the bottom edge of the V shape to form a reserved channel, the width of the first blanking opening is smaller than the width of the first side surface, and the second blanking opening The width of the second side is smaller than the width of the second side, wherein the width of the reserved passage is greater than the size of a single grain of rice and less than the size of a double-grained rice;

the collecting trough is arranged below the first blanking opening and the second blanking opening to collect the rice grains dropped from the first blanking opening and the second blanking opening;

Moreover, one end of each channel away from the feeding module is provided with a first light-passing hole, and the first light-passing hole is disposed opposite to the light source module.

Optionally, the light source module includes a light guide strip and a cold light source assembly, the cold light source assembly includes a light shield and a cold light source arranged inside the light shield, and the light shield is provided with a second light through hole, One end of the light guide strip is connected to the first light through hole, and the other end of the light guide strip is connected to the second light through hole.

Optionally, it also includes a cover plate, a bottom plate and a support frame, one end of the support frame is connected to the cover plate, and the other end of the support frame is connected to the bottom plate to form a closed cavity to accommodate the sorting module and the light source module.

Optionally, the main control unit includes a paddy detection subunit and a classification counting subunit, and the paddy detection subunit is used to identify the image, obtain the target frame corresponding to the paddy, and also be used to determine the target frame. The target frame coordinates and target frame type, the target frame coordinates are used to indicate the position of the paddy in the image, the target frame type is used to indicate the type of paddy in the image, and the paddy type includes internally damaged paddy or no internal paddy. damaged rice;

The classification counting subunit is used for classifying and processing based on the target frame coordinates and the target frame type to obtain the internal damage of the paddy, where the internal damage of the paddy includes the quantity of internally damaged paddy, the quantity of internal non-damaged paddy and the damage rate.

In a second aspect, the present application also provides a method for detecting damage to rice, which is applied to the device for detecting damage to rice described in the first aspect, and the method includes:

The feeding module drives the paddy to move towards the direction of the sorting module;

The sorting module arranges single grains of rice axially, and the sorting module is provided with light holes;

The light source module emits light to the light hole;

The visual sensor module acquires the image of the rice grains passing through the sorting module under the light;

The main control unit detects the internal damage of the rice according to the image information.

Optionally, the main control unit includes a paddy detection subunit and a classification counting subunit, and the main control unit detects the internal damage of the paddy according to the image information, including:

The rice detection subunit identifies the image, obtains the target frame corresponding to the rice, and determines the target frame coordinates and target frame type of the target frame, and the target frame coordinates are used to indicate the position of the rice in the image. The target box type is used to indicate the type of paddy in the image, and the paddy type includes inner damaged paddy or inner non-damaged paddy;

The classification counting subunit performs classification processing based on the coordinates of the target frame and the type of the target frame to obtain the internal damage condition of the paddy, where the internal damage condition of the paddy includes the quantity of internally damaged paddy, the quantity of internal non-damaged paddy and the damage rate.

Beneficial effects:

In the paddy damage detection device provided by the present application, a feeding module is used to drive the paddy to move towards the direction of the sorting module; the sorting module axially queues single grains of the paddy, and the sorting module is provided with a light hole; the light source module emits light to the light hole ; The visual sensor module obtains the image of the rice under the light passing through the sorting module; and the main control unit is used to detect the internal damage of the rice according to the image information. The detection of the internal damage of rice can accurately obtain the internal damage rate of a large amount of rice.

Description of drawings

1 is a front view of a rice damage detection device according to a preferred embodiment of the application;

Fig. 2 is the left side view of a kind of rice damage detection device according to the preferred embodiment of the application;

3 is an axial view of a rice damage detection device according to a preferred embodiment of the present application;

Fig. 4 is the schematic diagram of the reflection situation of the rice in the preferred embodiment of the present invention under the irradiation of light;

5 is a schematic diagram of a feeding module according to a preferred embodiment of the present application;

6 is a schematic diagram of a sorting module according to a preferred embodiment of the present application;

7 is a schematic diagram of a light source module according to a preferred embodiment of the present application;

8 is a schematic diagram of the connection between the sorting module and the light source module according to the preferred embodiment of the present invention;

Fig. 9 is the schematic diagram of the rice moving direction and coordinate value of the preferred embodiment of the present invention;

FIG. 10 is one of the flowcharts of a deep learning-based rice damage detection method according to a preferred embodiment of the present invention;

FIG. 11 is the second flow chart of a deep learning-based rice damage detection method according to a preferred embodiment of the present invention.

