| CPC A01M 1/026 (2013.01) [A01M 1/08 (2013.01); G06T 7/246 (2017.01); G06T 2207/20081 (2013.01); G06V 10/774 (2022.01); G06V 40/10 (2022.01)] | 5 Claims |
|
1. An intelligent Forcipomyia taiwana monitoring and management system comprising:
a catching mechanism having a housing, a first opening and a negative pressure device, the first opening being disposed on one side of the housing to enable an inner 5 space of the housing to communicate externally through the first opening, the negative pressure device being disposed on the housing, so that an air pressure in the inner space being lower than an external air pressure to be capable of sucking a to-be-identified target into the inner space from an external environment through the first opening, wherein the to-be-identified target is a flying insect belonging to the order Diptera of the class Insecta;
a database storing a preset datum, the datum comprising at least a predetermined number of example pictures of at least one flying insect category;
a model training module using the example pictures to perform calculations to establish a training model;
an image capture module disposed on the housing for shooting an image including the to-be-identified target;
an identifying module selecting a first segmented region including the to-be-identified target from the image by using a You Only Look Once detection framework technology, extracting at least one first identification feature from the to-be-identified target in the first segmented region, and inputting the at least one first identification feature into the training model for deep learning of image identification in order to identify a flying insect category to which the to-be-identified target belongs and producing an identification result;
a counting module recording a number of the to-be-identified target included in the identification result into the database; and
a predictive tracking module obtaining a marked object based on the identification result marked with the to-be-identified target identified in the image, and using a Monte-Carlo category algorithm to track and predict the marked object, thereby reducing a misjudgment rate in a tracking process;
wherein the catching mechanism further comprises:
a partition disposed in the housing and dividing the inner space into a first chamber and a second chamber to enable the first chamber to communicate externally through the first opening;
a through hole penetratingly disposed on the partition to enable the first chamber to communicate with the second chamber;
a second opening corresponding to a position of the second chamber and penetratingly provided on the housing to enable the second chamber to communicate externally through the second opening;
a tapered first sleeve located in the first chamber, one end opening of the first sleeve is abutted against and connected to a position of the housing corresponding to the first opening to enable the first sleeve to communicate externally through the first opening, and an inner diameter of the first sleeve gradually decreases toward a direction of the second chamber;
a tapered second sleeve located in the first chamber, one end opening of the second sleeve is abutted against and connected to a position of the partition corresponding to the through hole to enable the second sleeve to communicate with the second chamber through the through hole, and an inner diameter of the second sleeve gradually increases toward a direction of the second chamber; and
a connecting pipe bridged between the first sleeve and the second sleeve, so that the first sleeve and the second sleeve communicate with each other, and an inner diameter of the connecting pipe is equal to a minimum inner diameter of the first sleeve or equal to a minimum inner diameter of the second sleeve.
|