Development and evaluation of identification devices for vermin capture system

In the countryside of the world, wild monkeys have been appearing on farmland for several years, causing damage to crops. This area is a predominantly agricultural area. Damage caused by monkeys has a direct impact on crop production. Trap-based capture requires a detection sensor to activate the trap. Typically, detection sensors are realized with contact wire or infrared sensors, but they can only detect a single monkey. However, only one monkey could be detected. In the case of monkeys, it is necessary to capture the entire herd. This is because even if you capture one monkey, the other monkeys can't get close to the trap. Therefore, we are developing an artificial intelligence system to detect a herd of monkeys. This paper reports the results of the performance evaluation of an image recognizer with HOG feature extraction and a support vector machine. We created the discriminator with the cross-validation method. The performance of the HOG-SVM was compared the CNN discriminator. The results showed that there was little difference in performance between the HOG-SVM discriminator and the CNN discriminator. In general, the SVM discriminator is a small discrimination model. The vermin capture system is a stand-alone system that is installed in a remote field. The results show the possibility of embedding the discriminator in a micro-controller that controls the stand-alone system.