Reference number:

1. Feeding module; 1-1, base; 1-2, second shrapnel group; 1-3, hopper; 1-4, feeding chute; 1-5, first shrapnel group; 1-6, armature; 1 -7. Electromagnetic exciter; 2. Cover plate; 3. Vision sensor mounting frame; 4. Vision sensor module; 5. Light source module; 3-1, Sorting module; 3-2, Light guide strip; 3-3, Cold light source assembly; 3-1-1, channel; 3-1-2, first light-passing hole; 3-1-3, collecting trough; 3-1-4, first blanking port; 3-3- 1. Cold light source; 3-3-2, hood; 3-3-3, second light hole; 6, bottom plate; 7, support frame; 8, main control unit; 201, light source; 202, cutting template; 203, rice; 204, crack.

Detailed ways

The technical solutions of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

Unless otherwise defined, technical or scientific terms used in this application shall have the ordinary meaning as understood by those of ordinary skill in the art to which this application belongs. As used in this application, "first", "second" and similar words do not denote any order, quantity, or importance, but are merely used to distinguish the various components. Likewise, "a" or "an" and the like do not denote a quantitative limitation, but rather denote the presence of at least one. Words like "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "Down", "Left", "Right", etc. are only used to represent the relative positional relationship, and when the absolute position of the described object changes, the relative positional relationship also changes accordingly.

Please refer to FIG. 1-FIG. 3. A paddy damage detection device provided by the present application includes: a feeding module 1, a sorting module 3-1, a light source module 5, a visual sensor module 4 and a main control unit 8. The sorting module 3-1 is connected to the feeding module 1, the light source module 5 is arranged on the opposite side of the sorting module 3-1, the visual sensor is arranged on the opposite side of the sorting module 3-1, and the The visual sensor and the light source module 5 are located on the same vertical axis, and the main control unit 8 is connected to the visual sensor;

The feeding module 1 is used to drive the rice to move in the direction of the sorting module 3-1;

The sorting module 3-1 is used for single-grain axial queuing of rice, and the sorting module 3-1 is provided with a light hole;

The light source module 5 is used for emitting light to the light hole;

The visual sensor module 4 is used to obtain the image of the rice grains passing through the sorting module 3-1 under the light;

The main control unit 8 is used for detecting the internal damage of the paddy according to the image information.

It should be noted that, as shown in FIG. 4 , it is assumed that the paddy 203 is an ellipsoid-shaped entity, which is composed of outer chaff and inner rice, the paddy 203 is placed on the cutting board 202 , and the optical properties of the paddy 203 are used to make the paddy 203 A light source 201 is arranged at the bottom of one end, a part of the light is reflected, and the other part of the light enters the inside of the rice 203 through the husk, and the light is transmitted and refracted inside the rice 203. From the top of the rice 203, the internal conditions of the rice 203 can be observed and identified, and the rice is complete and without damage. The overall brightness of 203 is uniform and transparent, and the damaged area of rice 203 with internal cracks 204 can be observed in areas with darker brightness than those without damage. This is because the light that penetrates into the interior of rice 203 is damaged by cracks 204 and other damages. There will be strong reflection when the surface is damaged, the light intensity on the other side of the damaged area will be significantly weakened, and the brightness on both sides of the damaged area will have obvious contrast.

In this embodiment, the visual sensor module 4 can be installed in a position facing the rice through the visual sensor mounting bracket 3 , and the visual sensor module 4 can adopt a 1080p, 30f/s, high-definition camera with ten times optical zoom, and is arranged on the sorting module 3 Right above the first light-passing hole 3-1-2 on -1, the image information of the rice after being irradiated by cold light in the three V-shaped channels 3-1-1 is collected in real time, and transmitted to the main control unit 8 for detection. During the first inspection process, the rice collected in the collecting trough 3-1-3 of the sorting module 3-1 should be added to the hopper 1-3 of the feeding module 1 again, until all the rice is removed from the sorting module 3-1 v-channel 3-1 Pass through the light hole of -1 to complete the image acquisition.

