Automated mosquito vector system based on machine vision and deep learning

Abstract

This invention provides an automated mosquito vector control system based on machine vision and deep learning, comprising: a channel having an inlet end and an outlet end, the channel being tapered, and the inner diameter of the inlet end being larger than the inner diameter of the outlet end; an extraction device provided in the channel, the extraction device applying negative pressure at the inlet end to form an airflow, which outputs towards the outlet end; a base having an opening at one end corresponding to the outlet end, and having a plurality of chambers; a drive device corresponding to a rotating wheel, which is disposed in the accommodating space, the rotating wheel having a radially recessed conveying groove and a pushing block; an image capturing device disposed on the base and used to capture an image of an object to be identified towards the conveying groove; and a processing unit. The invention comprises an image capturing device equipped with a recognition model. This device captures mosquitoes using an extraction mechanism and a tapering channel design, allowing mosquitoes to be transported one by one into a conveying trough. An image capturing device captures images and classifies them using the recognition model. The transmission path is defined based on the category value, causing a drive device to drive a rotating wheel. This causes the conveying trough to align with the chamber corresponding to the category value, resulting in a sliding block that transports the mosquitoes to the corresponding category chamber. This allows for accurate and effective mosquito trapping, identification, and classification, facilitating the analysis of mosquito species in the environment. This enhances the reliability and effectiveness of intelligent epidemic prevention and can contribute to public health research and epidemic prevention decision-making.

Claims

1. An automated mosquito vector control system based on machine vision and deep learning, comprising: a channel having an inlet at one end and an outlet at the other end; the channel being provided with an extraction device that applies negative pressure at the inlet to generate airflow and outputs it toward the outlet; the channel being tapered, and the inner diameter of the inlet being larger than the inner diameter of the outlet; a base having an opening at one end corresponding to the outlet, and the base having a receiving space corresponding to the opening, and a plurality of chambers connected to the lower end of the receiving space; a drive device being correspondingly disposed at one end of the base; a rotating wheel being correspondingly disposed in the receiving space, and the drive device being correspondingly driven to rotate the rotating wheel; the rotating wheel having a radially recessed longitudinally arranged conveying groove, and a pushing block being slidably disposed within the conveying groove; and an image capturing device disposed on the base and used to capture an image of an object to be identified toward the conveying groove. A processing unit is correspondingly coupled to the driving device and the image capturing device. The processing unit is equipped with a recognition model for recognizing the image captured by the image capturing device and classifying the recognized object into a category value. The processing unit defines a transmission path corresponding to the driving device and the chamber based on the category value. Thus, when the conveying trough is located at the opening, the image capturing device captures an image. The processing unit recognizes the category value of the image through the recognition model and defines the transmission path to correspondingly drive the wheel, so that the conveying trough is located in the chamber corresponding to the category value, and the push block slides towards the corresponding chamber to push the conveyor.

2. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1, wherein, The extraction device is a bladeless fan, and a ring frame is formed at one end. A flow channel is formed inside the ring frame, and the inlet end is correspondingly located at the ring frame.

3. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1, wherein, The entrance end is further equipped with a mesh frame, which has multiple mesh openings.

4. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1 further includes a light-emitting unit, which is correspondingly arranged at the entrance end.

5. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1, wherein, The wheel is further provided with a groove corresponding to the conveying groove, and the push block is provided with a roller, which is correspondingly disposed in the groove.

6. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1 further includes a battery unit that is correspondingly coupled to the processing unit and correspondingly supplies power to the processing unit, the extraction device, the image capturing device and the driving device.

7. The automated mosquito vectoring system based on machine vision and deep learning as described in claim 6 further includes a solar panel that is correspondingly coupled to and charged to the battery unit.

8. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1, wherein, The base is formed by a recess on one side to create the accommodating space and the chamber; a cover is transparent and covers and shields the accommodating space, the chamber and the conveying trough.

9. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1, wherein, The object being identified is a mosquito.

10. The automated mosquito vector control system based on machine vision and deep learning as described in claim 1, wherein, The identification model is a YOLOv8 deep learning model.

Images