Authors

Abstract

With the development of society and the advancement of science and technology, the object detection technology driven by Artificial Intelligence is also constantly innovating. As an important task in the field of agricultural Computer Vision, fruit detection in real environments faces many challenges. This paper reviews and analyzes the latest breakthroughs and representative studies in this field. Based on the existing research, the existing fruit detection algorithms are roughly divided into 5 categories: (1) the traditional fruit detection algorithm based on manual features; (2) the fruit detection algorithm based on two stages; (3) the fruit detection algorithm based on one stage; (4) the fruit detection algorithm based on anchor-free frame, and (5) the fruit detection algorithm based on transfer learning. This paper also discusses various application scenarios, such as multi-object detection, complex backgrounds, and edge computing. It also summarizes the classic and cutting-edge technical methods at the current time point, providing valuable insights for fruit detection in agricultural production.

Abstract in Malay

Dengan perkembangan masyarakat dan kemajuan sains dan teknologi, teknologi pengesanan objek yang dipacu oleh kecerdasan buatan juga sentiasa berinovasi. Sebagai tugas penting dalam bidang penglihatan komputer pertanian, pengesanan buah-buahan dalam persekitaran sebenar menghadapi banyak cabaran. Kertas kerja ini mengulas dan menganalisis penemuan terkini dan kajian-kajian yang mewakili bidang ini. Berdasarkan kajian, algoritma pengesanan buah-buahan sedia ada secara kasarnya dibahagikan kepada lima kategori, iaitu: (1) algoritma pengesanan buah-buahan tradisional berdasarkan ciri manual; (2) algoritma pengesanan buah-buahan berdasarkan dua peringkat; (3) algoritma pengesanan buah-buahan berdasarkan satu peringkat; (4) algoritma pengesanan buah-buahan berdasarkan tanpa sauh; dan (5) algoritma pengesanan buah-buahan berdasarkan pembelajaran pemindahan. Kertas kerja ini juga membincangkan pelbagai senario aplikasi seperti pengesanan berbilang objek, latar belakang kompleks dan pengkomputeran tepi. Ia juga meringkaskan kaedah teknikal klasik dan canggih pada masa kini, memberikan pandangan berharga untuk pengesanan buah-buahan dalam pengeluaran pertanian.