The above-mentioned paddy damage detection device, through the cooperation of each module, sorts the paddy to be detected through the detection area illuminated by the light source, and utilizes the optical characteristics of the paddy to realize the non-destructive display of the characteristics of the inner damaged paddy, and the visual sensing device in the detection area is online. Obtain the rice image and transmit it to the main control unit 8, use the trained deep learning target detection and counting model set in the main control unit 8 to perform target detection, classification and counting on the internally damaged paddy and the non-damaged paddy, and calculate the damage rate. The detection of the internal damage of the paddy is automatically realized, and the internal damage rate of a large amount of rice can be accurately obtained.

Optionally, please refer to FIG. 5 , the feeding module 1 includes a base 1-1, a first elastic sheet group 1-5, a second elastic sheet group 1-2, an electromagnetic vibration exciter 1-7, and a feeding slot 1-4 , armature 1-6 and hopper 1-3, one end of the first elastic plate group 1-5 is connected with the feeding slot 1-4, and the other end of the first elastic plate group 1-5 is connected with the base 1- 1 connection, the second elastic sheet group 1-2 is arranged opposite to the first elastic sheet group 1-5, one end of the second elastic sheet group 1-2 is connected with the feeding slot 1-4, the The other end of the second elastic plate group 1-2 is connected to the base 1-1, the armature 1-6 is arranged on the first elastic plate group 1-5, and the electromagnetic vibration exciter 1-7 is arranged on the on the base 1-1, and the electromagnetic vibration exciter 1-7 is arranged facing the armature 1-6, the hopper 1-3 is arranged on the material trough, and is located at one end of the material trough, The other end of the trough is connected to the sorting module 3-1.

In this embodiment, the electromagnetic exciter 1-7 generates periodic pulse electromagnetic force through the half-wave rectified unidirectional pulse current, and cooperates with the first elastic sheet group 1-5 and the second elastic sheet group 1-2 to drive the feeding chute 1- 4. Do reciprocating vibration, so that the rice grains move forward directionally along the feeding trough 1-4, in this way, the process of automatic feeding can be realized without manual operation, which reduces the labor cost.

Optionally, please refer to FIG. 6 , the sorting module 3-1 includes N channels 3-1-1 and collecting troughs 3-1-3 arranged in parallel, where N is a positive integer, and the N channels 3- 1-1 are all V-shaped channels 3-1-1, each channel 3-1-1 includes a first side and a second side, and the end of the first side close to the feeding module 1 is provided with a first side. A blanking port 3-1-4, the end of the second side face close to the feeding module 1 is provided with a second blanking port, and the first blanking port 3-1-4 and the first blanking port The two blanking openings are both set away from the V-shaped bottom edge to form a reserved channel 3-1-1. The width of the first blanking opening 3-1-4 is smaller than the width of the first side surface. The width of the second blanking opening is smaller than the width of the second side surface, wherein the width of the reserved channel 3-1-1 is larger than the size of a single grain of rice and smaller than the size of a double-grained rice;

The collecting trough 3-1-3 is arranged below the first blanking port 3-1-4 and the second blanking port to collect the material collected from the first blanking port 3-1-4 and the second blanking port. the rice dropped from the second blanking port;

Moreover, one end of each channel 3-1-1 away from the feeding module 1 is provided with a first light-passing hole 3-1-2, and the first light-passing hole 3-1-2 is disposed opposite to the light source module 5 .

Optionally, referring to FIG. 7 , the light source module 5 includes a light guide strip 3-2 and a cold light source assembly 3-3, and the cold light source assembly 3-3 includes a light shield 3-3-2 and a The cold light source 3-3-1 inside the hood 3-3-2 is provided with a second light hole 3-3-3 on the hood 3-3-2. As shown in FIG. 8, the light guide One end of the strip 3-2 is connected to the first light-passing hole 3-1-2, and the other end of the light-guiding strip 3-2 is connected to the second light-passing hole 3-3-3. The schematic cross-sectional view A-A is the same as that in FIG. 7 .

