UK develops AI laser to remove cockroaches, the number one pest in households

thumbnail

Cockroaches are the number one pest in the home environment. Once there are cockroaches in the room, they are very difficult to get rid of because of their strong reproduction. Recently, British scientists have improved a pest-killing AI (artificial intelligence) laser that can directly identify and kill cockroaches, providing a new option for people to get rid of cockroaches.

A research team from the UK's historic Heriot-Watt University described an AI laser costing less than $250 in the journal Oriental Insects in late September that could operate without drugs or chemicals , effectively kill cockroaches and other pests.

This AI insecticidal method uses a fully automated tracking system to identify indoor pests and kill them. In the experiment, they carried out experiments with German cockroaches and German cockroaches, and proved that the AI ​​laser can selectively destroy individual pests within a distance of 4 feet (1.2 meters).

The machine uses low-power lasers to irradiate those pests, making them feel too hot to hide in the dark, to expel them, or choose to eliminate them further.

They used German cockroaches and German cockroaches because they are among the leading pests in households worldwide. These pests not only contaminate food and damage appliances, but also endanger human health, including triggering allergies, asthma and gastrointestinal problems.

Although the AI ​​laser before the improvement can effectively kill mosquitoes, locusts and caterpillars, the disadvantage is that the laser power is too high, which may cause damage to human eyes and the environment. The improved laser has a better 3D scanning recognition system and CPU, which improves the machine's ability to identify objects and reduces the machine's energy consumption.

This new AI insecticide laser consists of a camera, a laser, a galvanometer (for changing the position of the laser light), a galvanometer and a single-board computer (computer), and is modified by programming with python 3.7.

The computer uses a quad-core CPU and two Sony IMX219 cameras to determine the insect's position in 3D space, and then uses computer calculations to hit the laser at the galvanometer and then attack the target, and the AI ​​programming can be modified through python The target (insect) to attack.

The two Sony IMX219 cameras are cameras with an 8-megapixel matrix. In order to implement the neural network operation, the machine uses real-time video with a resolution of 416 × 416 pixels to search for a single object (worm), which can report the location of the insect more accurately and faster.

In addition, the scientists made 1,000 images of cockroaches, and marked the position of each cockroach with a red frame on each image, and trained the AI ​​system to remember the images of these cockroaches, increasing the AI's neural network response speed.

The improved AI laser insecticide this time uses neural network algorithms to further drive the galvanometer, so that the reflection angle of the galvanometer can effectively control the direction of the laser beam, and effectively target and attack the moving cockroaches to achieve the goal of eliminating and repelling cockroaches. Effect. The current accuracy of AI laser insecticides is about 1 cm, and it is expected that the accuracy will be higher in the future.

Current pest control methods rely on chemical substances and pesticides, which can leave the chemical substances in the environment for a long time, which is not only limited in effect, but also harmful to human health and the living environment. In addition, those cockroaches that are not killed by the drug may develop stronger resistance to the drug over time, forcing people to use more poisonous drugs, creating a vicious circle.

However, the research team warned that the use of strong laser beams may seriously damage human eyes, so they will do further research and improvements so that it can really enter practical applications in the future.

Latest Programming News and Information | GeekBar

Related Posts