Open Access

ARTICLE

An Intelligent Orchard Anti-Damage System Combining Real-Time AI Image Recognition and Laser-Based Deterrence for Multi-Target Monkeys

Shih-Ming Cho¹, Sung-Wen Wang¹, Min-Chie Chiu^2,*, Shao-Chun Chen¹

1 Department of Computer Science and Engineering, Tatung University, No.40, Sec. 3, Zhongshan N. Rd., Taipei City, 10452, Taiwan
2 Department of Mechanical and Materials Engineering, Tatung University, No. 40, Sec. 3, Zhongshan N. Rd., Taipei City, 10452, Taiwan

* Corresponding Author: Min-Chie Chiu. Email:

(This article belongs to the Special Issue: Development and Application of Deep Learning based Object Detection)

Computers, Materials & Continua 2026, 87(2), 37 https://doi.org/10.32604/cmc.2025.074911

Received 21 October 2025; Accepted 25 December 2025; Issue published 12 March 2026

Abstract

To address crop depredation by intelligent species (e.t, macaques) and the habituation from traditional methods, this study proposes an intelligent, closed-loop, adaptive laser deterrence system. A core contribution is an efficient multi-stage Semi-Supervised Learning (SSL) and incremental fine-tuning (IFT) framework, which reduced manual annotation by ~60% and training time by ~68%. This framework was benchmarked against YOLOv8n, v10n, and v11n. Our analysis revealed that YOLOv12n’s high Signal-to-Noise Ratio (SNR) (47.1% retention) pseudo-labels made it the only model to gain performance (+0.010 mAP) from SSL, allowing it to overtake competitors. Subsequently, in the IFT stress test, YOLOv12n proved most robust (a minimal −0.019 mAP decline), whereas YOLOv10n suffered catastrophic failure (−0.233 mAP), highlighting its incompatibility with IFT. The final model achieved high performance (mAP@0.5 of 0.947 for macaques, 0.946 for laser spots). In Multi-Object Tracking (MOT), this study quantitatively confirms that Bottom-Up Tracking by Sorting (BoT-SORT) (1.88 s avg. tracklet lifetime) significantly outperforms ByteTrack (0.81 s) in identity preservation for visually similar macaques. System integration achieved 480 Frames Per Second (FPS) real-time inference on edge devices. A quadratic polynomial fitting model ensured high-precision aiming (RMSE < 2 pixels; best 1.2 pixels) by compensating for distortion. To fundamentally solve habituation, an adaptive strategy driven by a Deep Deterministic Policy Gradient (DDPG) framework was introduced. By using a habituation penalty term (Rhabituation) to force unpredictable sequences, the DDPG strategy achieved a stable 88% average Intrusion Frequency Reduction Rate (IFRR) in field experiments, suppressing habituation in highly intelligent species. This study develops an efficient, precise, low-cost, and habituation-resistant automated wildlife defense system.

---

Keywords

---