In this embodiment, the sorting module 3-1 is installed at the front end of the feeding chute 1-4 of the feeding module 1. Specifically, the sorting module 3-1 is configured with three v-shaped channels 3-1-1, and the v-shaped channel 3-1 is configured with three v-shaped channels 3-1-1. -1 There are two blanking openings on both sides of the front end, and the two blanking openings are arranged in the direction away from the bottom edge of the V shape to form a reserved channel 3-1-1, which can ensure that the Only a single grain of rice can be arranged axially through the reserved channel 3-1-1 in turn, and the excess rice falls from the blanking port into the lower collecting trough 3-1-3, and each V-shaped channel 3-1-1 opens downward. There is a first light-passing hole 3-1-2; the light source module 5 is mainly composed of a cold light source 3-3-1 and an opaque hood 3-3-2. Two light holes 3-3-3 are connected to the first light hole 3-1-2 on the v-shaped channel 3-1-1 of the sorting module 3-1 and the light hood 3 on the light source module 5 through the light guide bar 3-2 - The second light hole 3-3-3 on 3-2, the light generated by the cold light source 3-3-1 goes down from the v-shaped channel 3-1-1 of the sorting module 3-1 along the light guide bar 3-2 The light aperture shines upwards.

Optionally, the above-mentioned paddy damage detection device further includes a cover plate 2, a bottom plate 6 and a support frame 7, one end of the support frame 7 is connected to the cover plate 2, and the other end of the support frame 7 is connected to the The bottom plate 6 is connected to form a closed cavity for accommodating the sorting module 3 - 1 and the light source module 5 . In this way, by forming a closed cavity, it can be ensured that the light emitted by the light source is not affected by external light when it irradiates the rice.

Optionally, the main control unit 8 includes a paddy detection subunit and a classification counting subunit, and the paddy detection subunit is used to identify the image, obtain the target frame corresponding to the paddy, and also be used to determine the target. The target frame coordinates and target frame type of the frame, the target frame coordinates are used to indicate the position of the paddy in the image, the target frame type is used to indicate the type of paddy in the image, and the paddy type includes internally damaged paddy or inner Undamaged rice;

The classification counting subunit is used for classifying and processing based on the target frame coordinates and the target frame type to obtain the internal damage of the paddy, where the internal damage of the paddy includes the quantity of internally damaged paddy, the quantity of internal non-damaged paddy and the damage rate.

In this embodiment, the rice detection sub-model can be the YOLOv4 model, which mainly includes three parts: BackBone, Neck, and Head. BackBone adopts a cross-stage local network CSPDarknet53, which performs 5 residuals after convolution processing on the input image. Network structure training; the Neck part includes the spatial pyramid pooling layer SPP and the path aggregation network PANe; the last layer of feature layers processed by CSPDarknet53 is imported into SPP for pooling and stacking after 3 convolution operations, and the processing results are processed by 3 After the second convolution operation, it is imported into PANet and CSPDarknet53 to obtain feature layer 1 and feature layer 2 for multiple upsampling and downsampling to complete feature fusion. Finally, the extracted features are used for prediction in the Head part, so as to obtain the image recognition result. In the embodiment, the recognition result of the image refers to identifying the type of paddy in the image, and the type of paddy refers to whether the paddy in the image is internally damaged paddy or internally non-damaged paddy. In this way, the rice type in the image can be quickly determined based on the rice detection sub-model.

Further, based on the sub-model of paddy detection and combined with the sub-model of classification and counting, the target recognition and detection of two types of paddy (inner damaged paddy and inner non-damaged paddy) can be realized at the same time to achieve classification and counting.

Next, the principle of the main control unit 8 detecting the internal damage of the rice based on the image information will be described in detail.

In an example, the rice detection sub-unit is provided with a rice detection sub-model, and the classification counting sub-unit is provided with a classification counting sub-model. First, the rice type in the image is determined based on the rice detection sub-model, and the rice type includes internally damaged rice. Or internal damage-free paddy; lock the paddy area in the image, generate the target frame corresponding to the paddy area, and determine the target frame coordinates and target frame type of the target frame. The target frame coordinates are used to indicate the position of the paddy in the image, and the target frame type It is used to indicate the type of rice in the image; the classification and counting sub-model is used to determine the detection result according to the coordinates and type of the target frame of the same rice in two adjacent frames of images.

Among them, the target frame can be a rectangular frame, and the determination method of the target frame is determined according to the size of the paddy and the position in the image. In other words, the target frame is used to frame the paddy in the image, and the target of the paddy in two adjacent frames Box coordinates and target box types can be used to reflect changes in the position of the rice.

In this way, the classification and counting sub-model is used to determine the detection result according to the target frame coordinates and target frame type of the same rice in two adjacent frames of images, which can quickly determine whether the change of the rice is in line with the normal situation, thereby avoiding the influence of errors in a certain frame of images. test results.

Specifically, the classification and counting sub-model can be used to determine the detection result according to the coordinates and type of the target frame of the same rice in two adjacent frames of images. The specific steps include: if the target frame of the rice in the current frame and the target frame in the previous frame The relationship between the target boxes satisfies the first preset condition, then it is recorded as the number of damaged paddy inside plus 1. If the relationship between the target frame in the current frame and the target frame in the previous frame satisfies the second preset If the relationship between the target frame in the current frame and the target frame in the previous frame satisfies the third preset condition, the count of the quantity of paddy remains unchanged.

In this way, by setting the first preset condition, the second preset condition and the third preset condition, the counting of different types of rice can be quickly realized.

Please refer to Figure 9, the direction of movement of the paddy is opposite to the direction in which the coordinate value increases, that is, the value of the target frame coordinate decreases frame by frame. YOLOv4 starts to recognize and detect from the left end of a grain of rice entering the light source until the right end of the paddy leaves the light source. Obtain several frames of rice images, identify the type of rice in them, and mark the type of the target detection frame. Because the rice moves from right to left, the target detection coordinates in the same obtained image of rice will decrease frame by frame, so compare the two frames before and after. The size of the target detection frame coordinates in the frame image. When the target detection frame coordinates in a certain frame image are greater than the previous frame, it means that the rice in the image appears to the right of the rice in the previous frame image, and it can be determined that a new grain of rice begins to pass through. In the detection area, at this time, according to the type of the target detection frame, the count of the corresponding type of rice is incremented by 1.

Wherein, the first preset condition includes:

If the paddy is internally damaged paddy, the value of the target frame coordinate of the paddy in the current frame is greater than the value of the target frame coordinate of the paddy in the previous frame; or, if the paddy is internally damaged paddy, the paddy is in the current frame. The value of the target frame coordinate in the frame is less than or equal to the value of the target frame coordinate of the paddy in the previous frame, and the target frame type of the paddy in the current frame is different from the target frame type of the paddy in the previous frame.

The second preset condition includes:

When the paddy is internally damaged paddy, the value of the target frame coordinate of the paddy in the current frame is greater than the value of the target frame coordinate of the paddy in the previous frame.

The third preset condition includes:

When the paddy is internally damaged paddy, the value of the target frame coordinate of the paddy in the current frame is less than or equal to the value of the target frame coordinate of the paddy in the previous frame, and the target frame type of the paddy in the current frame is the same as that of the paddy in the current frame. The target frame type in the previous frame is the same; or, if the paddy is internally undamaged paddy, the value of the target frame coordinate of the paddy in the current frame is less than or equal to the value of the target frame coordinate of the paddy in the previous frame.

Specifically, the classification and counting sub-model can also be used to determine the detection result according to the coordinates and type of the target frame of the same rice in two adjacent frames of images, and the specific steps include:

When the paddy is internally damaged paddy, if the value of the target frame coordinate of the paddy in the current frame is less than or equal to the value of the target frame coordinate of the paddy in the previous frame, and the target frame type of the paddy in the current frame is the same as that of the paddy. If the target box types in the previous frame are not the same, the number of damaged paddies inside is increased by 1, and the number of internal non-damaged paddies is decreased by 1.

It should be noted that there are a small number of internally damaged paddies that do not show damage characteristics in the first few frames. During the movement of the paddy, the type of the target frame will change from the internal damage type to the internal damage type. According to the above counting principle, there may be two types. Type paddy counts are incremented by 1. In order to avoid this problem, for the target frame with internal damage type whose coordinates are smaller than the previous frame, the comparison with the target frame type of the previous frame is added. The count of damaged paddy is increased by 1, and the count of internal non-damaged paddy is decreased by 1 (the first frame image of this grain of paddy is mistakenly increased by 1).

In an example, the detection process and counting can be monitored through the visualization window, and after the detection is completed, the number of internally damaged paddy and the number of internal non-damaged paddy, the total number of two types of paddy, and the ratio of internally damaged paddy are output; , and the pictures corresponding to each grain of rice can be classified and saved in the computer in real time, so as to facilitate the subsequent review of the detection results.

In other feasible embodiments, please refer to FIG. 10-FIG. 11, the working principle of the rice damage detection device provided by the present application is described by taking a complete work flow as an example. The hoppers 1-3 above are added, and the electromagnetic exciters 1-7 on the feeding module 1 generate periodic pulse electromagnetic force through the half-wave rectified unidirectional pulse current, which acts on the armature on the first shrapnel group 1-5 On 1-6, drive the first shrapnel group 1-5 to do reciprocating vibration, and the first shrapnel group 1-5 drives the feeding chute 1-4 to do reciprocating vibration, so that the rice grains move forward along the feeding chute 1-4; After reaching the sorting module 3-1, in each V-shaped channel 3-1-1, only a single grain of rice is axially lined up to pass in sequence, and the excess rice falls from the blanking port into the lower collecting trough 3-1-3; The visual sensing module collects in real time the image information of the rice after being irradiated by cold light in the three V-shaped channels 3-1-1, and transmits it to the main control unit 8 for detection. In each inspection process, the rice collected in the collecting trough 3-1-3 should be added to the hopper 1-3 of the feeding module 1 again, until all the rice is removed from the first place in the channel 3-1-1 of the sorting module 3-1. Through the light hole, the real-time acquisition of all images is completed.

After the visual sensing module transmits the image information of the rice in the detection area to the main control unit 8 in real time, the detection software system performs target detection and classification and counting on it. The detection software system of the present application consists of two parts: target detection and classification and counting. The deep learning target detection model YOLO V4 is used to perform real-time target detection on the collected and transmitted video, identify the type of rice in it (with internal damage, no internal damage), and mark the target frame to obtain the coordinate information of each target frame. Before applying for official use, collect samples, create a training set, and train and optimize the target detection model until the required detection accuracy is met. In order to realize the dynamic classification and counting of detection targets, the present application adds a classification and counting algorithm to the trained YOLOV4 target detection algorithm, so as to realize target recognition and detection of two types of paddy (internally damaged paddy and internally non-damaged paddy) and simultaneously classify and count.

The application also provides a method for detecting damage to rice, which is applied to the above-mentioned device for detecting damage to rice, and the method includes:

The feeding module 1 drives the paddy to move towards the direction of the sorting module 3-1;

The sorting module 3-1 arranges single grains of rice axially, and the sorting module 3-1 is provided with light holes;

The light source module 5 emits light to the light hole;

The visual sensor module 4 obtains the image of the rice grains passing through the sorting module 3-1 under the light;

The main control unit 8 detects the internal damage of the rice according to the image information.

The paddy damage detection method can implement the various embodiments of the above-mentioned paddy damage detection device, and can achieve the same beneficial effects, which will not be repeated here.

Embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned method steps are implemented. The readable storage medium can implement the various embodiments of the above-mentioned methods, and can achieve the same beneficial effects, which will not be repeated here.

The preferred specific embodiments of the present application are described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and changes based on the concept of the present application without creative efforts. Therefore, any technical solutions that can be obtained by those skilled in the art through logical analysis, reasoning or limited experiments on the basis of the prior art according to the concept of the present application shall fall within the protection scope determined by the claims.

Claims (8)

1. A paddy damage detection device, characterized in that it comprises: a feeding module, a sorting module, a light source module, a visual sensor module and a main control unit, the sorting module is connected with the feeding module, and the light source module is provided with Below the sorting module, the visual sensor is arranged above the sorting module, and the visual sensor and the light source module are located on the same vertical axis, and the main control unit is connected to the visual sensor ; The feeding module is used to drive the paddy to move towards the direction of the sorting module; The sorting module is used for single-grain axial queuing of the rice, and the sorting module is provided with a first light-passing hole; The light source module is used for emitting light to the light hole; The visual sensor module is used for acquiring the image of the rice grains passing through the sorting module under the light; The main control unit is used for detecting the internal damage of the rice according to the image information. 2 . The rice damage detection device according to claim 1 , wherein the feeding module comprises a base, a first shrapnel group, a second shrapnel group, an electromagnetic vibration exciter, a feeding trough, an armature and a hopper, 2 . One end of the first elastic sheet group is connected with the feeding chute, the other end of the first elastic sheet group is connected with the base, the second elastic sheet group is arranged opposite to the first elastic sheet group, and the second elastic sheet group is opposite to the first elastic sheet group. One end of the shrapnel group is connected to the feeding chute, the other end of the second shrapnel group is connected to the base, the armature is arranged on the first shrapnel group, and the electromagnetic exciter is arranged on the base and the electromagnetic vibration exciter is arranged facing the armature, the hopper is arranged on the material trough, and is located at one end of the material trough, and the other end of the material trough is connected to the sorting module. 3 . The paddy damage detection device according to claim 1 , wherein the sorting module comprises N channels and collecting troughs arranged in parallel, N is a positive integer, and the N channels are all V-shaped channels, 4 . Each channel includes a first side surface and a second side surface, and the end of the first side surface close to the feeding module is provided with a first blanking port, and the end of the second side surface close to the feeding module is provided with a first blanking port. There is a second blanking port, and both the first blanking port and the second blanking port are arranged in the direction away from the V-shaped bottom edge to form a reserved channel, and the width of the first blanking port is smaller than the width of the first blanking port. The width of the first side, the width of the second blanking port is less than the width of the second side, wherein, the width of the reserved channel is greater than the size of a single grain of rice and smaller than the size of two grains of rice; the collecting trough is arranged below the first blanking opening and the second blanking opening to collect the rice grains dropped from the first blanking opening and the second blanking opening; In addition, one end of each channel away from the feeding module is provided with the first light-passing hole, and the first light-passing hole is disposed opposite to the light source module. 4 . The rice damage detection device according to claim 3 , wherein the light source module comprises a light guide strip and a cold light source assembly, and the cold light source assembly comprises a light shield and a cold light source arranged inside the light shield. 5 . The light shield is provided with a second light-passing hole, one end of the light guide strip is connected to the first light-passing hole, and the other end of the light-guiding strip is connected to the second light-passing hole. 5 . The rice damage detection device according to claim 1 , further comprising a cover plate, a bottom plate and a support frame, one end of the support frame is connected to the cover plate, and the other end of the support frame is connected to the cover plate. 6 . The bottom plate is connected to form a closed cavity for accommodating the sorting module and the light source module. 6 . The paddy damage detection device according to claim 1 , wherein the main control unit comprises a paddy detection subunit and a classification counting subunit, and the paddy detection subunit is used to identify the image and obtain the The target frame corresponding to the rice is also used to determine the target frame coordinates and target frame type of the target frame, the target frame coordinates are used to indicate the position of the rice in the image, and the target frame type is used to indicate the rice in the image. type, the type of paddy includes internally damaged paddy or internally non-damaged paddy; The classification counting subunit is used for classifying and processing based on the target frame coordinates and the target frame type to obtain the internal damage of the paddy, where the internal damage of the paddy includes the quantity of internally damaged paddy, the quantity of internal non-damaged paddy and the damage rate. 7. A method for detecting damage to rice, applied to the device for detecting damage to rice according to any one of claims 1-6, wherein the method comprises: The feeding module drives the paddy to move towards the direction of the sorting module; The sorting module arranges single grains of rice axially, and the sorting module is provided with light holes; The light source module emits light to the light hole; The visual sensor module acquires the image of the rice grains passing through the sorting module under the light; The main control unit detects the internal damage of the rice according to the image information. 8 . The rice damage detection method according to claim 7 , wherein the main control unit comprises a rice detection subunit and a classification counting subunit, and the main control unit detects the internal damage of the rice according to the image information, comprising: 8 . : The rice detection subunit identifies the image, obtains the target frame corresponding to the rice, and determines the target frame coordinates and target frame type of the target frame, and the target frame coordinates are used to indicate the position of the rice in the image. The target box type is used to indicate the type of paddy in the image, and the paddy type includes inner damaged paddy or inner non-damaged paddy; The classification counting subunit performs classification processing based on the coordinates of the target frame and the type of the target frame to obtain the internal damage condition of the paddy, where the internal damage condition of the paddy includes the quantity of internally damaged paddy, the quantity of internal non-damaged paddy and the damage rate.

Images